WO2019085754A1 - 应用清理方法、装置、存储介质及电子设备 - Google Patents

应用清理方法、装置、存储介质及电子设备 Download PDF

Info

Publication number
WO2019085754A1
WO2019085754A1 PCT/CN2018/110632 CN2018110632W WO2019085754A1 WO 2019085754 A1 WO2019085754 A1 WO 2019085754A1 CN 2018110632 W CN2018110632 W CN 2018110632W WO 2019085754 A1 WO2019085754 A1 WO 2019085754A1
Authority
WO
WIPO (PCT)
Prior art keywords
ridge regression
error
parameter
feature set
application
Prior art date
Application number
PCT/CN2018/110632
Other languages
English (en)
French (fr)
Inventor
曾元清
Original Assignee
Oppo广东移动通信有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Oppo广东移动通信有限公司 filed Critical Oppo广东移动通信有限公司
Publication of WO2019085754A1 publication Critical patent/WO2019085754A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F1/00Details not covered by groups G06F3/00 - G06F13/00 and G06F21/00
    • G06F1/26Power supply means, e.g. regulation thereof
    • G06F1/32Means for saving power
    • G06F1/3203Power management, i.e. event-based initiation of a power-saving mode
    • G06F1/3234Power saving characterised by the action undertaken
    • G06F1/329Power saving characterised by the action undertaken by task scheduling
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00Arrangements for program control, e.g. control units
    • G06F9/06Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46Multiprogramming arrangements
    • G06F9/50Allocation of resources, e.g. of the central processing unit [CPU]
    • G06F9/5005Allocation of resources, e.g. of the central processing unit [CPU] to service a request
    • G06F9/5011Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resources being hardware resources other than CPUs, Servers and Terminals
    • G06F9/5022Mechanisms to release resources
    • YGENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02DCLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00Energy efficient computing, e.g. low power processors, power management or thermal management

