CN116108732A

CN116108732A - Index time sequence prediction method, equipment, storage medium and product of application program

Info

Publication number: CN116108732A
Application number: CN202111320565.4A
Authority: CN
Inventors: 万明阳; 马国俊; 黄冶; 吴昊
Original assignee: Beijing Zitiao Network Technology Co Ltd
Current assignee: Beijing Zitiao Network Technology Co Ltd
Priority date: 2021-11-09
Filing date: 2021-11-09
Publication date: 2023-05-12

Abstract

The embodiment of the disclosure provides an index time sequence prediction method, equipment, a storage medium and a product of an application program, wherein the method, the equipment, the storage medium and the product are used for acquiring scene information of a target application scene and historical data of an index to be predicted of the application program in the target application scene at different moments in a past preset time window; constructing an input vector at any moment according to scene information of a target application scene and historical data of an index to be predicted at any moment; and inputting each input vector into a preset index time sequence prediction model according to time sequence to predict an index to be predicted of the application program in a target application scene, wherein the preset index time sequence prediction model is an index time sequence prediction model capable of performing time sequence prediction on various indexes of the application program in different application scenes. When predicting the index to be predicted in the target application scene, the scene information is also used as a part of characteristic variables of the model entry, and the universal index time sequence prediction model is used for time sequence prediction, so that the cost is reduced and the efficiency is improved.

Description

Index time sequence prediction method, equipment, storage medium and product of application program

Technical Field

The embodiment of the disclosure relates to the technical field of computer and network communication, in particular to an index time sequence prediction method, equipment, a storage medium and a product of an application program.

Background

The time sequence prediction is based on historical statistical data to predict future trend, and the prior art is divided into four modes, namely single variable time sequence prediction index, multiple variable time sequence prediction index and multiple variable time sequence prediction index.

In the prior art, no matter what way, the scene learned by the time sequence prediction model is unique, a model needs to be learned again for time sequence prediction of different scenes, for example, when time sequence prediction is performed on service indexes of an application program, for example, the number of active users on day, the number of users remaining on day and the like, because future trends of the service indexes of the application program are related to the application scene where the application program is located, for example, when the service of the application program relates to different areas, different operating systems and different product lines, the future trends of the service indexes are different, and the time sequence prediction model needs to be learned respectively under various application scenes.

The prior art needs to learn different time sequence prediction models for different application scenes, has higher cost and lower efficiency, and can not be efficiently solved when time sequence prediction is needed in more application scenes.

Disclosure of Invention

The embodiment of the disclosure provides an index time sequence prediction method, equipment, a storage medium and a product of an application program, which are used for performing time sequence prediction on various business indexes of the application program in different application scenes by adopting a universal index time sequence prediction model, so that the cost is reduced and the efficiency is improved.

In a first aspect, an embodiment of the present disclosure provides a method for predicting an index timing of an application program, including:

acquiring scene information of a target application scene of an application program, and historical data of to-be-predicted indexes of the application program in the target application scene at different moments in a preset time window in the past;

fusing and processing scene information of the target application scene and historical data of the index to be predicted at any moment to construct an input vector corresponding to the moment, wherein the input vector comprises a scene information part and an index historical data part;

and inputting each input vector into a preset index time sequence prediction model according to time sequence, and predicting the index to be predicted of the application program in a target application scene through the preset index time sequence prediction model, wherein the preset index time sequence prediction model is an index time sequence prediction model capable of performing time sequence prediction on various indexes of the application program in different application scenes.

In a second aspect, an embodiment of the present disclosure provides a training method of an index timing prediction model, including:

acquiring scene information of different application scenes of an application program and historical data of different indexes of the application program under the different application scenes;

acquiring an input vector set corresponding to different indexes of the application program in different application scenes according to scene information of the different application scenes and historical data of the different indexes of the application program in the different application scenes, wherein the input vector set comprises a plurality of input vectors arranged according to time sequence, and the input vector comprises a scene information part and an index historical data part;

training an initial index time sequence prediction model according to input vector sets corresponding to different indexes of an application program in different application scenes, so that the index time sequence prediction model can conduct time sequence prediction on various indexes of the application program in different application scenes.