Definitions

  • the present invention relates to the field of electronic device communication technologies, and in particular, to an application cleaning method, device, storage medium, and electronic device.
  • the embodiment of the present application provides an application cleaning method, a device, a storage medium, and an electronic device, which can improve the running fluency of the electronic device and reduce power consumption.
  • an application cleaning method provided by the embodiment of the present application includes:
  • the ridge regression model is trained according to the applied training feature set, and the trained ridge regression model is obtained;
  • an application cleaning apparatus provided by the embodiment of the present application includes:
  • a training feature acquiring unit configured to acquire a multi-dimensional feature of the application, and obtain a training feature set of the application
  • a training unit configured to train the ridge regression model according to the applied training feature set, and obtain a trained ridge regression model
  • a prediction feature acquiring unit configured to acquire a multi-dimensional feature of the application, to obtain a predicted feature set of the application
  • a prediction unit configured to predict whether the application is cleanable according to the predicted feature set and the trained ridge regression model.
  • a storage medium provided by an embodiment of the present application has a computer program stored thereon, and when the computer program is run on a computer, the computer is caused to perform an application cleaning method according to any embodiment of the present application.
  • an electronic device provided by an embodiment of the present application includes a processor and a memory, where the memory has a computer program, and the processor is used to execute an application provided by any embodiment of the present application by calling the computer program. Cleaning method.
  • FIG. 1 is a schematic diagram of an application scenario of an application cleaning method according to an embodiment of the present disclosure.
  • FIG. 2 is a schematic flowchart of an application cleaning method provided by an embodiment of the present application.
  • FIG. 3 is another schematic flowchart of an application cleaning method provided by an embodiment of the present application.
  • FIG. 4 is a schematic structural diagram of an application cleaning device according to an embodiment of the present application.
  • FIG. 5 is another schematic structural diagram of an application cleaning device according to an embodiment of the present application.
  • FIG. 6 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
  • FIG. 7 is another schematic structural diagram of an electronic device according to an embodiment of the present application.
  • references to "an embodiment” herein mean that a particular feature, structure, or characteristic described in connection with the embodiments can be included in at least one embodiment of the present application.
  • the appearances of the phrases in various places in the specification are not necessarily referring to the same embodiments, and are not exclusive or alternative embodiments that are mutually exclusive. Those skilled in the art will understand and implicitly understand that the embodiments described herein can be combined with other embodiments.
  • An embodiment of the present application provides an application cleaning method, including:
  • the ridge regression model is trained according to the applied training feature set, and the trained ridge regression model is obtained;
  • the ridge regression model is trained according to the applied training samples to obtain a trained ridge regression model, including:
  • a trained ridge regression model is obtained according to the target ridge regression parameter and the ridge regression model.
  • acquiring a target ridge regression parameter of the ridge regression model according to the training feature set and the error determination function includes:
  • the ridge regression parameters including: a ridge parameter and a regression parameter;
  • Corresponding target ridge regression parameters are selected from the plurality of sets of ridge regression parameters according to the error corresponding to each set of ridge regression parameters.
  • acquiring a plurality of sets of ridge regression parameters according to the error determination function comprises:
  • regression parameters corresponding to each preset ridge parameter are obtained, and multiple sets of ridge regression parameters are obtained.
  • the error of the training feature set for the ridge regression model under the ridge regression parameter is obtained according to the training feature set, the ridge regression parameter, and the error judgment function, including:
  • the error of the training feature set for the ridge regression model under the ridge regression parameter is obtained according to a sub-error corresponding to each sub-training feature set, including:
  • the corresponding target ridge regression parameters are selected from the plurality of sets of ridge regression parameters according to the error corresponding to each set of ridge regression parameters, including:
  • a ridge regression parameter corresponding to the minimum error is selected from the plurality of sets of ridge regression parameters as a target ridge regression parameter.
  • obtaining an error of the training feature set for the ridge regression model under the ridge regression parameter according to the average error comprises:
  • the average error is directly taken as the error of the training feature set for the ridge regression model under the ridge regression parameter.
  • predicting whether the application is cleanable based on the predicted feature set and the trained ridge regression model comprises:
  • the embodiment of the present application provides an application cleaning method, which may be a background application cleaning device provided by an embodiment of the present application, or an electronic device integrated with the application cleaning device, where the application cleaning device may adopt hardware. Or software implementation.
  • the electronic device may be a device such as a smart phone, a tablet computer, a palmtop computer, a notebook computer, or a desktop computer.
  • FIG. 1 is a schematic diagram of an application scenario of an application cleaning method according to an embodiment of the present application.
  • the application device is integrated into an electronic device as an example, and the electronic device can acquire a multi-dimensional feature of the application and obtain a training feature set of the application.
  • the ridge regression model is trained to obtain the trained ridge regression model; the applied multi-dimensional features are obtained to obtain the applied prediction feature set; and the predicted feature set and the trained ridge regression model are used to predict whether the application can be used. Clean up.
  • electronic devices can be cleaned up with cleanable applications.
  • the multi-dimensional feature of the application a can be collected in a historical time period (for example, the application a is The duration of the background running, the time information of the application a, etc.), the feature set of the application a is obtained, and the ridge regression model is trained according to the feature set (for example, the time length of the application a running in the background, the time information of the application a running, etc.)
  • the ridge regression model after training; collecting the multi-dimensional features corresponding to the application according to the prediction time (such as t) (for example, the time length of the application a running in the background at time t, the time information of the application a running, etc.), and obtaining the predicted feature set of the application a; It is predicted whether the application a can be cleaned based on the predicted feature set and the trained ridge regression model.
  • the prediction time such as t
  • FIG. 2 is a schematic flowchart of an application cleaning method according to an embodiment of the present application.
  • the specific process of the application cleaning method provided by the embodiment of the present application may be as follows:
  • the application mentioned in the embodiment of the present application may be any application installed on the electronic device, such as an office application, a communication application, a game application, a shopping application, and the like.
  • the application may include an application running in the foreground, that is, a foreground application, and may also include an application running in the background, that is, a background application.
  • the application cleanup request may be received, the application to be cleaned is determined according to the application cleanup request, and then the multidimensional feature of the application is obtained, and the applied training feature set is obtained.
  • the multi-dimensional feature of the application may be obtained from the feature database, wherein the multi-dimensional feature may be a multi-dimensional feature collected by the historical time, that is, a historical multi-dimensional feature.
  • the feature database stores various features of the application at historical time.
  • the training feature set may include multi-dimensional features of the application, that is, multiple features of the application.
  • the multi-dimensional feature of the application has a dimension of a certain length, and the parameters in each dimension correspond to a feature information that represents the application, that is, the multi-dimensional feature is composed of multiple features.
  • the plurality of features may include feature information related to the application itself, for example, the duration of the application cutting into the background; the duration of the electronic device when the application is cut into the background; the number of times the application enters the foreground; the time when the application is in the foreground; the application is in the background The time, the way the application enters the background, such as being switched by the home button (home button), being switched back by the return button, being switched in by other applications, etc.; the type of application, including the first level (common application), the second level (other applications) )Wait.
  • the multi-dimensional feature information may further include related feature information of the electronic device where the application is located, for example, the time when the electronic device is off, the time of the bright screen, the current power, the wireless network connection status of the electronic device, whether the electronic device is in the charging state, or the like.
  • the applied training samples include multi-dimensional features of the application.
  • the multi-dimensional feature may be a plurality of features acquired at a preset frequency during a historical time period.
  • the historical time period may be, for example, the past 7 days or 10 days;
  • the preset frequency may be, for example, collected every 10 minutes and collected every half hour. It can be understood that the multi-dimensional feature data of the once collected application constitutes a training feature set.
  • the feature information that is not directly represented by the value in the multi-dimensional feature information of the application may be quantized by a specific value, for example, the feature information of the wireless network connection state of the electronic device may be used.
  • the value 1 indicates the normal state, and the value 0 indicates the abnormal state (or vice versa); for example, for the characteristic information of whether the electronic device is in the charging state, the value 1 indicates the state of charge, and the value 0 indicates the uncharged state ( The opposite is also possible).
  • the ridge regression model can be a machine learning algorithm, ridge regression (Tikhonov regularization) algorithm, also known as ridge regression is a biased estimation regression method dedicated to collinear data analysis, which is essentially an improved
  • the least squares estimation method by abandoning the unbiasedness of the least squares method, obtains a more realistic and reliable regression method with the regression coefficient at the cost of losing part of the information and reducing the accuracy, and fits the ill-conditioned data more strongly than the least squares method. .
  • the ridge regression model can be used to predict whether the application can be cleaned, wherein the output of the ridge regression model includes cleanable or non-cleanable.
  • the ridge regression model it is necessary to use the existing feature information to train the model to improve the accuracy of the prediction.
  • the process of training the ridge regression model is a process of solving the ridge regression parameter of the ridge regression model.
  • the ridge regression parameter required by the ridge regression model may be calculated first, and then the ridge is determined based on the ridge regression parameter.
  • the regression model is set.
  • the step "training the ridge regression model according to the applied training samples to obtain the post-training ridge regression model” may include:
  • the target ridge regression parameter of the ridge regression model is obtained according to the training feature set and the error judgment function, and the target ridge regression parameter includes a ridge parameter and a regression parameter;
  • the trained ridge regression model is obtained according to the target ridge regression parameter and the ridge regression model.
  • the ridge regression parameter may include a ridge parameter and a regression parameter
  • Ridge Regression is to add a regular term on the basis of the square error, and the balance between the variance and the deviation can be achieved by determining the value of ⁇ : with ⁇
  • the ridge parameter can normalize the parameter ⁇ , which can be the model parameter w of the ridge regression model to be solved.
  • the error judgment function is a loss function of the ridge regression model, and is used to calculate an error between the output value and the true value of the ridge regression model on the sample.
  • the error judgment function of the ridge regression model may include the following functions:
  • is the ridge parameter, ie the regularization parameter, x is the characteristic of the sample, w is the regression parameter of the ridge regression model, and n is the dimension of the feature.
  • the error judgment function of the ridge regression model may be deformed to obtain a regression parameter acquisition function, and then the ridge regression parameter is obtained based on the regression parameter acquisition function.
  • the error judgment function can be derived to obtain a regression parameter acquisition function, and then the ridge regression parameter is obtained based on the regression parameter acquisition function and the training feature set.
  • the error judgment function of the ridge regression model may include the following functions:
  • the error judgment function can be derived to obtain a function:
  • the regression parameter can be calculated based on the formula and the training feature set. Finally, the ridge parameter ⁇ and the corresponding regression parameters are obtained.
  • the step "acquiring the target ridge regression parameter of the ridge regression model according to the training feature set and the error judgment function" may include:
  • the ridge regression parameters include: ridge parameters and regression parameters;
  • the ridge regression parameter and the error judgment function the error of the training feature set for the ridge regression model under the ridge regression parameter is obtained, and the error corresponding to each set of ridge regression parameters is obtained;
  • the corresponding target ridge regression parameters are selected from the plurality of sets of ridge regression parameters
  • the trained ridge regression model is obtained according to the target ridge regression parameter and the ridge regression model.
  • the error corresponding to the ridge regression parameter is the ridge regression model under the ridge regression parameter, and the error between the predicted value and the true value obtained by inputting the training sample set.
  • m can be a positive integer greater than 2, which can be set according to actual needs, for example, 20, 30, 40...
  • the ridge regression parameter And an error judgment function to obtain regression parameters in the set
  • the errors corresponding to each set of ridge regression parameters are obtained, such as F1, F2, ..., Fk...Fm. Error F from the ridge regression parameters based on the regression parameters of each set of ridges Select the corresponding target ridge regression parameter
  • the error judgment function of the ridge regression model may be deformed to obtain a regression parameter acquisition function, and then the regression parameter is obtained based on the regression parameter acquisition function and the plurality of preset ridge parameters ⁇ .
  • Get multiple sets of ridge regression parameters For example, the error judgment function may be derived to obtain a regression parameter acquisition function, and then the ridge regression parameter is obtained based on the regression parameter acquisition function and the training feature set.
  • the error judgment function of the ridge regression model may include the following functions:
  • the error judgment function can be derived to obtain a function:
  • the regression parameter can be calculated based on the formula and a plurality of preset ridge parameters ⁇ . Finally, the ridge parameter ⁇ and the corresponding regression parameters are obtained.
  • the training feature set may be divided into a plurality of sub-training feature sets, and each sub-training feature set is obtained under the ridge regression parameter.
  • the error f of the ridge regression model is then obtained based on the error of the sub-training feature set for the ridge regression model under the ridge regression parameter to obtain the error F of the entire training feature set for the ridge regression model under the ridge regression parameter.
  • the step "acquiring the error of the training feature set to the ridge regression model under the ridge regression parameter according to the training feature set, the ridge regression parameter, and the error judgment function" may include:
  • the sub-training feature set the ridge regression parameter and the error judgment function, the sub-errors of the sub-training set for the ridge regression model under the ridge regression parameter are obtained, and the sub-error corresponding to each sub-training feature set is obtained;
  • the error of the training feature set for the ridge regression model under the ridge regression parameter is obtained.
  • the number of sub-training feature set partitions can be set according to actual needs, such as 10, 20, and so on.
  • the sub-training feature set includes the same number of features, that is, the training feature set is equally divided into a plurality of sub-training feature sets.
  • the training feature set D may be divided into M sub-training feature sets to obtain sub-training feature sets D1, D2, . . . DM; wherein M is a positive integer greater than one.
  • M is a positive integer greater than one.
  • the error of the training feature set in the m-group ridge regression parameter for the ridge regression model can be calculated, and the errors F1, F2, ... Fm are obtained.
  • the embodiment of the present application can obtain the error of the entire training feature set for the ridge regression model based on the sub-error, and the obtaining manner can be various.
  • the average value of the sub-errors may be calculated, and then the error of the entire training feature set for the ridge regression model is obtained based on the average value.
  • the step “acquiring the error of the training feature set to the ridge regression model under the ridge regression parameter according to the sub-error corresponding to each sub-training feature set” may include:
  • the error of the training feature set for the ridge regression model under the ridge regression parameter is obtained according to the average error.
  • the average error can be used as an error of the training feature set for the ridge regression model under the ridge regression parameter.
  • the ridge regression parameter is For example, in the calculation of the sub-errors f11, D2 for the ridge regression model under the ridges in the ridge regression parameters Sub-errors f12, ... DM in the ridge regression model
  • the sub-error f1M for the ridge regression model is based on the ridge regression parameter for each sub-training feature set.
  • the error F1 for the ridge regression model is For example, in the calculation of the sub-errors f11, D2 for the ridge regression model under the ridges in the ridge regression parameters.
  • the sub-error f1M for the ridge regression model is based on the ridge regression parameter for each sub-training feature set.
  • the average error f' (f11+f12+...+
  • the ridge regression parameter corresponding to the smallest error may be selected as the target ridge regression parameter of the ridge regression model, that is, finally parameter.
  • the ridge regression parameter corresponding to Fk can be selected.
  • the target ridge regression parameter as a ridge regression model.
  • the selection process of the target ridge regression parameters that is, the training process of the ridge regression model, will be introduced as follows: the ridge regression parameter is 20 groups and the number of sub-training feature sets is 10.
  • step A the following error judgment formula respectively calculates the different error values of each sub-feature set of 10 equal parts for the ridge regression, and obtains 10 different error values:
  • the 20 sets of characteristic errors obtained from (5) take the minimum value corresponding to And ⁇ value
  • the The ⁇ value is the ridge regression fitting to obtain the ridge regression parameters, that is, the parameters finally selected by the ridge regression model.
  • the ridge regression parameters corresponding to each application can be calculated by the above steps (1)-(6).
  • the multidimensional features of the application can be collected as prediction samples based on the predicted time.
  • the prediction time can be set according to requirements, such as the current time.
  • a multi-dimensional feature of an application can be acquired as a prediction sample at a predicted time point.
  • the multi-dimensional features collected in steps 201 and 203 are the same type of features, for example, the length of time when the application is cut into the background; the time when the application is cut into the background, the duration of the electronic device; the number of times the application enters the foreground; The time at the front desk; the way the app enters the background.
  • the probability that the application can be cleaned can be calculated based on the ridge regression model and the predicted feature set, and when the probability is greater than a certain threshold, the application can be cleaned and the like.
  • the embodiment of the present application obtains the multi-dimensional features of the application, and obtains the applied training feature set; the ridge regression model is trained according to the applied training feature set, and the trained ridge regression model is obtained; the multi-dimensional feature of the application is obtained, and the application is obtained.
  • the embodiment of the present application can make the cleaning of the corresponding application more personalized and intelligent.
  • the application of the cleanup prediction based on the ridge regression model can improve the accuracy of the user behavior prediction, thereby improving the accuracy of the cleanup.
  • multiple sets of ridge regression parameters can be calculated during the training of the model, and the ridge regression parameter with the lowest error of the feature error is used as the final parameter of the ridge regression model, and the ridge regression model can be further improved. The accuracy of the forecast for application cleanup.
  • the application cleaning method may include:
  • the multi-dimensional feature of the application is obtained from the feature database, wherein the multi-dimensional feature can be a multi-dimensional feature collected by the historical time, that is, a historical multi-dimensional feature.
  • the feature database stores various features of the application at historical time.
  • the training feature set may include multi-dimensional features of the application, that is, multiple features of the application.
  • the multi-dimensional feature of the application has a dimension of a certain length, and the parameters in each dimension correspond to a feature information that represents the application, that is, the multi-dimensional feature is composed of multiple features.
  • the plurality of features may include feature information related to the application itself, for example, the duration of the application cutting into the background; the duration of the electronic device when the application is cut into the background; the number of times the application enters the foreground; the time when the application is in the foreground; the application is in the background The time, the way the application enters the background, such as being switched by the home button (home button), being switched back by the return button, being switched in by other applications, etc.; the type of application, including the first level (common application), the second level (other applications) )Wait.
  • the multi-dimensional feature information may further include related feature information of the electronic device where the application is located, for example, the time when the electronic device is off, the time of the bright screen, the current power, the wireless network connection status of the electronic device, whether the electronic device is in the charging state, or the like.
  • a specific training feature set may be as follows, including feature information of multiple dimensions (30 dimensions). It should be noted that the feature information shown below is only an example. In practice, a feature included in a training feature set is included. The number of the information may be more than the number of the information shown below, or may be less than the number of the information shown below. The specific feature information may be different from the feature information shown below, and is not specifically limited herein.
  • the number of times the APP enters the foreground in the day (the rest day is divided by the working day and the rest day). For example, if the current forecasting time is the working day, the feature usage value is the average number of working days in the foreground for each working day.
  • the background APP is followed by the number of times the current foreground APP is opened, regardless of the workday rest day statistics;
  • the background APP is closely followed by the number of times the current foreground APP is opened, and is counted as a workday rest day;
  • the way in which the target APP is switched is divided into a home key switch, a recent key switch, and another APP switch;
  • the screen of the mobile phone is off;
  • the current screen is on and off
  • the feature indicates that the target app is being used every day from 8:00 to 12:00. Length of time used;
  • the current front-end APP enters the background to the target APP and enters the foreground according to the average interval of daily statistics;
  • the current foreground APP goes to the background until the target APP enters the foreground and the average screen is extinguished by daily statistics;
  • the target APP stays in the background for the first bin of the histogram (the proportion of times corresponding to 0-5 minutes);
  • the target APP stays in the background for the first bin of the histogram (5-10 minutes corresponds to the proportion of times);
  • the target APP stays in the background for the first bin of the histogram (10-15 minutes corresponds to the proportion of times);
  • the target APP stays in the background for the first bin of the histogram (15-20 minutes corresponds to the proportion of times);
  • the target APP stays in the background for the first bin of the histogram (15-20 minutes corresponds to the proportion of times);
  • the target APP stays in the background for the first bin of the histogram (25-30 minutes corresponds to the proportion of times);
  • the target APP stays in the background for the first bin of the histogram (the proportion of the corresponding number of times after 30 minutes);
  • the ridge regression model can be a machine learning algorithm, ridge regression (Tikhonov regularization) algorithm, also known as ridge regression is a biased estimation regression method dedicated to collinear data analysis, which is essentially an improved
  • the least squares estimation method by abandoning the unbiasedness of the least squares method, obtains a more realistic and reliable regression method with the regression coefficient at the cost of losing part of the information and reducing the accuracy, and fits the ill-conditioned data more strongly than the least squares method. .
  • the error judgment function of the ridge regression model may include the following functions:
  • is the ridge parameter, ie the regularization parameter, x is the characteristic of the sample, w is the regression parameter of the ridge regression model, and n is the dimension of the feature.
  • the error judgment function can be derived to obtain a function:
  • the ridge regression parameters include the ridge parameter ⁇ and the corresponding regression parameters
  • the number of sub-training feature set partitions can be set according to actual needs, such as 10, 20, and so on.
  • the sub-training feature set includes the same number of features, that is, the training feature set is equally divided into a plurality of sub-training feature sets.
  • the training feature set D may be divided into M sub-training feature sets to obtain sub-training feature sets D1, D2, . . . DM; wherein M is a positive integer greater than one.
  • M is a positive integer greater than one.
  • the sub-errors f11 and D2 for the ridge regression model are in the ridge regression parameters.
  • Sub-errors f12, ... DM in the ridge regression model For the sub-error f1M of the ridge regression model, the sub-training feature set is obtained in the ridge regression parameter.
  • each sub-training set for the ridge regression model under the ridge regression parameter obtains the error of the training feature set for the ridge regression model under the ridge regression parameter, and repeat steps 305 and 306 to obtain the training characteristics of each set of ridge regression parameters.
  • the error for the ridge regression model obtains the error of the training feature set for the ridge regression model under the ridge regression parameter, and repeat steps 305 and 306 to obtain the training characteristics of each set of ridge regression parameters.
  • the average error of the sub-training feature set is obtained; and the error of the training feature set for the ridge regression model under the ridge regression parameter is obtained according to the average error.
  • the average error can be used as an error of the training feature set for the ridge regression model under the ridge regression parameter.
  • the ridge regression parameter is For example, in the calculation of the sub-errors f11, D2 for the ridge regression model under the ridges in the ridge regression parameters Sub-errors f12, ... DM in the ridge regression model
  • the sub-error f1M for the ridge regression model is based on the ridge regression parameter for each sub-training feature set.
  • the error F1 for the ridge regression model is For example, in the calculation of the sub-errors f11, D2 for the ridge regression model under the ridges in the ridge regression parameters.
  • the sub-error f1M for the ridge regression model is based on the ridge regression parameter for each sub-training feature set.
  • the average error f' (f11+f12+...+
  • the ridge regression parameter corresponding to Fk can be selected.
  • the target ridge regression parameter as a ridge regression model.
  • the above steps 301-307 are repeated to obtain the ridge regression parameters corresponding to each application.
  • the value of the regression parameter w in the ridge regression model is updated.
  • repeating the above steps 301-308 can obtain a post-training ridge regression model corresponding to each application.
  • the prediction time can be set according to requirements, such as the current time.
  • a multi-dimensional feature of an application can be acquired as a prediction sample at a predicted time point.
  • the multi-dimensional feature collected in the step is the same type of feature as the feature acquired in step 301, that is, the predicted feature set and the feature set included in the training feature set are the same, for example, the time length of the application is cut into the background.
  • the time when the application is cut into the background the duration of the electronic device; the number of times the application enters the foreground; the time the application is in the foreground; and the way the application enters the background.
  • the probability that the application can be cleaned can be calculated based on the ridge regression model and the predicted feature set, and when the probability is greater than a certain threshold, the application can be cleaned and the like.
  • the post-training ridge regression model of each background application can be obtained through the above steps 301-308; then, based on the post-training ridge regression model of each background application, it is predicted whether multiple applications running in the background can be cleaned up, As shown in Table 1, it is determined that the application A1 and the application A3 running in the background can be cleaned, while the state in which the application A2 is running in the background is maintained.
  • the embodiment of the present application obtains the multi-dimensional features of the application, and obtains the applied training feature set; the ridge regression model is trained according to the applied training feature set, and the trained ridge regression model is obtained; the multi-dimensional feature of the application is obtained, and the application is obtained.
  • the embodiment of the present application can make the cleaning of the corresponding application more personalized and intelligent.
  • the application of the cleanup prediction based on the ridge regression model can improve the accuracy of the user behavior prediction, thereby improving the accuracy of the cleanup.
  • multiple sets of ridge regression parameters can be calculated during the training of the model, and the ridge regression parameter with the lowest error of the feature error is used as the final parameter of the ridge regression model, and the ridge regression model can be further improved. The accuracy of the forecast for application cleanup.
  • the embodiment of the present application further provides an application cleaning device, including:
  • a training feature acquiring unit configured to acquire a multi-dimensional feature of the application, and obtain a training feature set of the application
  • a training unit configured to train the ridge regression model according to the applied training feature set, and obtain a trained ridge regression model
  • a prediction feature acquiring unit configured to acquire a multi-dimensional feature of the application, to obtain a predicted feature set of the application
  • a prediction unit configured to predict whether the application is cleanable according to the predicted feature set and the trained ridge regression model.
  • the training unit includes:
  • a parameter obtaining subunit configured to acquire a target ridge regression parameter of the ridge regression model according to the training feature set and the error judgment function, where the target ridge regression parameter includes a ridge parameter and a regression parameter;
  • a training subunit configured to obtain a trained ridge regression model according to the target ridge regression parameter and the ridge regression model.
  • the parameter acquisition subunit is configured to:
  • the ridge regression parameters including: a ridge parameter and a regression parameter;
  • Corresponding target ridge regression parameters are selected from the plurality of sets of ridge regression parameters according to the error corresponding to each set of ridge regression parameters.
  • the parameter acquisition subunit is specifically configured to:
  • regression parameters corresponding to each preset ridge parameter are obtained, and multiple sets of ridge regression parameters are obtained.
  • the parameter acquisition subunit is specifically configured to:
  • the parameter acquisition subunit is specifically configured to:
  • the ridge regression parameters including: a ridge parameter and a regression parameter;
  • a ridge regression parameter corresponding to the minimum error is selected from the plurality of sets of ridge regression parameters as a target ridge regression parameter.
  • the parameter acquisition subunit is specifically configured to: directly use the average error as an error of the training feature set for the ridge regression model under the ridge regression parameter.
  • the predicting unit is configured to:
  • FIG. 4 is a schematic structural diagram of an application cleaning apparatus according to an embodiment of the present application.
  • the application cleaning device is applied to an electronic device, and the application cleaning device includes a training feature acquisition unit 401, a training unit 402, a prediction feature acquisition unit 403, and a prediction unit 404, as follows:
  • the training feature acquiring unit 401 is configured to acquire a multi-dimensional feature of the application, and obtain a training feature set of the application;
  • the training unit 402 is configured to train the ridge regression model according to the applied training feature set to obtain a trained ridge regression model
  • a prediction feature acquisition unit 403 configured to acquire a multi-dimensional feature of the application, to obtain a prediction feature set of the application;
  • the prediction unit 404 is configured to predict whether the application is cleanable according to the predicted feature set and the trained ridge regression model.
  • the training unit 402 includes:
  • a parameter obtaining sub-unit 4022 configured to acquire a target ridge regression parameter of the ridge regression model according to the training feature set and the error determination function, where the target ridge regression parameter includes a ridge parameter and a regression parameter;
  • the training sub-unit 4023 is configured to obtain a trained ridge regression model according to the target ridge regression parameter and the ridge regression model.
  • the parameter acquisition subunit 4022 can be used to:
  • the ridge regression parameters including: a ridge parameter and a regression parameter;
  • Corresponding target ridge regression parameters are selected from the plurality of sets of ridge regression parameters according to the error corresponding to each set of ridge regression parameters.
  • the parameter obtaining subunit 4022 may be specifically configured to:
  • regression parameters corresponding to each preset ridge parameter are obtained, and multiple sets of ridge regression parameters are obtained.
  • the parameter obtaining subunit 4022 may be specifically configured to:
  • the parameter obtaining subunit 4022 may be specifically configured to:
  • the parameter obtaining subunit 4022 may be specifically configured to:
  • a ridge regression parameter corresponding to the minimum error is selected from the plurality of sets of ridge regression parameters as a target ridge regression parameter.
  • the parameter obtaining subunit 4022 may be specifically configured to:
  • the parameter obtaining subunit 4022 may be specifically configured to:
  • the ridge regression parameters including: a ridge parameter and a regression parameter;
  • a ridge regression parameter corresponding to the minimum error is selected from the plurality of sets of ridge regression parameters as a target ridge regression parameter.
  • the parameter obtaining sub-unit 4022 may be specifically configured to: directly use the average error as an error of the training feature set for the ridge regression model under the ridge regression parameter.
  • the predicting unit 404 is configured to:
  • the steps performed by each unit in the application cleaning device may refer to the method steps described in the foregoing method embodiments.
  • the application cleaning device can be integrated in an electronic device such as a mobile phone, a tablet computer, or the like.
  • the foregoing various units may be implemented as an independent entity, and may be implemented in any combination, and may be implemented as the same entity or a plurality of entities.
  • the foregoing units refer to the foregoing embodiments, and details are not described herein again.
  • module unit
  • module may be taken to mean a software object that is executed on the computing system.
  • the different components, modules, engines, and services described herein can be considered as implementation objects on the computing system.
  • the apparatus and method described herein may be implemented in software, and may of course be implemented in hardware, all of which are within the scope of the present application.
  • the application cleaning device can obtain the multi-dimensional feature of the application by the training feature acquiring unit 401, and obtain the applied training feature set; the training unit 402 trains the ridge regression model according to the applied training feature set, and obtains the training.
  • the ridge regression model; the multi-dimensional feature of the application is obtained by the prediction feature acquisition unit 403 to obtain the applied prediction feature set; the prediction unit 404 predicts whether the application can be cleaned according to the predicted feature set and the trained ridge regression model;
  • the application is cleaned; the solution can automatically clean the application, improve the running fluency of the electronic device, reduce power consumption and save resources.
  • the electronic device 500 includes a processor 501 and a memory 502.
  • the processor 501 is electrically connected to the memory 502.
  • the processor 500 is a control center of the electronic device 500 that connects various portions of the entire electronic device using various interfaces and lines, by running or loading a computer program stored in the memory 502, and recalling data stored in the memory 502, The various functions of the electronic device 500 are performed and the data is processed to perform overall monitoring of the electronic device 500.
  • the memory 502 can be used to store software programs and modules, and the processor 501 executes various functional applications and data processing by running computer programs and modules stored in the memory 502.
  • the memory 502 may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, a computer program required for at least one function (such as a sound playing function, an image playing function, etc.), and the like; the storage data area may be stored according to Data created by the use of electronic devices, etc.
  • memory 502 can include high speed random access memory, and can also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid state storage device. Accordingly, memory 502 can also include a memory controller to provide processor 501 access to memory 502.
  • the processor 501 in the electronic device 500 loads the instructions corresponding to the process of one or more computer programs into the memory 502 according to the following steps, and is stored in the memory 502 by the processor 501.
  • the computer program in which to implement various functions, as follows:
  • the ridge regression model is trained according to the applied training feature set, and the trained ridge regression model is obtained;
  • the processor 501 may specifically perform the following steps:
  • a trained ridge regression model is obtained according to the target ridge regression parameter and the ridge regression model.
  • the processor 501 when acquiring the target ridge regression parameter of the ridge regression model according to the training feature set and the error judgment function, the processor 501 may specifically perform the following steps:
  • the ridge regression parameters including: a ridge parameter and a regression parameter;
  • Corresponding target ridge regression parameters are selected from the plurality of sets of ridge regression parameters according to the error corresponding to each set of ridge regression parameters.
  • the processor 501 may specifically perform the following steps:
  • regression parameters corresponding to each preset ridge parameter are obtained, and multiple sets of ridge regression parameters are obtained.
  • the processor 501 can specifically perform the following steps:
  • the processor 501 may specifically perform the following steps. :
  • the processor 501 may specifically perform the following steps:
  • a ridge regression parameter corresponding to the minimum error is selected from the plurality of sets of ridge regression parameters as a target ridge regression parameter.
  • the processor 501 when acquiring an error of the training feature set for the ridge regression model under the ridge regression parameter according to the average error, the processor 501 may specifically perform the following steps:
  • the average error is directly taken as the error of the training feature set for the ridge regression model under the ridge regression parameter.
  • the processor 501 may specifically perform the following steps:
  • the electronic device in the embodiment of the present application acquires the multi-dimensional feature of the application, obtains the applied training feature set, and trains the ridge regression model according to the applied training feature set to obtain the trained ridge regression model;
  • Feature obtain the applied feature set; according to the predicted feature set and the trained ridge regression model, predict whether the application can be cleaned; to clean up the cleanable application; the solution can automatically clean the application and improve the electronic device Smooth operation and reduced power consumption.
  • the electronic device 500 may further include: a display 503, a radio frequency circuit 504, an audio circuit 505, and a power source 506.
  • the display 503, the radio frequency circuit 504, the audio circuit 505, and the power source 506 are electrically connected to the processor 501, respectively.
  • the display 503 can be used to display information entered by a user or information provided to a user, as well as various graphical user interfaces, which can be composed of graphics, text, icons, video, and any combination thereof.
  • the display 503 can include a display panel.
  • the display panel can be configured in the form of a liquid crystal display (LCD) or an organic light-emitting diode (OLED).
  • LCD liquid crystal display
  • OLED organic light-emitting diode
  • the radio frequency circuit 504 can be used to transmit and receive radio frequency signals to establish wireless communication with a network device or other electronic device through wireless communication, and to transmit and receive signals with a network device or other electronic device.
  • the audio circuit 505 can be used to provide an audio interface between a user and an electronic device through a speaker or a microphone.
  • the power source 506 can be used to power various components of the electronic device 500.
  • the power source 506 can be logically coupled to the processor 501 through a power management system to enable functions such as managing charging, discharging, and power management through the power management system.
  • the electronic device 500 may further include a camera, a Bluetooth module, and the like, and details are not described herein again.
  • the embodiment of the present application further provides a storage medium, where the storage medium stores a computer program, and when the computer program runs on a computer, causes the computer to execute an application cleaning method in any of the above embodiments, such as: obtaining Applying the multi-dimensional characteristics, the applied training feature set is obtained; the ridge regression model is trained according to the applied training feature set, and the trained ridge regression model is obtained; the applied multi-dimensional features are obtained, and the applied predicted feature set is obtained; according to the predicted feature set And the trained ridge regression model to predict whether the application can be cleaned up.
  • an application cleaning method in any of the above embodiments, such as: obtaining Applying the multi-dimensional characteristics, the applied training feature set is obtained; the ridge regression model is trained according to the applied training feature set, and the trained ridge regression model is obtained; the applied multi-dimensional features are obtained, and the applied predicted feature set is obtained; according to the predicted feature set And the trained ridge regression model to predict whether the application can be cleaned up.
  • the storage medium may be a magnetic disk, an optical disk, a read only memory (ROM), or a random access memory (RAM).
  • ROM read only memory
  • RAM random access memory
  • the computer program may be stored in a computer readable storage medium, such as in a memory of the electronic device, and executed by at least one processor in the electronic device, and may include, for example, an application cleaning method during execution.
  • the storage medium may be a magnetic disk, an optical disk, a read only memory, a random access memory, or the like.
  • each functional module may be integrated into one processing chip, or each module may exist physically separately, or two or more modules may be integrated into one module.
  • the above integrated modules can be implemented in the form of hardware or in the form of software functional modules.
  • the integrated module if implemented in the form of a software functional module and sold or used as a standalone product, may also be stored in a computer readable storage medium, such as a read only memory, a magnetic disk or an optical disk, etc. .