In a third aspect, an embodiment of the present disclosure provides an index timing prediction apparatus of an application program, including:

the data acquisition unit is used for acquiring scene information of a target application scene of the application program and historical data of to-be-predicted indexes of the application program in the target application scene at different moments in a past preset time window;

The preprocessing unit is used for fusing and processing the scene information of the target application scene and the historical data of the index to be predicted at any moment to construct an input vector corresponding to the moment, wherein the input vector comprises a scene information part and an index historical data part;

the prediction unit is used for inputting each input vector into a preset index time sequence prediction model according to time sequence, and predicting the index to be predicted of the application program in a target application scene through the preset index time sequence prediction model, wherein the preset index time sequence prediction model is an index time sequence prediction model capable of performing time sequence prediction on various indexes of the application program in different application scenes.

In a fourth aspect, an embodiment of the present disclosure provides a training apparatus of an index timing prediction model, including:

the data acquisition unit is used for acquiring scene information of different application scenes of the application program and historical data of different indexes of the application program under the different application scenes;

the preprocessing unit is used for acquiring an input vector set corresponding to different indexes of the application program in different application scenes according to scene information of the different application scenes and historical data of the different indexes of the application program in the different application scenes, wherein the input vector set comprises a plurality of input vectors arranged according to time sequence, and the input vectors comprise scene information parts and index historical data parts;

The training unit is used for training an initial index time sequence prediction model according to input vector sets corresponding to different indexes of the application program in different application scenes, so that the index time sequence prediction model can conduct time sequence prediction on various indexes of the application program in different application scenes.

In a fifth aspect, embodiments of the present disclosure provide an electronic device, including: at least one processor and memory;

the memory stores computer-executable instructions;

the at least one processor executes the computer-executable instructions stored by the memory, causing the at least one processor to perform the method of index timing prediction for an application as described above in the first aspect and the various possible designs of the first aspect.

In a sixth aspect, embodiments of the present disclosure provide an electronic device, including: at least one processor and memory;

the memory stores computer-executable instructions;

the at least one processor executes the computer-executable instructions stored by the memory, causing the at least one processor to perform the training method of the indicator timing prediction model as described above in the second aspect and the various possible designs of the second aspect.

In a seventh aspect, embodiments of the present disclosure provide a computer readable storage medium having stored therein computer executable instructions that, when executed by a processor, implement the method for index timing prediction of an application according to the above first aspect and the various possible designs of the first aspect, and/or the method for training the index timing prediction model according to the above second aspect and the various possible designs of the second aspect.

In an eighth aspect, embodiments of the present disclosure provide a computer program product comprising computer-executable instructions which, when executed by a processor, implement the method for index timing prediction of an application as described in the first aspect and the various possible designs of the first aspect, and/or the method for training the index timing prediction model as described in the second aspect and the various possible designs of the second aspect.

According to the index time sequence prediction method, the device, the storage medium and the product of the application program, scene information of a target application scene of the application program is obtained, and historical data of indexes to be predicted of the application program in the target application scene at different moments in a past preset time window are obtained; fusing and processing scene information of a target application scene and historical data of an index to be predicted at any moment, and constructing an input vector corresponding to the moment, wherein the input vector comprises a scene information part and an index historical data part; and inputting each input vector into a preset index time sequence prediction model according to time sequence, and predicting the index to be predicted of the application program in a target application scene through the preset index time sequence prediction model, wherein the preset index time sequence prediction model is an index time sequence prediction model capable of performing time sequence prediction on various indexes of the application program in different application scenes. When predicting the index to be predicted in the target application scene, the scene information is also used as a part of characteristic variables of the model entry, so that the universal index time sequence prediction model can be used for time sequence prediction, the same universal index time sequence prediction model can be used for predicting various indexes of the application program in different application scenes, the cost is reduced, and the efficiency is improved.

When training the index time sequence prediction model, by acquiring historical data of different indexes under scene information of different application scenes of an application program, an input vector set of different indexes of the application program under different application scenes is constructed, so that training of multiple angles is performed on one index time sequence prediction model, the index time sequence prediction model can perform time sequence prediction on various indexes of the application program under different application scenes, the universality is high, and therefore the index time sequence prediction model does not need to be learned respectively for each application scene, the cost is reduced, and the efficiency is improved.

Drawings

In order to more clearly illustrate the embodiments of the present disclosure or the solutions in the prior art, a brief description will be given below of the drawings that are needed in the embodiments or the description of the prior art, it being obvious that the drawings in the following description are some embodiments of the present disclosure, and that other drawings may be obtained from these drawings without inventive effort to a person of ordinary skill in the art.