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Image Analysis (AREA)

Abstract

本申请实施例公开了一种应用清理方法、装置、存储介质及电子设备,其中,本申请实施例获取应用的多维特征得到应用的训练特征集合;根据训练特征集合对岭回归模型进行训练,得到训练后的岭回归模型;获取应用的多维特征得到应用的预测特征集合;根据预测特征集合以及训练后的岭回归模型预测应用是否可清理。

Description

应用清理方法、装置、存储介质及电子设备
本申请要求于2017年10月31日提交中国专利局、申请号为201711050187.6、发明名称为“应用清理方法、装置、存储介质及电子设备”的中国专利申请的优先权,其全部内容通过引用结合在本申请中。
技术领域
本发明涉及电子设备通信技术领域,尤其涉及一种应用清理方法、装置、存储介质及电子设备。
背景技术
目前,智能手机等电子设备上,通常会有多个应用同时运行,其中,一个应用在前台运行,其他应用在后台运行。如果长时间不清理后台运行的应用,则会导致电子设备的可用内存变小、中央处理器(central processing unit,CPU)占用率过高,导致电子设备出现运行速度变慢,卡顿,耗电过快等问题。
发明内容
本申请实施例提供了一种应用清理方法、装置、存储介质及电子设备,能够提高电子设备的运行流畅度,降低功耗。
第一方面,本申请实施例了提供了的一种应用清理方法,包括:
获取应用的多维特征,得到所述应用的训练特征集合;
根据所述应用的训练特征集合对岭回归模型进行训练,得到训练后的岭回归模型;
获取所述应用的多维特征,得到所述应用的预测特征集合;
根据所述预测特征集合以及所述训练后的岭回归模型,预测所述应用是否可清理。
第二方面,本申请实施例了提供了的一种应用清理装置,包括:
训练特征获取单元,用于获取应用的多维特征,得到所述应用的训练特征集合;
训练单元,用于根据所述应用的训练特征集合对岭回归模型进行训练,得到训练后的岭回归模型;
预测特征获取单元,用于获取所述应用的多维特征,得到所述应用的预测特征集合;
预测单元,用于根据所述预测特征集合以及所述训练后的岭回归模型,预测所述应用是否可清理。
第三方面,本申请实施例提供的存储介质,其上存储有计算机程序,当所述计算机程序在计算机上运行时,使得所述计算机执行如本申请任一实施例提供的应用清理方法。
第四方面,本申请实施例提供的电子设备,包括处理器和存储器,所述存储器有计算机程序,所述处理器通过调用所述计算机程序,用于执行如本申请任一实施例提供的应用清理方法。
附图说明
为了更清楚地说明本申请实施例中的技术方案,下面将对实施例描述中所需要使用的附图作简单地介绍,显而易见地,下面描述中的附图仅仅是本申请的一些实施例,对于本领域技术人员来讲,在不付 出创造性劳动的前提下,还可以根据这些附图获得其他的附图。
图1为本申请实施例提供的应用清理方法的应用场景示意图。
图2是本申请实施例提供的应用清理方法的一个流程示意图。
图3是本申请实施例提供的应用清理方法的另一个流程示意图。
图4是本申请实施例提供的应用清理装置的一个结构示意图。
图5是本申请实施例提供的应用清理装置的另一结构示意图。
图6是本申请实施例提供的电子设备的一个结构示意图。
图7是本申请实施例提供的电子设备的另一结构示意图。
具体实施方式
请参照图式,本申请的原理是以实施在一适当的运算环境中来举例说明。以下的说明是基于所例示的本申请具体实施例,其不应被视为限制本申请未在此详述的其它具体实施例。
在本文中提及“实施例”意味着,结合实施例描述的特定特征、结构或特性可以包含在本申请的至少一个实施例中。在说明书中的各个位置出现该短语并不一定均是指相同的实施例,也不是与其它实施例互斥的独立的或备选的实施例。本领域技术人员显式地和隐式地理解的是,本文所描述的实施例可以与其它实施例相结合。
本申请实施例提供了一种应用清理方法,包括:
获取应用的多维特征,得到所述应用的训练特征集合;
根据所述应用的训练特征集合对岭回归模型进行训练,得到训练后的岭回归模型;
获取所述应用的多维特征,得到所述应用的预测特征集合;
根据所述预测特征集合以及所述训练后的岭回归模型,预测所述应用是否可清理。
在一些实施例中,根据所述应用的训练样本对岭回归模型进行训练,得到训练后的述岭回归模型,包括:
建立所述岭回归模型的误差判断函数;
根据所述训练特征集合和所述误差判断函数获取所述岭回归模型的目标岭回归参数,所述目标岭回归参数包括岭参数和回归参数;
根据所述目标岭回归参数和所述岭回归模型得到训练后的岭回归模型。
在一些实施例中,根据所述训练特征集合和所述误差判断函数获取所述岭回归模型的目标岭回归参数,包括:
根据所述误差判断函数获取多组岭回归参数,所述岭回归参数包括:岭参数和回归参数;
根据所述训练特征集合、所述岭回归参数和所述误差判断函数,获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差,得到每组岭回归参数对应的误差;
根据每组岭回归参数对应的误差,从所述多组岭回归参数中选取相应的目标岭回归参数。
在一些实施例中,根据所述误差判断函数获取多组岭回归参数,包括:
根据所述误差判断函数获取相应的回归参数获取函数;
根据多个预设岭参数以及所述回归参数获取函数,获取每个预设岭参数对应的回归参数,得到多组岭回归参数。
在一些实施例中,根据所述训练特征集合、所述岭回归参数和所述误差判断函数,获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差,包括:
将所述训练特征集合划分成多个子训练特征集合;
根据所述子训练特征集合、所述岭回归参数以及所述误差判断函数,获取在所述岭回归参数下所述子训练集合对于岭回归模型的子误差,得到每个子训练特征集合对应的所述子误差;
根据每个子训练特征集合对应的子误差,获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差。
在一些实施例中,根据每个子训练特征集合对应的子误差,获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差,包括:
根据每个子训练特征集合对应的子误差,获取子训练特征集合的平均误差;
根据所述平均误差获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差。
在一些实施例中,根据每组岭回归参数对应的误差,从所述多组岭回归参数中选取相应的目标岭回归参数,包括:
从每组岭回归参数对应的误差中确定最小误差;
从所述多组岭回归参数中选取所述最小误差对应的岭回归参数作为目标岭回归参数。
在一些实施例中,根据所述平均误差获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差,包括:
将所述平均误差直接作为在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差。
在一些实施例中,根据所述预测特征集合以及所述训练后的岭回归模型,预测所述应用是否可清理,包括:
根据所述预测特征集合以及所述训练后的岭回归模型计算出应用可清理的概率;
当所述概率大于预设阈值时,确定所述应用可清理。
本申请实施例提供一种应用清理方法,该应用清理方法的执行主体可以是本申请实施例提供的后台应用清理装置,或者集成了该应用清理装置的电子设备,其中该应用清理装置可以采用硬件或者软件的方式实现。其中,电子设备可以是智能手机、平板电脑、掌上电脑、笔记本电脑、或者台式电脑等设备。
请参阅图1,图1为本申请实施例提供的应用清理方法的应用场景示意图,以应用清理装置集成在电子设备中为例,电子设备可以获取应用的多维特征,得到应用的训练特征集合;根据应用的训练特征集合对岭回归模型进行训练,得到训练后的岭回归模型;获取应用的多维特征,得到应用的预测特征集合;根据预测特征集合以及训练后的岭回归模型,预测应用是否可清理。此外,电子设备还可以可清理的应用进行清理。
具体地,例如图1所示,以判断后台运行的应用程序a(如邮箱应用、游戏应用等)是否可以清理为例,可以在历史时间段内,采集应用a的多维特征(例如应用a在后台运行的时长、应用a运行的时 间信息等),得到应用a的特征集合,根据特征集合(例如应用a在后台运行的时长、应用a运行的时间信息等)对岭回归模型进行训练,得到训练后的岭回归模型;根据预测时间(如t)采集应用对应的多维特征(例如在t时刻应用a在后台运行的时长、应用a运行的时间信息等),得到应用a的预测特征集合;根据预测特征集合和训练后的岭回归模型预测应用a是否可清理。此外,当预测应用a可清理时,电子设备对应用a进行清理。
请参阅图2,图2为本申请实施例提供的应用清理方法的流程示意图。本申请实施例提供的应用清理方法的具体流程可以如下:
201、获取应用的多维特征,得到应用的训练特征集合。
本申请实施例所提及的应用,可以是电子设备上安装的任何一个应用,例如办公应用、通信应用、游戏应用、购物应用等。其中,应用可以包括前台运行的应用,即前台应用,也可以包括后台运行的应用,即后台应用。
在一实施例中,可以接收应用清理请求,根据应用清理请求确定待清理的应用,然后,获取应用的多维特征,得到应用的训练特征集合。
具体地,可以从特征数据库中获取应用的多维特征,其中,多维特征可以为历史时间采集到的多维特征,也即历史多维特征。特征数据库中存储有应用在历史时间的多种特征。
其中,训练特征集合可以包括应用的多维特征,即应用的多个特征。
其中,应用的多维特征具有一定长度的维度,其每个维度上的参数均对应表征应用的一种特征信息,即该多维特征息由多个特征构成。该多个特征可以包括应用自身相关的特征信息,例如:应用切入到后台的时长;应用切入到后台期间,电子设备的灭屏时长;应用进入前台的次数;应用处于前台的时间;应用处于后台的时间、应用进入后台的方式,例如被主页键(home键)切换进入、被返回键切换进入,被其他应用切换进入等;应用的类型,包括一级(常用应用)、二级(其他应用)等。
该多维特征信息还可以包括应用所在的电子设备的相关特征信息,例如:电子设备的灭屏时间、亮屏时间、当前电量,电子设备的无线网络连接状态,电子设备是否在充电状态等。
其中,应用的训练样本包括应用的多维特征。该多维特征可以是在历史时间段内,按照预设频率采集的多个特征。历史时间段,例如可以是过去7天、10天;预设频率,例如可以是每10分钟采集一次、每半小时采集一次。可以理解的是,一次采集的应用的多维特征数据构成一个训练特征集合。
在一实施例中,为便于应用清理,可以将应用的多维特征信息中,未用数值直接表示的特征信息用具体的数值量化出来,例如针对电子设备的无线网连接状态这个特征信息,可以用数值1表示正常的状态,用数值0表示异常的状态(反之亦可);再例如,针对电子设备是否在充电状态这个特征信息,可以用数值1表示充电状态,用数值0表示未充电状态(反之亦可)。
202、根据应用的训练特征集合对岭回归模型进行训练,得到训练后的岭回归模型。
其中,岭回归模型可以一种机器学习算法,岭回归(ridge regression,Tikhonov regularization)算法又称为脊回归是一种专用于共线性数据分析的有偏估计回归方法,实质上是一种改良的最小二乘估计法,通过放弃最小二乘法的无偏性,以损失部分信息、降低精度为代价获得回归系数更为符合实际、 更可靠的回归方法,对病态数据的拟合要强于最小二乘法。
本申请实施例中,可以利用岭回归模型来预测应用是否可清理,其中,岭回归模型的输出包括可清理、或不可清理。在利用岭回归模型预测应用是否可清理时,需要利用已有的特征信息对模型进行训练,提升预测的准确性。
在一实施例中,对岭回归模型训练的过程就是求解岭回归模型的岭回归参数的过程,比如,可以先计算出岭回归模型所需的岭回归参数,然后,基于该岭回归参数对岭回归模型进行设置。比如,步骤“根据应用的训练样本对岭回归模型进行训练,得到训练后的述岭回归模型”,可以包括:
建立岭回归模型的误差判断函数;
根据训练特征集合和误差判断函数获取岭回归模型的目标岭回归参数,目标岭回归参数包括岭参数和回归参数;
根据目标岭回归参数和岭回归模型得到训练后的岭回归模型。
其中,岭回归参数可以包括岭参数和回归参数,岭回归(Ridge Regression)是在平方误差的基础上增加正则项,通过确定λ的值可以使得在方差和偏差之间达到平衡:随着λ的增大,模型方差减小而偏差增大。岭参数可以正则化参数λ,该回归参数可以为待求解的岭回归模型的模型参数w。
本申请实施例中,误差判断函数为岭回归模型的损失函数,用于计算岭回归模型在样本上的输出值与真实值之间的误差。
在一实施例中,岭回归模型的误差判断函数可以包括如下函数:
Figure PCTCN2018110632-appb-000001
其中,λ为岭参数,即正则化参数,x为样本的特征,w为岭回归模型的回归参数,n为特征的维度。
在一实施例中,可以对岭回归模型的误差判断函数进行变形,得到回归参数获取函数,然后,基于回归参数获取函数来获取岭回归参数。比如,可以对误差判断函数进行求导,以得到回归参数获取函数,然后,基于该回归参数获取函数和训练特征集合获取岭回归参数。
例如,岭回归模型的误差判断函数可以包括如下函数:
Figure PCTCN2018110632-appb-000002
可以对误差判断函数进行求导,得到函数:
2X T(Y-XW)-2λW,X为特征x的矩阵或向量,X T为X的转置,Y为y的矩阵或向量;
然后,令2X T(Y-XW)-2λW等零,可以得到如下的回归参数计算公式:
Figure PCTCN2018110632-appb-000003
其中,
Figure PCTCN2018110632-appb-000004
为待求解的回归参数。
在得到回归参数计算公式后,便可以基于该公式和训练特征集合来计算回归参数
Figure PCTCN2018110632-appb-000005
最终得到岭参数λ以及相应的回归参数
Figure PCTCN2018110632-appb-000006
在一实施例中,为了提升预测准确性,可以计算出多组岭回归参数,然后,选取最合适的岭回归参数。比如,步骤“根据训练特征集合和误差判断函数获取岭回归模型的目标岭回归参数”,可以包括:
根据误差判断函数获取多组岭回归参数,岭回归参数包括:岭参数和回归参数;
根据训练特征集合、岭回归参数和误差判断函数,获取在岭回归参数下训练特征集合对于岭回归模型的误差,得到每组岭回归参数对应的误差;
根据每组岭回归参数对应的误差,从多组岭回归参数中选取相应的目标岭回归参数;
根据目标岭回归参数和岭回归模型得到训练后的岭回归模型。
其中,岭回归参数对应的误差为该岭回归参数下的岭回归模型,输入训练样本集合得出的预测值与真实值之间的误差。
比如,可以获取岭回归参数
Figure PCTCN2018110632-appb-000007
其中,m可以为大于2的正整数,可以根据实际需求设定,比如,20、30、40……。
然后,根据训练特征集合、岭回归参数
Figure PCTCN2018110632-appb-000008
以及误差判断函数,获取在该组回归参数
Figure PCTCN2018110632-appb-000009
Figure PCTCN2018110632-appb-000010
下训练特征集合对于岭回归模型的误差Fk,得到每组岭回归参数对应的误差如F1、F2……Fk……Fm。基于每组岭回归参数对应的误差F从岭回归参数
Figure PCTCN2018110632-appb-000011
选取相应的目标岭回归参数
Figure PCTCN2018110632-appb-000012
在一实施例中,可以对岭回归模型的误差判断函数进行变形,得到回归参数获取函数,然后,基于回归参数获取函数以及多个预设岭参数λ来获取回归参数
Figure PCTCN2018110632-appb-000013
得到多组岭回归参数
Figure PCTCN2018110632-appb-000014
比如, 可以对误差判断函数进行求导,以得到回归参数获取函数,然后,基于该回归参数获取函数和训练特征集合获取岭回归参数。
例如,岭回归模型的误差判断函数可以包括如下函数:
Figure PCTCN2018110632-appb-000015
可以对误差判断函数进行求导,得到函数:
2X T(Y-XW)-2λW,X为特征x的矩阵或向量,X T为X的转置,Y为y的矩阵或向量;
然后,令2X T(Y-XW)-2λW等零,可以得到如下的回归参数计算公式:
Figure PCTCN2018110632-appb-000016
其中,
Figure PCTCN2018110632-appb-000017
为待求解的回归参数。
在得到回归参数计算公式后,便可以基于该公式和多个预设岭参数λ来计算回归参数
Figure PCTCN2018110632-appb-000018
最终得到岭参数λ以及相应的回归参数
Figure PCTCN2018110632-appb-000019
例如,初始化岭参数λ的值为1,利用公式
Figure PCTCN2018110632-appb-000020
计算求得λ=1对应的
Figure PCTCN2018110632-appb-000021
值;λ加1,重复利用公式
Figure PCTCN2018110632-appb-000022
求得λ=2对应的
Figure PCTCN2018110632-appb-000023
值;λ再加1,重复利用公式
Figure PCTCN2018110632-appb-000024
求得λ=3对应的
Figure PCTCN2018110632-appb-000025
值……直到求得λ=m对应的
Figure PCTCN2018110632-appb-000026
值,比如m=20。此时,便可以得到m组如20组不同的
Figure PCTCN2018110632-appb-000027
值,进而得到m组如20组
Figure PCTCN2018110632-appb-000028
在一实施例中,为了能够准确性地和快速地获取训练特征集合对于岭回归模型的误差,可以将训练特征集合划分成多个子训练特征集合,获取各子训练特征集合在岭回归参数下对于岭回归模型的误差f,然后,基于各子训练特征集合在岭回归参数下对于岭回归模型的误差得到整个训练特征集合在岭回归参数下对于岭回归模型的误差F。
比如,步骤“根据训练特征集合、岭回归参数和误差判断函数,获取在岭回归参数下训练特征集合 对于岭回归模型的误差”,可以包括:
将训练特征集合划分成多个子训练特征集合;
根据子训练特征集合、岭回归参数以及误差判断函数,获取在岭回归参数下子训练集合对于岭回归模型的子误差,得到每个子训练特征集合对应的子误差;
根据每个子训练特征集合对应的子误差,获取在岭回归参数下训练特征集合对于岭回归模型的误差。
其中,子训练特征集合划分数量可以根据实际需求设定,比如10个、20个等等。在一实施例中,为提升误差获取的准确度,子训练特征集合包含的特征数量相等,也即将训练特征集合等分成多个子训练特征集合。
例如,可以将训练特征集合D划分成M个子训练特征集合,得到子训练特征集合D1、D2……DM;其中,M为大于1的正整数。然后,根据误差判断函数以及岭回归参数,计算每个子训练特征集合在岭回归参数下对于岭回归模型的子误差,如D1在岭回归参数
Figure PCTCN2018110632-appb-000029
下对于岭回归模型的子误差f11、D2在岭回归参数
Figure PCTCN2018110632-appb-000030
下对于岭回归模型的子误差f12、……DM在岭回归参数
Figure PCTCN2018110632-appb-000031
下对于岭回归模型的子误差f1M,基于每个子训练特征集合在该在岭回归参数
Figure PCTCN2018110632-appb-000032
下对于岭回归模型的子误差,即f11、f12……f1M获取训练特征D在岭回归参数
Figure PCTCN2018110632-appb-000033
下对于岭回归模型的误差F1。
接着,根据误差判断函数以及下一组岭回归参数,计算每个子训练特征集合在岭回归参数下对于下一组岭回归模型的子误差,如D1在岭回归参数
Figure PCTCN2018110632-appb-000034
下对于岭回归模型的子误差f21、D2在岭回归参数
Figure PCTCN2018110632-appb-000035
下对于岭回归模型的子误差f22、……DM在岭回归参数
Figure PCTCN2018110632-appb-000036
下对于岭回归模型的子误差f2M,基于每个子训练特征集合在该在岭回归参数
Figure PCTCN2018110632-appb-000037
下对于岭回归模型的子误差,即f21、f22……f2M获取训练特征D在岭回归参数
Figure PCTCN2018110632-appb-000038
下对于岭回归模型的误差F2。
依次类推,可以计算出训练特征集合在m组岭回归参数对于岭回归模型的误差,得到误差F1、F2……Fm。
本申请实施例在得到每个子训练特征集合对应的子误差后,便可以基于子误差获取整个训练特征集合对于岭回归模型的误差,该获取方式可以有多种。比如,在一实施例中,为了提升误差的准确性,可以计算子误差的平均值,然后,基于平均值获取整个训练特征集合对于岭回归模型的误差。比如,步骤“根据每个子训练特征集合对应的子误差,获取在岭回归参数下训练特征集合对于岭回归模型的误差”,可以包括:
根据每个子训练特征集合对应的子误差,获取子训练特征集合的平均误差;
根据平均误差获取在岭回归参数下训练特征集合对于岭回归模型的误差。
在一实施例中,可以将该平均误差作为在岭回归参数下训练特征集合对于岭回归模型的误差。
比如,以岭回归参数为
Figure PCTCN2018110632-appb-000039
为例,在计算出各子下对于岭回归模型的子误差f11、D2在岭回归参数
Figure PCTCN2018110632-appb-000040
下对于岭回归模型的子误差f12、……DM在岭回归参数
Figure PCTCN2018110632-appb-000041
下对于岭回归模型的子误差f1M,基于每个子训练特征集合在该在岭回归参数
Figure PCTCN2018110632-appb-000042
下对于岭回归模型的子误差,即f11、f12……f1M后,可以计算平均误差f’=(f11+f12+……+f1M)/M;该f’即为训练特征D在岭回归参数
Figure PCTCN2018110632-appb-000043
下对于岭回归模型的误差F1。
在一实施例中,为了提升参数准确性以及预测精确性,可以在得到每组岭回归参数对应的误差后,可以选取误差最小对应的岭回归参数作为岭回归模型的目标岭回归参数,即最终参数。
比如,在得到每组岭回归参数对应的误差如F1、F2……Fk……Fm后,假设Fk最小,此时,可以选取Fk对应的岭回归参数
Figure PCTCN2018110632-appb-000044
作为岭回归模型的目标岭回归参数。
根据上述描述,下面将以岭回归参数为20组、子训练特征集合数量为10来介绍目标岭回归参数的选取过程,也即岭回归模型的训练过程,如下:
(1)、建立岭回归的误差判断函数为:
Figure PCTCN2018110632-appb-000045
对进行求导,结果为:
2X T(Y-XW)-2λW
令其值为0可求得w的值为:
Figure PCTCN2018110632-appb-000046
(2)、初始化λ的值为1,按照(3)步骤中的
Figure PCTCN2018110632-appb-000047
公式计算求得相应的
Figure PCTCN2018110632-appb-000048
值。
(3)、λ加1,重复(2)步骤求得20组不同的
Figure PCTCN2018110632-appb-000049
值;
(4)、将特征集合分为10等分,选(3)步骤中
Figure PCTCN2018110632-appb-000050
的一个数值,下面的误差判断公式分别计算10等分的各个子特征集合对于岭回归的不同误差值,得到10个不同的误差值:
Figure PCTCN2018110632-appb-000051
然后计算将10等分各子特征集合对于岭回归的误差值的平均误差值,并将平均误差值作为特征集合在选取的
Figure PCTCN2018110632-appb-000052
和λ下对岭回归的误差;
(5)、重复(4)步骤,分别计算出特征集合在20组不同的
Figure PCTCN2018110632-appb-000053
值下对于岭回归的特征误差;
(6)、从(5)中求得的20组特征误差取最小值对应的
Figure PCTCN2018110632-appb-000054
和λ值,该
Figure PCTCN2018110632-appb-000055
和λ值即为岭回归拟合得到岭回归参数,即岭回归模型最终选用的参数。
通过上述步骤(1)-(6)可以计算出每个应用对应的岭回归参数。
203、获取应用的多维特征,得到应用的预测特征集合。
比如,可以根据预测时间采集应用的多维特征作为预测样本。
其中,预测时间可以根据需求设定,如可以为当前时间等。
比如,可以在预测时间点采集应用的多维特征作为预测样本。
本申请实施例中,步骤201和203中采集的多维特征是相同类型特征,例如:应用切入到后台的时长;应用切入到后台期间,电子设备的灭屏时长;应用进入前台的次数;应用处于前台的时间;应用进入后台的方式。
204、根据预测特征集合以及训练后的岭回归模型,预测应用是否可清理。
比如,可以基于岭回归模型和预测特征集合计算出应用可清理的概率,当概率大于某个阈值时,确定该应用可清理等等。
由上可知,本申请实施例获取应用的多维特征,得到应用的训练特征集合;根据应用的训练特征集合对岭回归模型进行训练,得到训练后的岭回归模型;获取应用的多维特征,得到应用的预测特征集合;根据预测特征集合以及训练后的岭回归模型,预测应用是否可清理;以便对可清理的应用进行清理;该方案可以实现应用的自动清理,提高了电子设备的运行流畅度,降低了功耗和节省了资源。
进一步地,由于特征集合中,包括了反映用户使用应用的行为习惯的多个特征信息,因此本申请实施例可以使得对对应应用的清理更加个性化和智能化。
进一步地,基于岭回归模型来实现应用清理预测,可以提升用户行为预测的准确性,进而提高清理的准确度。此外,本申请实施例在对模型训练时还可以计算出多组岭回归参数,并采用特征误差选取误差最下的岭回归参数,以作为岭回归模型的最终参数,可以进一步地提升岭回归模型对应用清理预测的准确性。
下面将在上述实施例描述的方法基础上,对本申请的清理方法做进一步介绍。参考图3,该应用清理方法可以包括:
301、获取应用的多维特征,得到应用的训练特征集合。
比如,从特征数据库中获取应用的多维特征,其中,多维特征可以为历史时间采集到的多维特征,也即历史多维特征。特征数据库中存储有应用在历史时间的多种特征。
其中,训练特征集合可以包括应用的多维特征,即应用的多个特征。
其中,应用的多维特征具有一定长度的维度,其每个维度上的参数均对应表征应用的一种特征信息,即该多维特征息由多个特征构成。该多个特征可以包括应用自身相关的特征信息,例如:应用切入到后台的时长;应用切入到后台期间,电子设备的灭屏时长;应用进入前台的次数;应用处于前台的时间;应用处于后台的时间、应用进入后台的方式,例如被主页键(home键)切换进入、被返回键切换进入,被其他应用切换进入等;应用的类型,包括一级(常用应用)、二级(其他应用)等。