Fig. 1 is an application scenario schematic diagram of an index timing prediction method of an application program according to an embodiment of the disclosure;

FIG. 2 is a flowchart illustrating a method for predicting an index timing of an application according to an embodiment of the disclosure;

FIG. 3 is a schematic diagram of a process for creating an input vector according to one embodiment of the present disclosure;

FIG. 4 is a schematic diagram of an indicator timing prediction model according to an embodiment of the disclosure;

FIG. 5 is a flowchart illustrating a method for index timing prediction of an application according to another embodiment of the present disclosure;

FIG. 6 is a flowchart of a training method of an index timing prediction model according to an embodiment of the disclosure;

FIG. 7 is a block diagram of an indicator timing prediction apparatus for an application according to an embodiment of the present disclosure;

FIG. 8 is a block diagram of a training apparatus for an index timing prediction model according to an embodiment of the present disclosure;

fig. 9 is a schematic hardware structure of an electronic device according to an embodiment of the disclosure.

Detailed Description

For the purposes of making the objects, technical solutions and advantages of the embodiments of the present disclosure more apparent, the technical solutions of the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present disclosure, and it is apparent that the described embodiments are some embodiments of the present disclosure, but not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art without inventive effort, based on the embodiments in this disclosure are intended to be within the scope of this disclosure.

In any way, the scene learned by the time sequence prediction model is unique, and when time sequence prediction is performed on service indexes of an application program, for example, the number of active users newly added, the number of remaining users next day and the like, a model is needed to be learned again for time sequence prediction of different scenes, for example, when the service of the application program relates to different areas, different operating systems and different product lines, the future trend of the service index is different, and the time sequence prediction model needs to be learned respectively under various application scenes.

In order to solve the above technical problems, in the embodiments of the present disclosure, by training a general index timing prediction model, scene information of an application scene is also used as a part of features (variables) of a model entry, so that the index timing prediction model can perform timing prediction on various indexes of an application program in different application scenes, so that it is not necessary to learn an index timing prediction model for each scene, thereby reducing cost and improving efficiency.

Specifically, when the application model predicts the time sequence of the application program index, the scene information of the target application scene of the application program and the historical data of the target application scene to be predicted of the application program at different moments in a past preset time window can be obtained; fusing and processing scene information of the target application scene and historical data of the index to be predicted at any moment to construct an input vector corresponding to the moment, wherein the input vector comprises a scene information part and an index historical data part; and inputting each input vector into a preset index time sequence prediction model according to time sequence, and predicting the index to be predicted of the application program in a target application scene through the preset index time sequence prediction model, wherein the preset index time sequence prediction model is an index time sequence prediction model capable of performing time sequence prediction on various indexes of the application program in different application scenes.

The method for predicting the index time sequence of the application program provided in this embodiment is suitable for a system as shown in fig. 1, and includes a database 101 and a processing device 102, where the database 101 is configured to provide historical data of indexes to be predicted of the application program in a target application scenario at different moments in the past preset time window, and the processing device 102 may adopt the method for predicting the index time sequence of the application program, and obtain a predicted value of the indexes to be predicted in the target application scenario based on a preset index time sequence prediction model.

The following description of the technical solutions in the embodiments of the present disclosure will be made clearly and completely with reference to the accompanying drawings in the embodiments of the present disclosure, and it is apparent that the described embodiments are only some embodiments of the present disclosure, and not all embodiments. Based on the embodiments in this disclosure, all other embodiments that a person of ordinary skill in the art would obtain without making any inventive effort are within the scope of protection of this disclosure.

The method and the device are based on the same disclosure, and because the principles of solving the problems by the method and the device are similar, the implementation of the device and the method can be referred to each other, and the repetition is not repeated.

Referring to fig. 2, fig. 2 is a flowchart illustrating a method for predicting an index timing of an application program according to an embodiment of the disclosure. The method of the embodiment can be applied to a terminal device or a server, and the index time sequence prediction method of the application program comprises the following steps:

s201, acquiring scene information of a target application scene of an application program, and historical data of to-be-predicted indexes of the application program in the target application scene at different moments in a past preset time window.