该多维特征信息还可以包括应用所在的电子设备的相关特征信息,例如:电子设备的灭屏时间、亮屏时间、当前电量,电子设备的无线网络连接状态,电子设备是否在充电状态等。
一个具体的训练特征集合可如下所示,包括多个维度(30个维度)的特征信息,需要说明的是,如下所示的特征信息仅为举例,实际中,一个训练特征集合所包含的特征信息的数量,可以多于如下所示信息的数量,也可以少于如下所示信息的数量,所取的具体特征信息也可以与如下所示特征信息不同,此处不作具体限定。
APP上一次切入后台到现在的时长;
APP上一次切入后台到现在的期间中,累计屏幕关闭时间长度;
APP一天里(按每天统计)进入前台的次数;
APP一天里(休息日按工作日、休息日分开统计)进入前台的次数,比如若当前预测时间为工作日,则该特征使用数值为工作日统计到的平均每个工作日在前台使用次数;
APP一天中(按每天统计)处于前台的时间;
该后台APP紧跟当前前台APP后被打开次数,不分工作日休息日统计所得;
该后台APP紧跟当前前台APP后被打开次数,分工作日休息日统计;
目标APP被切换的方式,分为被home键切换、被recent键切换、被其他APP切换;
目标APP一级类型(常用应用);
目标APP二级类型(其他应用);
手机屏幕灭屏时间;
手机屏幕亮屏时间;
当前屏幕亮灭状态;
当前的电量;
当前wifi状态;
App上一次切入后台到现在的时长;
APP上一次在前台被使用时长;
APP上上一次在前台被使用时长;
APP上上上一次在前台被使用时长;
若一天分了6个时间段,每段4小时,当前预测时间点为早上8:30,则处于第3段,则该特征表示的是目标app每天在8:00-12:00这个时段被使用的时间长度;
当前前台APP进入后台到目标APP进入前台按每天统计的平均间隔时间;
当前前台APP进入后台到目标APP进入前台期间按每天统计的平均屏幕熄灭时间;
目标APP在后台停留时间直方图第一个bin(0-5分钟对应的次数占比);
目标APP在后台停留时间直方图第一个bin(5-10分钟对应的次数占比);
目标APP在后台停留时间直方图第一个bin(10-15分钟对应的次数占比);
目标APP在后台停留时间直方图第一个bin(15-20分钟对应的次数占比);
目标APP在后台停留时间直方图第一个bin(15-20分钟对应的次数占比);
目标APP在后台停留时间直方图第一个bin(25-30分钟对应的次数占比);
目标APP在后台停留时间直方图第一个bin(30分钟以后对应的次数占比);
当前是否有在充电。
302、建立岭回归模型的误差判断函数,根据误差判断函数获取相应的回归参数获取函数。
其中,岭回归模型可以一种机器学习算法,岭回归(ridge regression,Tikhonov regularization)算法又称为脊回归是一种专用于共线性数据分析的有偏估计回归方法,实质上是一种改良的最小二乘估 计法,通过放弃最小二乘法的无偏性,以损失部分信息、降低精度为代价获得回归系数更为符合实际、更可靠的回归方法,对病态数据的拟合要强于最小二乘法。
比如,岭回归模型的误差判断函数可以包括如下函数:
Figure PCTCN2018110632-appb-000056
其中,λ为岭参数,即正则化参数,x为样本的特征,w为岭回归模型的回归参数,n为特征的维度。
可以对误差判断函数进行求导,得到函数:
2X T(Y-XW)-2λW,X为特征x的矩阵或向量,X T为X的转置,Y为y的矩阵或向量;
然后,令2X T(Y-XW)-2λW等零,可以得到如下的回归参数计算公式:
Figure PCTCN2018110632-appb-000057
其中,
Figure PCTCN2018110632-appb-000058
为待求解的回归参数。
303、根据多个预设岭参数以及回归参数获取函数,获取相应的多个回归参数,得到多组岭回归参数。
其中,岭回归参数包括岭参数λ以及相应的回归参数
Figure PCTCN2018110632-appb-000059
例如,初始化岭参数λ的值为1,利用公式
Figure PCTCN2018110632-appb-000060
计算求得λ=1对应的
Figure PCTCN2018110632-appb-000061
值;λ加1,重复利用公式
Figure PCTCN2018110632-appb-000062
求得λ=2对应的
Figure PCTCN2018110632-appb-000063
值;λ再加1,重复利用公式
Figure PCTCN2018110632-appb-000064
求得λ=3对应的
Figure PCTCN2018110632-appb-000065
值……直到求得λ=m对应的
Figure PCTCN2018110632-appb-000066
值,比如m=20。此时,便可以得到m组如20组不同的
Figure PCTCN2018110632-appb-000067
值,进而得到m组如20组
Figure PCTCN2018110632-appb-000068
304、将训练特征集合划分成多个子训练特征集合。
其中,子训练特征集合划分数量可以根据实际需求设定,比如10个、20个等等。在一实施例中,为提升误差获取的准确度,子训练特征集合包含的特征数量相等,也即将训练特征集合等分成多个子训 练特征集合。
305、根据子训练特征集合、岭回归参数以及误差判断函数,获取在岭回归参数下子训练集合对于岭回归模型的子误差。
例如,可以将训练特征集合D划分成M个子训练特征集合,得到子训练特征集合D1、D2……DM;其中,M为大于1的正整数。然后,根据误差判断函数以及岭回归参数,计算每个子训练特征集合在岭回归参数下对于岭回归模型的子误差,如D1在岭回归参数
Figure PCTCN2018110632-appb-000069
下对于岭回归模型的子误差f11、D2在岭回归参数
Figure PCTCN2018110632-appb-000070
下对于岭回归模型的子误差f12、……DM在岭回归参数
Figure PCTCN2018110632-appb-000071
下对于岭回归模型的子误差f1M,得到每个子训练特征集合在该在岭回归参数
Figure PCTCN2018110632-appb-000072
下对于岭回归模型的子误差,f11、f12……f1M。
306、根据在岭回归参数下每个子训练集合对于岭回归模型的子误差,获取在岭回归参数下训练特征集合对于岭回归模型的误差,重复步骤305和306得到每组岭回归参数下训练特征对于岭回归模型的误差。
比如,根据每个子训练特征集合对应的子误差,获取子训练特征集合的平均误差;根据平均误差获取在岭回归参数下训练特征集合对于岭回归模型的误差。
在一实施例中,可以将该平均误差作为在岭回归参数下训练特征集合对于岭回归模型的误差。
比如,以岭回归参数为
Figure PCTCN2018110632-appb-000073
为例,在计算出各子下对于岭回归模型的子误差f11、D2在岭回归参数
Figure PCTCN2018110632-appb-000074
下对于岭回归模型的子误差f12、……DM在岭回归参数
Figure PCTCN2018110632-appb-000075
下对于岭回归模型的子误差f1M,基于每个子训练特征集合在该在岭回归参数
Figure PCTCN2018110632-appb-000076
下对于岭回归模型的子误差,即f11、f12……f1M后,可以计算平均误差f’=(f11+f12+……+f1M)/M;该f’即为训练特征D在岭回归参数
Figure PCTCN2018110632-appb-000077
下对于岭回归模型的误差F1。
接着重复步骤305和306可以计算出每组岭回归参数下训练特征集合对于岭回归模型的误差;如岭 回归参数
Figure PCTCN2018110632-appb-000078
分别对应的误差F1、F2……Fk……Fm。
307、选取误差最小对应的岭回归参数作为岭回归模型的目标岭回归参数。
比如,在得到每组岭回归参数对应的误差如F1、F2……Fk……Fm后,假设Fk最小,此时,可以选取Fk对应的岭回归参数
Figure PCTCN2018110632-appb-000079
作为岭回归模型的目标岭回归参数。
在一实施例中,重复上述步骤301-307可以得到每个应用对应的岭回归参数。
308、根据目标岭回归参数对岭回归模型中相应参数更新,得到训练后的岭回归模型。
比如,对岭回归模型中回归参数w的值进行更新。
在一实施例中,重复上述步骤301-308可以得到每个应用对应的训练后岭回归模型
309、获取应用的多维特征,得到应用的预测特征集合。
其中,预测时间可以根据需求设定,如可以为当前时间等。
比如,可以在预测时间点采集应用的多维特征作为预测样本。
本申请实施例中,该步骤采集的多维特征与步骤301中获取的特征是相同类型特征,也即预测特征集合与训练特征集合所包含的特征类型相同,例如均包括:应用切入到后台的时长;应用切入到后台期间,电子设备的灭屏时长;应用进入前台的次数;应用处于前台的时间;应用进入后台的方式。
310、根据预测特征集合以及训练后的岭回归模型,预测应用是否可清理。
比如,可以基于岭回归模型和预测特征集合计算出应用可清理的概率,当概率大于某个阈值时,确定该应用可清理等等。
在一个具体的例子中,可以通过上述步骤301-308获取每个后台应用的训练后岭回归模型;然后,基于每个后台应用的训练后岭回归模型预测后台运行的多个应用是否可清理,如表1所示,则确定可以清理后台运行的应用A1和应用A3,而保持应用A2在后台运行的状态不变。
应用 预测结果
应用A1 可清理
应用A2 不可清理
应用A3 可清理
表1
由上可知,本申请实施例获取应用的多维特征,得到应用的训练特征集合;根据应用的训练特征集合对岭回归模型进行训练,得到训练后的岭回归模型;获取应用的多维特征,得到应用的预测特征集合;根据预测特征集合以及训练后的岭回归模型,预测应用是否可清理;以便对可清理的应用进行清理;该 方案可以实现应用的自动清理,提高了电子设备的运行流畅度,降低了功耗和节省了资源。
进一步地,由于特征集合中,包括了反映用户使用应用的行为习惯的多个特征信息,因此本申请实施例可以使得对对应应用的清理更加个性化和智能化。
进一步地,基于岭回归模型来实现应用清理预测,可以提升用户行为预测的准确性,进而提高清理的准确度。此外,本申请实施例在对模型训练时还可以计算出多组岭回归参数,并采用特征误差选取误差最下的岭回归参数,以作为岭回归模型的最终参数,可以进一步地提升岭回归模型对应用清理预测的准确性。
本申请实施例还提供了一种应用清理装置,包括:
训练特征获取单元,用于获取应用的多维特征,得到所述应用的训练特征集合;
训练单元,用于根据所述应用的训练特征集合对岭回归模型进行训练,得到训练后的岭回归模型;
预测特征获取单元,用于获取所述应用的多维特征,得到所述应用的预测特征集合;
预测单元,用于根据所述预测特征集合以及所述训练后的岭回归模型,预测所述应用是否可清理。
在一些实施例中,所述训练单元,包括:
建立子单元,用于建立所述岭回归模型的误差判断函数;
参数获取子单元,用于根据所述训练特征集合和所述误差判断函数获取所述岭回归模型的目标岭回归参数,所述目标岭回归参数包括岭参数和回归参数;
训练子单元,用于根据所述目标岭回归参数和所述岭回归模型得到训练后的岭回归模型。
在一些实施例中,所述参数获取子单元,用于:
根据所述误差判断函数获取多组岭回归参数,所述岭回归参数包括:岭参数和回归参数;
根据所述训练特征集合、所述岭回归参数和所述误差判断函数,获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差,得到每组岭回归参数对应的误差;
根据每组岭回归参数对应的误差,从所述多组岭回归参数中选取相应的目标岭回归参数。
在一些实施例中,所述参数获取子单元,具体用于:
根据所述误差判断函数获取相应的回归参数获取函数;
根据多个预设岭参数以及所述回归参数获取函数,获取每个预设岭参数对应的回归参数,得到多组岭回归参数。
在一些实施例中,所述参数获取子单元,具体用于:
将所述训练特征集合划分成多个子训练特征集合;
根据所述子训练特征集合、所述岭回归参数以及所述误差判断函数,获取在所述岭回归参数下所述子训练集合对于岭回归模型的子误差,得到每个子训练特征集合对应的所述子误差;
根据每个子训练特征集合对应的子误差,获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差。
在一些实施例中,所述参数获取子单元,具体用于:
根据所述误差判断函数获取多组岭回归参数,所述岭回归参数包括:岭参数和回归参数;
根据所述训练特征集合、所述岭回归参数和所述误差判断函数,获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差,得到每组岭回归参数对应的误差;
从每组岭回归参数对应的误差中确定最小误差;
从所述多组岭回归参数中选取所述最小误差对应的岭回归参数作为目标岭回归参数。
在一些实施例中,所述参数获取子单元,具体用于:将所述平均误差直接作为在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差。
在一些实施例中,所述预测单元,用于:
根据所述预测特征集合以及所述训练后的岭回归模型计算出应用可清理的概率;
当所述概率大于预设阈值时,确定所述应用可清理。
在一实施例中还提供了一种应用清理装置。请参阅图4,图4为本申请实施例提供的应用清理装置的结构示意图。其中该应用清理装置应用于电子设备,该应用清理装置包括训练特征获取单元401、训练单元402、预测特征获取单元403、和预测单元404,如下:
训练特征获取单元401,用于获取应用的多维特征,得到应用的训练特征集合;
训练单元402,用于根据所述应用的训练特征集合对岭回归模型进行训练,得到训练后的岭回归模型;
预测特征获取单元403,用于获取所述应用的多维特征,得到所述应用的预测特征集合;
预测单元404,用于根据所述预测特征集合以及所述训练后的岭回归模型,预测所述应用是否可清理。
在一实施例中,参考图5,其中,训练单元402,包括:
建立子单元4021,用于建立所述岭回归模型的误差判断函数;
参数获取子单元4022,用于根据所述训练特征集合和所述误差判断函数获取所述岭回归模型的目标岭回归参数,所述目标岭回归参数包括岭参数和回归参数;
训练子单元4023,用于根据所述目标岭回归参数和所述岭回归模型得到训练后的岭回归模型。
在一实施例中,参数获取子单元4022,可以用于:
根据所述误差判断函数获取多组岭回归参数,所述岭回归参数包括:岭参数和回归参数;
根据所述训练特征集合、所述岭回归参数和所述误差判断函数,获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差,得到每组岭回归参数对应的误差;
根据每组岭回归参数对应的误差,从所述多组岭回归参数中选取相应的目标岭回归参数。
在一实施例中,参数获取子单元4022,可以具体用于:
根据所述误差判断函数获取相应的回归参数获取函数;
根据多个预设岭参数以及所述回归参数获取函数,获取每个预设岭参数对应的回归参数,得到多组岭回归参数。
在一实施例中,参数获取子单元4022,可以具体用于:
将所述训练特征集合划分成多个子训练特征集合;
根据所述子训练特征集合、所述岭回归参数以及所述误差判断函数,获取在所述岭回归参数下所述子训练集合对于岭回归模型的子误差,得到每个子训练特征集合对应的所述子误差;
根据每个子训练特征集合对应的子误差,获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差。
在一实施例中,参数获取子单元4022,可以具体用于:
根据每个子训练特征集合对应的子误差,获取子训练特征集合的平均误差;
根据所述平均误差获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差。
在一实施例中,参数获取子单元4022,可以具体用于:
从每组岭回归参数对应的误差中确定最小误差;
从所述多组岭回归参数中选取所述最小误差对应的岭回归参数作为目标岭回归参数。
在一实施例中,参数获取子单元4022,可以具体用于:
将所述训练特征集合划分成多个子训练特征集合;
根据所述子训练特征集合、所述岭回归参数以及所述误差判断函数,获取在所述岭回归参数下所述子训练集合对于岭回归模型的子误差,得到每个子训练特征集合对应的所述子误差;
根据每个子训练特征集合对应的子误差,获取子训练特征集合的平均误差;
根据所述平均误差获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差。
在一实施例中,参数获取子单元4022,可以具体用于:
根据所述误差判断函数获取多组岭回归参数,所述岭回归参数包括:岭参数和回归参数;
根据所述训练特征集合、所述岭回归参数和所述误差判断函数,获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差,得到每组岭回归参数对应的误差;
从每组岭回归参数对应的误差中确定最小误差;
从所述多组岭回归参数中选取所述最小误差对应的岭回归参数作为目标岭回归参数。
在一实施例中,参数获取子单元4022,可以具体用于:将所述平均误差直接作为在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差。
在一实施例中,所述预测单元404,用于:
根据所述预测特征集合以及所述训练后的岭回归模型计算出应用可清理的概率;
当所述概率大于预设阈值时,确定所述应用可清理。其中,应用清理装置中各单元执行的步骤可以参考上述方法实施例描述的方法步骤。该应用清理装置可以集成在电子设备中,如手机、平板电脑等。
具体实施时,以上各个单元可以作为独立的实体实现,也可以进行任意组合,作为同一或若干个实体来实现,以上各个单位的具体实施可参见前面的实施例,在此不再赘述。
本文所使用的术语“模块”“单元”可看做为在该运算***上执行的软件对象。本文所述的不同组件、模块、引擎及服务可看做为在该运算***上的实施对象。而本文所述的装置及方法可以以软件的方式进行实施,当然也可在硬件上进行实施,均在本申请保护范围之内。
由上可知,本实施例应用清理装置可以由训练特征获取单元401获取应用的多维特征,得到应用的 训练特征集合;由训练单元402根据应用的训练特征集合对岭回归模型进行训练,得到训练后的岭回归模型;由预测特征获取单元403获取应用的多维特征,得到应用的预测特征集合;由预测单元404根据预测特征集合以及训练后的岭回归模型,预测应用是否可清理;以便对可清理的应用进行清理;该方案可以实现应用的自动清理,提高了电子设备的运行流畅度,降低了功耗和节省了资源。
本申请实施例还提供一种电子设备。请参阅图6,电子设备500包括处理器501以及存储器502。其中,处理器501与存储器502电性连接。
所述处理器500是电子设备500的控制中心,利用各种接口和线路连接整个电子设备的各个部分,通过运行或加载存储在存储器502内的计算机程序,以及调用存储在存储器502内的数据,执行电子设备500的各种功能并处理数据,从而对电子设备500进行整体监控。
所述存储器502可用于存储软件程序以及模块,处理器501通过运行存储在存储器502的计算机程序以及模块,从而执行各种功能应用以及数据处理。存储器502可主要包括存储程序区和存储数据区,其中,存储程序区可存储操作***、至少一个功能所需的计算机程序(比如声音播放功能、图像播放功能等)等;存储数据区可存储根据电子设备的使用所创建的数据等。此外,存储器502可以包括高速随机存取存储器,还可以包括非易失性存储器,例如至少一个磁盘存储器件、闪存器件、或其他易失性固态存储器件。相应地,存储器502还可以包括存储器控制器,以提供处理器501对存储器502的访问。
在本申请实施例中,电子设备500中的处理器501会按照如下的步骤,将一个或一个以上的计算机程序的进程对应的指令加载到存储器502中,并由处理器501运行存储在存储器502中的计算机程序,从而实现各种功能,如下:
获取应用的多维特征,得到应用的训练特征集合;
根据所述应用的训练特征集合对岭回归模型进行训练,得到训练后的岭回归模型;
获取所述应用的多维特征,得到所述应用的预测特征集合;
根据所述预测特征集合以及所述训练后的岭回归模型,预测所述应用是否可清。
在某些实施方式中,在根据所述应用的训练样本对岭回归模型进行训练,得到训练后的述岭回归模型时,处理器501可以具体执行以下步骤:
建立所述岭回归模型的误差判断函数;
根据所述训练特征集合和所述误差判断函数获取所述岭回归模型的目标岭回归参数,所述目标岭回归参数包括岭参数和回归参数;
根据所述目标岭回归参数和所述岭回归模型得到训练后的岭回归模型。
在某些实施方式中,在根据所述训练特征集合和所述误差判断函数获取所述岭回归模型的目标岭回归参数时,处理器501可以具体执行以下步骤:
根据所述误差判断函数获取多组岭回归参数,所述岭回归参数包括:岭参数和回归参数;
根据所述训练特征集合、所述岭回归参数和所述误差判断函数,获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差,得到每组岭回归参数对应的误差;
根据每组岭回归参数对应的误差,从所述多组岭回归参数中选取相应的目标岭回归参数。
在某些实施方式中,在根据所述误差判断函数获取多组岭回归参数时,处理器501可以具体执行以下步骤:
根据所述误差判断函数获取相应的回归参数获取函数;
根据多个预设岭参数以及所述回归参数获取函数,获取每个预设岭参数对应的回归参数,得到多组岭回归参数。
在某些实施方式中,在根据所述训练特征集合、所述岭回归参数和所述误差判断函数,获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差时,处理器501可以具体执行以下步骤:
将所述训练特征集合划分成多个子训练特征集合;
根据所述子训练特征集合、所述岭回归参数以及所述误差判断函数,获取在所述岭回归参数下所述子训练集合对于岭回归模型的子误差,得到每个子训练特征集合对应的所述子误差;
根据每个子训练特征集合对应的子误差,获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差。
在某些实施方式中,在根据每个子训练特征集合对应的子误差,获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差时,处理器501可以具体执行以下步骤:
根据每个子训练特征集合对应的子误差,获取子训练特征集合的平均误差;
根据所述平均误差获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差。
在某些实施方式中,在根据每组岭回归参数对应的误差,从所述多组岭回归参数中选取相应的目标岭回归参数时,处理器501可以具体执行以下步骤:
从每组岭回归参数对应的误差中确定最小误差;
从所述多组岭回归参数中选取所述最小误差对应的岭回归参数作为目标岭回归参数。
在某些实施方式中,在根据所述平均误差获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差时,处理器501可以具体执行以下步骤:
将所述平均误差直接作为在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差。
在某些实施方式中,在根据所述预测特征集合以及所述训练后的岭回归模型,预测所述应用是否可清理时,处理器501可以具体执行以下步骤:
根据所述预测特征集合以及所述训练后的岭回归模型计算出应用可清理的概率;
当所述概率大于预设阈值时,确定所述应用可清理。
由上述可知,本申请实施例的电子设备,获取应用的多维特征,得到应用的训练特征集合;根据应用的训练特征集合对岭回归模型进行训练,得到训练后的岭回归模型;获取应用的多维特征,得到应用的预测特征集合;根据预测特征集合以及训练后的岭回归模型,预测应用是否可清理;以便对可清理的应用进行清理;该方案可以实现应用的自动清理,提高了电子设备的运行流畅度,降低了功耗。
请一并参阅图7,在某些实施方式中,电子设备500还可以包括:显示器503、射频电路504、音频电路505以及电源506。其中,其中,显示器503、射频电路504、音频电路505以及电源506分别与处理器501电性连接。
所述显示器503可以用于显示由用户输入的信息或提供给用户的信息以及各种图形用户接口,这些图形用户接口可以由图形、文本、图标、视频和其任意组合来构成。显示器503可以包括显示面板,在某些实施方式中,可以采用液晶显示器(Liquid Crystal Display,LCD)、或者有机发光二极管(Organic Light-Emitting Diode,OLED)等形式来配置显示面板。
所述射频电路504可以用于收发射频信号,以通过无线通信与网络设备或其他电子设备建立无线通讯,与网络设备或其他电子设备之间收发信号。
所述音频电路505可以用于通过扬声器、传声器提供用户与电子设备之间的音频接口。
所述电源506可以用于给电子设备500的各个部件供电。在一些实施例中,电源506可以通过电源管理***与处理器501逻辑相连,从而通过电源管理***实现管理充电、放电、以及功耗管理等功能。
尽管图7中未示出,电子设备500还可以包括摄像头、蓝牙模块等,在此不再赘述。
本申请实施例还提供一种存储介质,所述存储介质存储有计算机程序,当所述计算机程序在计算机上运行时,使得所述计算机执行上述任一实施例中的应用清理方法,比如:获取应用的多维特征,得到应用的训练特征集合;根据应用的训练特征集合对岭回归模型进行训练,得到训练后的岭回归模型;获取应用的多维特征,得到应用的预测特征集合;根据预测特征集合以及训练后的岭回归模型,预测应用是否可清理。
在本申请实施例中,存储介质可以是磁碟、光盘、只读存储器(Read Only Memory,ROM,)、或者随机存取记忆体(Random Access Memory,RAM)等。
在上述实施例中,对各个实施例的描述都各有侧重,某个实施例中没有详述的部分,可以参见其他实施例的相关描述。
需要说明的是,对本申请实施例的应用清理方法而言,本领域普通测试人员可以理解实现本申请实施例的应用清理方法的全部或部分流程,是可以通过计算机程序来控制相关的硬件来完成,所述计算机程序可存储于一计算机可读取存储介质中,如存储在电子设备的存储器中,并被该电子设备内的至少一个处理器执行,在执行过程中可包括如应用清理方法的实施例的流程。其中,所述的存储介质可为磁碟、光盘、只读存储器、随机存取记忆体等。
对本申请实施例的应用清理装置而言,其各功能模块可以集成在一个处理芯片中,也可以是各个模块单独物理存在,也可以两个或两个以上模块集成在一个模块中。上述集成的模块既可以采用硬件的形式实现,也可以采用软件功能模块的形式实现。所述集成的模块如果以软件功能模块的形式实现并作为独立的产品销售或使用时,也可以存储在一个计算机可读取存储介质中,所述存储介质譬如为只读存储器,磁盘或光盘等。
以上对本申请实施例所提供的一种应用清理方法、装置、存储介质及电子设备进行了详细介绍,本文中应用了具体个例对本申请的原理及实施方式进行了阐述,以上实施例的说明只是用于帮助理解本申请的方法及其核心思想;同时,对于本领域的技术人员,依据本申请的思想,在具体实施方式及应用范围上均会有改变之处,综上所述,本说明书内容不应理解为对本申请的限制。

Claims (20)

  1. 一种应用清理方法,其中,包括:
    获取应用的多维特征,得到所述应用的训练特征集合;
    根据所述应用的训练特征集合对岭回归模型进行训练,得到训练后的岭回归模型;
    获取所述应用的多维特征,得到所述应用的预测特征集合;
    根据所述预测特征集合以及所述训练后的岭回归模型,预测所述应用是否可清理。
  2. 如权利要求1所述的应用清理方法,其中,根据所述应用的训练样本对岭回归模型进行训练,得到训练后的述岭回归模型,包括:
    建立所述岭回归模型的误差判断函数;
    根据所述训练特征集合和所述误差判断函数获取所述岭回归模型的目标岭回归参数,所述目标岭回归参数包括岭参数和回归参数;
    根据所述目标岭回归参数和所述岭回归模型得到训练后的岭回归模型。
  3. 如权利要求2所述的应用清理方法,其中,根据所述训练特征集合和所述误差判断函数获取所述岭回归模型的目标岭回归参数,包括:
    根据所述误差判断函数获取多组岭回归参数,所述岭回归参数包括:岭参数和回归参数;
    根据所述训练特征集合、所述岭回归参数和所述误差判断函数,获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差,得到每组岭回归参数对应的误差;
    根据每组岭回归参数对应的误差,从所述多组岭回归参数中选取相应的目标岭回归参数。
  4. 如权利要求3所述的应用清理方法,其中,根据所述误差判断函数获取多组岭回归参数,包括:
    根据所述误差判断函数获取相应的回归参数获取函数;
    根据多个预设岭参数以及所述回归参数获取函数,获取每个预设岭参数对应的回归参数,得到多组岭回归参数。
  5. 如权利要求3所述的应用清理方法,其中,根据所述训练特征集合、所述岭回归参数和所述误差判断函数,获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差,包括:
    将所述训练特征集合划分成多个子训练特征集合;
    根据所述子训练特征集合、所述岭回归参数以及所述误差判断函数,获取在所述岭回归参数下所述子训练集合对于岭回归模型的子误差,得到每个子训练特征集合对应的所述子误差;
    根据每个子训练特征集合对应的子误差,获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差。
  6. 如权利要求5所述的应用清理方法,其中,根据每个子训练特征集合对应的子误差,获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差,包括:
    根据每个子训练特征集合对应的子误差,获取子训练特征集合的平均误差;
    根据所述平均误差获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差。
  7. 如权利要求3所述的应用清理方法,其中,根据每组岭回归参数对应的误差,从所述多组岭回归参数中选取相应的目标岭回归参数,包括:
    从每组岭回归参数对应的误差中确定最小误差;
    从所述多组岭回归参数中选取所述最小误差对应的岭回归参数作为目标岭回归参数。
  8. 如权利要求6所述的应用清理方法,其中,根据所述平均误差获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差,包括:
    将所述平均误差直接作为在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差。
  9. 如权利要求1所述的应用清理方法,其中,根据所述预测特征集合以及所述训练后的岭回归模型,预测所述应用是否可清理,包括:
    根据所述预测特征集合以及所述训练后的岭回归模型计算出应用可清理的概率;
    当所述概率大于预设阈值时,确定所述应用可清理。
  10. 一种应用清理装置,其中,包括:
    训练特征获取单元,用于获取应用的多维特征,得到所述应用的训练特征集合;
    训练单元,用于根据所述应用的训练特征集合对岭回归模型进行训练,得到训练后的岭回归模型;
    预测特征获取单元,用于获取所述应用的多维特征,得到所述应用的预测特征集合;
    预测单元,用于根据所述预测特征集合以及所述训练后的岭回归模型,预测所述应用是否可清理。
  11. 如权利要求10所述的应用清理装置,其中,所述训练单元,包括:
    建立子单元,用于建立所述岭回归模型的误差判断函数;
    参数获取子单元,用于根据所述训练特征集合和所述误差判断函数获取所述岭回归模型的目标岭回归参数,所述目标岭回归参数包括岭参数和回归参数;
    训练子单元,用于根据所述目标岭回归参数和所述岭回归模型得到训练后的岭回归模型。
  12. 如权利要求11所述的应用清理装置,其中,所述参数获取子单元,用于:
    根据所述误差判断函数获取多组岭回归参数,所述岭回归参数包括:岭参数和回归参数;
    根据所述训练特征集合、所述岭回归参数和所述误差判断函数,获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差,得到每组岭回归参数对应的误差;
    根据每组岭回归参数对应的误差,从所述多组岭回归参数中选取相应的目标岭回归参数。
  13. 如权利要求12所述的应用清理装置,其中,所述参数获取子单元,具体用于:
    根据所述误差判断函数获取相应的回归参数获取函数;
    根据多个预设岭参数以及所述回归参数获取函数,获取每个预设岭参数对应的回归参数,得到多组岭回归参数。
  14. 如权利要求12所述的应用清理装置,其中,所述参数获取子单元,具体用于:
    将所述训练特征集合划分成多个子训练特征集合;
    根据所述子训练特征集合、所述岭回归参数以及所述误差判断函数,获取在所述岭回归参数下所述子训练集合对于岭回归模型的子误差,得到每个子训练特征集合对应的所述子误差;
    根据每个子训练特征集合对应的子误差,获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差。
  15. 如权利要求14所述的应用清理装置,其中,所述参数获取子单元,具体用于:
    将所述训练特征集合划分成多个子训练特征集合;
    根据所述子训练特征集合、所述岭回归参数以及所述误差判断函数,获取在所述岭回归参数下所述子训练集合对于岭回归模型的子误差,得到每个子训练特征集合对应的所述子误差;
    根据每个子训练特征集合对应的子误差,获取子训练特征集合的平均误差;
    根据所述平均误差获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差。
  16. 如权利要求12所述的应用清理装置,其中,所述参数获取子单元,具体用于:
    根据所述误差判断函数获取多组岭回归参数,所述岭回归参数包括:岭参数和回归参数;
    根据所述训练特征集合、所述岭回归参数和所述误差判断函数,获取在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差,得到每组岭回归参数对应的误差;
    从每组岭回归参数对应的误差中确定最小误差;
    从所述多组岭回归参数中选取所述最小误差对应的岭回归参数作为目标岭回归参数。
  17. 如权利要求15所述的应用清理装置,其中,所述参数获取子单元,具体用于:将所述平均误差直接作为在所述岭回归参数下所述训练特征集合对于所述岭回归模型的误差。
  18. 如权利要求10所述的应用清理装置,其中,所述预测单元,用于:
    根据所述预测特征集合以及所述训练后的岭回归模型计算出应用可清理的概率;
    当所述概率大于预设阈值时,确定所述应用可清理。
  19. 一种存储介质,其上存储有计算机程序,其中,当所述计算机程序在计算机上运行时,使得所 述计算机执行如权利要求1至9任一项所述的应用清理方法。
  20. 一种电子设备,包括处理器和存储器,所述存储器有计算机程序,其中,所述处理器通过调用所述计算机程序,用于执行如权利要求1至9任一项所述的应用清理方法。
PCT/CN2018/110632 2017-10-31 2018-10-17 应用清理方法、装置、存储介质及电子设备 WO2019085754A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201711050187.6A CN107807730B (zh) 2017-10-31 2017-10-31 应用清理方法、装置、存储介质及电子设备
CN201711050187.6 2017-10-31

Publications (1)

Publication Number Publication Date
WO2019085754A1 true WO2019085754A1 (zh) 2019-05-09

Family

ID=61583133

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/110632 WO2019085754A1 (zh) 2017-10-31 2018-10-17 应用清理方法、装置、存储介质及电子设备

Country Status (2)

Country Link
CN (1) CN107807730B (zh)
WO (1) WO2019085754A1 (zh)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107807730B (zh) * 2017-10-31 2019-12-03 Oppo广东移动通信有限公司 应用清理方法、装置、存储介质及电子设备
CN111050385A (zh) * 2018-10-15 2020-04-21 中兴通讯股份有限公司 一种应用清理方法、装置、设备及可读存储介质
CN111797859A (zh) * 2019-04-09 2020-10-20 Oppo广东移动通信有限公司 特征处理方法、装置、存储介质及电子设备
CN113705107B (zh) * 2021-09-01 2023-09-08 桂林电子科技大学 一种基于均值岭回归的功耗分析方法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103902266A (zh) * 2012-12-26 2014-07-02 中兴通讯股份有限公司 清理后台运行应用的方法和装置
CN104298549A (zh) * 2014-09-30 2015-01-21 北京金山安全软件有限公司 移动终端中应用程序的清理方法、装置和移动终端
CN104866069A (zh) * 2015-06-12 2015-08-26 广东小天才科技有限公司 一种自动清理后台应用程序的方法及装置
CN107133094A (zh) * 2017-06-05 2017-09-05 努比亚技术有限公司 应用管理方法、移动终端及计算机可读存储介质
CN107807730A (zh) * 2017-10-31 2018-03-16 广东欧珀移动通信有限公司 应用清理方法、装置、存储介质及电子设备

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101763503B (zh) * 2009-12-30 2012-08-22 中国科学院计算技术研究所 一种姿态鲁棒的人脸识别方法
CN104159294B (zh) * 2014-08-01 2018-05-25 西南科技大学 一种基于蓝牙4.0技术的云定位平台
US9910481B2 (en) * 2015-02-13 2018-03-06 Intel Corporation Performing power management in a multicore processor
CN104899474B (zh) * 2015-06-09 2018-02-09 大连三生科技发展有限公司 基于岭回归矫正MB‑seq甲基化水平的方法及***
CN106709295A (zh) * 2016-12-08 2017-05-24 湖南文理学院 一种基于手指触屏信息分析的智能终端程序权限控制方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103902266A (zh) * 2012-12-26 2014-07-02 中兴通讯股份有限公司 清理后台运行应用的方法和装置
CN104298549A (zh) * 2014-09-30 2015-01-21 北京金山安全软件有限公司 移动终端中应用程序的清理方法、装置和移动终端
CN104866069A (zh) * 2015-06-12 2015-08-26 广东小天才科技有限公司 一种自动清理后台应用程序的方法及装置
CN107133094A (zh) * 2017-06-05 2017-09-05 努比亚技术有限公司 应用管理方法、移动终端及计算机可读存储介质
CN107807730A (zh) * 2017-10-31 2018-03-16 广东欧珀移动通信有限公司 应用清理方法、装置、存储介质及电子设备

Also Published As

Publication number Publication date
CN107807730B (zh) 2019-12-03
CN107807730A (zh) 2018-03-16

Similar Documents

Publication Publication Date Title
WO2019062417A1 (zh) 应用清理方法、装置、存储介质及电子设备
WO2019085754A1 (zh) 应用清理方法、装置、存储介质及电子设备
WO2019062414A1 (zh) 应用程序管控方法、装置、存储介质及电子设备
US10216558B1 (en) Predicting drive failures
EP3502878B1 (en) Method for preloading application and terminal device
WO2019062418A1 (zh) 应用清理方法、装置、存储介质及电子设备
CN107832132B (zh) 应用控制方法、装置、存储介质及电子设备
WO2019062342A9 (zh) 后台应用清理方法、装置、存储介质及电子设备
WO2019062460A1 (zh) 应用控制方法、装置、存储介质以及电子设备
CN107870810B (zh) 应用清理方法、装置、存储介质及电子设备
WO2019062405A1 (zh) 应用程序的处理方法、装置、存储介质及电子设备
JP2020537213A (ja) 端末の電力消費を低減するための方法、および端末
US11422831B2 (en) Application cleaning method, storage medium and electronic device
CN107943582B (zh) 特征处理方法、装置、存储介质及电子设备
WO2022161325A1 (zh) 提示方法和电子设备
WO2019062419A1 (zh) 应用清理方法、装置、存储介质及电子设备
CN107943571B (zh) 后台应用管控方法、装置、存储介质及电子设备
WO2019062462A1 (zh) 应用控制方法、装置、存储介质以及电子设备
CN107748697B (zh) 应用关闭方法、装置、存储介质及电子设备
WO2019062404A1 (zh) 应用程序的处理方法、装置、存储介质及电子设备
WO2019062341A1 (zh) 后台应用清理方法、装置、存储介质及电子设备
DE112018006292T5 (de) Batteriezustandsschätzung
WO2019062409A1 (zh) 后台应用程序管控方法、存储介质及电子设备
CN107797832B (zh) 应用清理方法、装置、存储介质及电子设备
WO2020133418A1 (zh) 应用图标显示方法、装置、存储介质及电子设备

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18873385

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18873385

Country of ref document: EP

Kind code of ref document: A1