In this embodiment, when an application program performs time-sequential prediction on a certain to-be-predicted index in a target application scene, since a future trend of the to-be-predicted index is related to the application scene where the to-be-predicted index is located, it is necessary to determine scene information of the target application scene, for example, at least one to-be-predicted index in service indexes such as a daily active user number (Daily Active User, DAU, for short term daily activity) of the application program, a newly added daily active user number, a next daily remaining user number, etc., where the scene information of the target application scene may include region information, operating system information, product line information, etc., for example, if the requirement is to predict the DAU under the region A, IOS operating system, music short video community application production line, it may be determined that the scene information includes: the region A, IOS operating system, the short video community application line, and multiple historical data for the DAU index under the region A, IOS operating system, the short video community application line, over a past preset time window, such as the DAU index values for each day over the past 30 days, may be obtained from a database or system.

The preset time window in this embodiment may be set according to the actual situation, that is, it is determined how long the data within the range is used to obtain the data of the business index of the future day, for example, the preset time window is 30 days, and then the data within the past 30 days may be used to predict the data of the business index of the future day.

S202, fusing and processing scene information of the target application scene and historical data of the index to be predicted at any moment, and constructing an input vector corresponding to the moment, wherein the input vector comprises a scene information part and an index historical data part.

In this embodiment, the scene information of the target application scene is also used as the parameter of the index time sequence prediction model, which can be used as a feature (variable) of one aspect of the parameter, and the scene information of the target application scene and the historical data of the index to be predicted are fused and processed to construct the input vector corresponding to the moment.

As an example, the above-mentioned example of the historical data of the region A, IOS operating system, the music short video community application line, and the DAU, the corresponding input vector structure is schematically shown as follows:

information of region A
	Information of IOS operating system
Information of music short video community application production line
	Numerical value of history data of DAU

The first three parts are scene information parts, and the last part is index historical data part.

Alternatively, since each scene information is a discrete variable, in order to convert the discrete scene information into a form that is easy to be utilized by a model, a one-hot (one-hot) encoding mode is adopted for processing in this embodiment. Specifically, as shown in the one-hot process in fig. 3, the scene information of the target application scene may be subjected to one-hot encoding to obtain an initial vector of the scene information part, where for the information of the region a, since there may be a region a, a region B, and a region C … … in relation to the region, the information of the region a is subjected to one-hot encoding to obtain (1, 0, …) for representing the information of the region a; similarly, there may be an IOS operating system and an Android operating system for the operating system, so information of the IOS operating system is obtained by one-hot encoding (1, 0) to be used for representing information of the IOS operating system; for the production line, there may be a short video community application production line of music class and another short video community application production line, so that the information of the short video community application production line of music class is obtained by one-hot encoding (1, 0) to be used for representing the information of the short video community application production line of music class.

For the index of the application program, there are a plurality of business indexes such as DAU, number of active users in new day, number of reserved users in next day, etc., so that one-hot encoding is performed on the index to be predicted, the index to be predicted is assumed to be obtained (0, 1,0, …), and the dimension corresponding to the index to be predicted in the one-hot encoding is set as the historical data of the index to be predicted at any moment, so as to obtain the initial vector of the historical data part of the index, for example, the historical data of the DAU at a certain moment is 1177434, and the initial vector of the historical data part of the index is (0,1177434,0, …)

Further, the initial vector of the scene information part and the initial vector of the index history data part are merged into one vector, for example, the input vector is obtained by combining according to a predetermined sequence.

Based on the above embodiment, considering that the one-hot encoding dimension is relatively large and the encoding is sparse, and the model calculation cost is increased, therefore, in this embodiment, the one-hot encoding may be optionally further reduced, specifically, an embedding layer (ebedding) process may be performed on an initial vector of a scene information portion, the initial vector of the scene information portion is mapped into a continuous vector, specifically, an ebedding process in fig. 3, after the one-hot encoding (1, 0, …) is performed on the information of the area a, the one-hot encoding is obtained (0.22,0.45,0.78), and after the ebedding process is performed on the information of the IOS operating system, the one-hot encoding (1, 0) is obtained (0.34,0.65, -0.25).

On the basis of the above embodiment, considering that the magnitudes of different indexes are different, for example, the magnitude of DAU is typically in the millions, and the magnitude of newly added daily activity is typically in the tens of thousands, it is necessary to normalize the service indexes of different magnitudes to the same data interval, and preserve the data timing trend, that is, normalize the initial vector of the index history data portion, and convert the numerical value in the initial vector of the index history data portion into the numerical value with the absolute value not greater than 1.

Optionally, when normalization processing is performed, historical data of one moment can be selected from the historical data of the to-be-predicted index at different moments in a preset time window to serve as reference historical data; and for the initial vector of the index historical data part at any moment, acquiring a difference value between the historical data of the index to be predicted and the reference historical data, and acquiring a ratio of the difference value to the reference historical data.

That is, = (history of the index to be predicted-reference history) reference history = history of the index to be predicted/reference history-1 after normalization

For example, the historical data before and after normalization is shown in the following table:

Before normalization	1150363	1177434	1130687	1185133	1129250	1126184	…
								Normalized to	0	0.023	-0.017	0.030	-0.018	-0.021	…

The final input vector of fig. 3 can be obtained through the above-described processing.

Of course, it should be noted that, in this embodiment, other manners may be adopted to encode the scene information of the target application scene and the history data of any of the to-be-predicted indexes, and convert the encoded history data into a vector form, so as to obtain the fused input vector.

S203, inputting each input vector into a preset index time sequence prediction model according to time sequence, and predicting the index to be predicted of the application program in a target application scene through the preset index time sequence prediction model, wherein the preset index time sequence prediction model is an index time sequence prediction model capable of performing time sequence prediction on various indexes of the application program in different application scenes.

In this embodiment, by performing the above processing on the scene information of the target application scene and the historical data of each index to be predicted, a plurality of input vectors may be obtained, which may be arranged according to the time sequence and input into a preset index time sequence prediction model, where the preset index time sequence prediction model is an index time sequence prediction model capable of performing time sequence prediction on various indexes of the application program in different application scenes, so that the predicted value of the index to be predicted in the target application scene may be obtained through the preset index time sequence prediction model. For example, if the predicted value of the index x to be predicted of the application program in the application scene M needs to be predicted, input data including a scene information M portion and an index x history data portion may be input; if the predicted value of the index y to be predicted of the application program in the application scene M is required to be predicted, inputting input data comprising a scene information M part and an index y historical data part; if the predicted value of the index x to be predicted of the application program in the application scene N is required to be predicted, inputting input data comprising a scene information N part and an index x historical data part; and so on.

In an alternative embodiment, as shown in fig. 4, the preset index timing prediction model is a BiLSTM (Bi-directional Long Short-Term Memory network) model, including a forward LSTM network and a backward LSTM network; the forward LSTM network comprises a plurality of LSTM units which are sequentially connected according to a time sequence, the backward LSTM network comprises a plurality of LSTM units which are sequentially connected according to a reverse sequence, the LSTM network can learn which information and which information is forgotten through a training process, and the forward LSTM network and the backward LSTM network of the BiLSTM model can better capture bidirectional semantic dependence; only a partial structure of the BiLSTM model is shown in fig. 4.

Further, as shown in fig. 5, in the above embodiment, predicting, by the preset indicator timing prediction model, the indicator to be predicted of the application program in the target application scenario in S203 may include:

s301, respectively inputting each input vector into a corresponding LSTM unit in a forward LSTM network according to a time sequence for processing; respectively inputting the input vectors into corresponding LSTM units in a backward LSTM network according to the reverse order for processing;

s302, combining the forward processing result and the backward processing result of the same input vector to obtain a corresponding result vector; summing or averaging the result vectors corresponding to the input vectors to obtain a final output vector;

S303, processing the output vector through a full connection layer to obtain a predicted value of the index to be predicted in the target application scene.

In the present embodiment, as shown in fig. 4, taking 3 input vectors as an example, an input vector x _t-2 、x _t-1 、x _t Respectively inputting the LSTM units into the corresponding LSTM units in the forward LSTM network according to the time sequence, and respectively inputting the LSTM units into the backward LSTM network according to the reverse order for processing, wherein the processing process can adopt a conventional LSTM processing process; further, the forward processing result and the backward processing result of the same input vector are combined to obtain a corresponding result vector, and the input vector x is _t-2 The forward processing result and the backward processing result are connected to obtain O _t-1 For input vector x _t-1 The forward processing result and the backward processing result are connected to obtain O _t For input vector x _t The forward processing result and the backward processing result are connected to obtain O _t+1 。

Further, a result vector O corresponding to each input vector _t-1 、O _t 、O _t+1 Summing or averaging to obtain final output vector y _t+1 Finally according to the final output vector y _t+1 Through the full connection layer (Fully Connected L)and enabling the layers to be processed and FC) to obtain the predicted value of the index to be predicted in the target application scene.

It should be noted that, in the embodiment of the disclosure, the preset indicator timing prediction model is not limited to BiLSTM, and other timing prediction models may be adopted, which is not described herein.

According to the index time sequence prediction method of the application program, scene information of a target application scene of the application program is obtained, and historical data of indexes to be predicted of the application program in the target application scene at different moments in a past preset time window are obtained; fusing and processing scene information of a target application scene and historical data of an index to be predicted at any moment, and constructing an input vector corresponding to the moment, wherein the input vector comprises a scene information part and an index historical data part; and inputting each input vector into a preset index time sequence prediction model according to time sequence, and predicting the index to be predicted of the application program in a target application scene through the preset index time sequence prediction model, wherein the preset index time sequence prediction model is an index time sequence prediction model capable of performing time sequence prediction on various indexes of the application program in different application scenes. When predicting the index to be predicted in the target application scene, the scene information is also used as a part of characteristic variables of the model entry, so that the universal index time sequence prediction model can be used for time sequence prediction, the same universal index time sequence prediction model can be used for predicting various indexes of the application program in different application scenes, the cost is reduced, and the efficiency is improved.

Referring to fig. 6, fig. 6 is a flowchart illustrating a training method of the index timing prediction model according to an embodiment of the disclosure. The method of the embodiment may be applied to a terminal device or a server, and may be executed on the same device as the embodiment of the method for predicting the index time sequence of the application program, or may be executed on a different device, where the training method of the index time sequence prediction model includes:

s401, acquiring scene information of different application scenes of an application program and historical data of different indexes of the application program under different application scenes;

s402, acquiring an input vector set corresponding to different indexes of the application program in different application scenes according to scene information of the different application scenes and historical data of the different indexes of the application program in the different application scenes, wherein the input vector set comprises a plurality of input vectors arranged according to time sequence, and the input vector comprises a scene information part and an index historical data part;

s403, training an initial index time sequence prediction model according to input vector sets corresponding to different indexes of the application program in different application scenes, so that the index time sequence prediction model can conduct time sequence prediction on various indexes of the application program in different application scenes.

In this embodiment, training data needs to be acquired for model training, and because the universal index timing prediction model can perform timing prediction on various service indexes of different scenes, scene information of different application scenes of an application program and history data of different indexes of the application program in different application scenes are required to be acquired, for example, for the above example, a plurality of history data of DAU indexes under a region A, IOS operating system and a music short video community application production line, a region B, IOS operating system and a plurality of history data of DAU indexes under a music short video community application production line, and a region C, IOS operating system and a plurality of history data … … of DAU indexes under a music short video community application production line can be acquired; the method can also obtain a plurality of historical data of DAU indexes under a region A, android operating system and a music short video community application production line, a plurality of historical data of DAU indexes under a region B, android operating system and a music short video community application production line, and a plurality of historical data … … of DAU indexes under a region C, android operating system and a music short video community application production line; the method can also obtain a plurality of historical data of DAU indexes under a region A, IOS operating system and other production lines, a region B, IOS operating system, a plurality of historical data … … of DAU indexes under other production lines, a region A, android operating system, a plurality of historical data of DAU indexes under other production lines, a region B, android operating system, a plurality of historical data … … of DAU indexes under other production lines; the method can also acquire a plurality of historical data of the newly added daily activity index under the area A, IOS operating system and the music short video community application production line, and acquire a plurality of historical data … … of the newly added daily activity index under the area B, IOS operating system and the music short video community application production line; etc.

That is, through the combination of different regions, different operating systems, different production lines and different service indexes, various situations can be combined and respectively used as different application scenes, in order to ensure the universality of the trained preset index time sequence prediction model, the combined application scenes are more comprehensive and better, the historical data sequences of indexes of corresponding application programs in each application scene are obtained, and the input vector set in each situation is obtained and used as training data through the construction of the same processing procedure as in the index time sequence prediction method embodiment of the application programs. The treatment process is as follows:

in one embodiment, the step S402 of obtaining the input vector set corresponding to the different indexes of the application program in the different application scenes according to the scene information of the different application scenes and the historical data of the different indexes of the application program in the different application scenes includes:

for scene information of any application scene and history data of any index of an application program in the application scene at any moment, performing one-hot coding on the scene information to acquire an initial vector of the scene information part;

Performing one-hot encoding on the index, and setting the dimension corresponding to the index in the one-hot encoding as the historical data of the index at the moment to obtain an initial vector of the historical data part of the index;

and merging the initial vector of the scene information part and the initial vector of the index history data part into a vector, and determining the vector as the input vector corresponding to the moment.

Further, optionally, after performing one-hot encoding on the scene information to obtain the initial vector of the scene information part, the method further includes:

carrying out ebedding processing on the initial vector of the scene information part, and mapping the initial vector of the scene information part into a continuous vector; and/or

After the initial vector of the index historical data part is obtained, the method further comprises the following steps:

and carrying out normalization processing on the initial vector of the index historical data part, and converting the numerical value in the initial vector of the index historical data part into a numerical value with the absolute value not more than 1.

The specific principle of the above process may be referred to the above embodiments, and will not be described herein.

The index timing prediction model in this embodiment may be a BiLSTM model, or may be another model.

On the basis of the foregoing embodiment, step S403 of training an initial indicator timing prediction model according to the input vector set corresponding to the different indicators of the application program in the different application scenarios includes:

and obtaining an index predicted value output by the index time sequence prediction model, calculating the root mean square error loss of the index predicted value and the index actual value, and carrying out adaptive training on the index time sequence prediction model according to the root mean square error loss.

In this embodiment, after acquiring input vector sets of different indexes under different application scene information of an application program, sequentially selecting input vectors in a preset time window from each set, inputting the input vectors into an index time sequence prediction model according to time sequence, predicting an index predicted value at the next moment, and calculating root mean square error loss (Root Mean Square Error, RMSE) of the index predicted value and an index actual value, and performing adaptive training (Adaptive moment estimation) on the index time sequence prediction model according to the root mean square error loss. Of course, other training methods may be used in this embodiment, and other loss functions may be used for the loss function, which is not illustrated here.

According to the training method of the index time sequence prediction model, the historical data of different indexes under the scene information of different application scenes of the application program are obtained, and the input vector sets of the different indexes of the application program under the different application scenes are constructed, so that the index time sequence prediction model is trained in multiple angles, various indexes of the application program under the different application scenes can be subjected to time sequence prediction by the index time sequence prediction model, the universality is high, and therefore the index time sequence prediction model does not need to be respectively learned for each application scene, the cost is reduced, and the efficiency is improved.

Fig. 7 is a block diagram of the configuration of the index timing prediction apparatus of the application program according to the embodiment of the present disclosure, corresponding to the index timing prediction method of the application program according to the above embodiment. For ease of illustration, only portions relevant to embodiments of the present disclosure are shown. Referring to fig. 7, the index timing prediction apparatus 500 of an application program includes: a data acquisition unit 501, a preprocessing unit 502, and a prediction unit 503.

A data obtaining unit 501, configured to obtain scene information of a target application scene of an application program, and historical data of indicators to be predicted of the application program in the target application scene at different moments in a preset time window in the past;

The preprocessing unit 502 is configured to fuse and process the scene information of the target application scene and the historical data of the index to be predicted at any moment, and construct an input vector corresponding to the moment, where the input vector includes a scene information part and an index historical data part;

the prediction unit 503 is configured to input each input vector into a preset indicator time sequence prediction model according to time sequence, and predict the indicator to be predicted of the application program in a target application scenario through the preset indicator time sequence prediction model, where the preset indicator time sequence prediction model is an indicator time sequence prediction model capable of performing time sequence prediction on various indicators of the application program in different application scenarios.

According to one or more embodiments of the present disclosure, when the preprocessing unit 502 fuses and processes the scene information of the target application scene and the historical data of the to-be-predicted index at any time, it is configured to:

performing one-hot independent encoding on scene information of the target application scene to obtain an initial vector of the scene information part;

performing one-hot encoding on an index to be predicted, and setting the dimension corresponding to the index to be predicted in the one-hot encoding as the historical data of the index to be predicted at any moment to obtain an initial vector of the index historical data part;

In accordance with one or more embodiments of the present disclosure, the preprocessing unit 502 is further configured to, before merging the initial vector of the scene information part and the initial vector of the index history data part into one vector:

carrying out embedding processing on the initial vector of the scene information part, and mapping the initial vector of the scene information part into a continuous vector; and/or

The preprocessing unit 502 is further configured to, after obtaining the initial vector of the index history data portion:

In accordance with one or more embodiments of the present disclosure, the preprocessing unit 502 is configured, in performing an initial vector normalization processing on the index history data portion, to:

selecting historical data of one moment from the historical data of the index to be predicted at different moments in the preset time window as reference historical data;

And for the initial vector of the index historical data part at any moment, acquiring a difference value between the historical data of the index to be predicted and the reference historical data, and acquiring a ratio of the difference value to the reference historical data.

According to one or more embodiments of the present disclosure, the preset index timing prediction model is a two-way long and short memory network BiLSTM model, including a forward LSTM network and a backward LSTM network; the forward LSTM network comprises a plurality of LSTM units which are sequentially connected according to a time sequence, and the backward LSTM network comprises a plurality of LSTM units which are sequentially connected according to a reverse sequence;

the prediction unit 503 is configured to, when predicting the to-be-predicted index of the application program in the target application scenario by using the preset index timing prediction model:

respectively inputting each input vector into a corresponding LSTM unit in a forward LSTM network according to a time sequence for processing; respectively inputting the input vectors into corresponding LSTM units in a backward LSTM network according to the reverse order for processing;

the forward processing result and the backward processing result of the same input vector are combined to obtain a corresponding result vector; summing or averaging the result vectors corresponding to the input vectors to obtain a final output vector;

And processing the output vector through a full connection layer to obtain the predicted value of the index to be predicted in the target application scene.

The index timing prediction device of the application program provided in this embodiment is used to execute the flow of the index timing prediction method embodiment of the application program, and its implementation principle and technical effects are similar, and this embodiment is not repeated here.

Corresponding to the training method of the index timing prediction model of the above embodiment, fig. 8 is a block diagram of the training apparatus of the index timing prediction model provided by the embodiment of the present disclosure. For ease of illustration, only portions relevant to embodiments of the present disclosure are shown. Referring to fig. 8, the training apparatus 600 of the index timing prediction model includes: a data acquisition unit 601, a preprocessing unit 602, and a training unit 603.

A data obtaining unit 601, configured to obtain scene information of different application scenes of an application program, and historical data of different indexes of the application program in the different application scenes;

the preprocessing unit 602 is configured to obtain an input vector set corresponding to different indexes of the application program in different application scenes according to scene information of the different application scenes and historical data of different indexes of the application program in the different application scenes, where the input vector set includes a plurality of input vectors arranged according to time sequence, and the input vector includes a scene information part and an index historical data part;

The training unit 603 is configured to train an initial indicator timing prediction model according to input vector sets corresponding to different indicators of an application program in different application scenarios, so that the indicator timing prediction model can perform timing prediction on various indicators of the application program in different application scenarios.

According to one or more embodiments of the present disclosure, when the training unit 603 trains an initial indicator timing prediction model according to input vector sets corresponding to different indicators of an application program in different application scenarios, the training unit is configured to:

According to one or more embodiments of the present disclosure, the preprocessing unit 602 is configured to, when acquiring input vector sets corresponding to different indexes of an application program in different application scenarios according to scenario information of the different application scenarios and historical data of different indexes of the application program in the different application scenarios:

In accordance with one or more embodiments of the present disclosure, the preprocessing unit 602 is further configured to, before merging the initial vector of the scene information part and the initial vector of the index history data part into one vector:

The preprocessing unit 602 is further configured to, after obtaining the initial vector of the indicator history data portion:

The training device of the index time sequence prediction model provided in this embodiment is used for executing the flow of the training method embodiment of the index time sequence prediction model, and its implementation principle and technical effects are similar, and this embodiment is not repeated here.

Referring to fig. 9, a schematic diagram of an electronic device 900 suitable for use in implementing embodiments of the present disclosure is shown, where the electronic device 900 may be a terminal device or a server. The terminal device may include, but is not limited to, a mobile terminal such as a mobile phone, a notebook computer, a digital broadcast receiver, a personal digital assistant (Personal Digital Assistant, PDA for short), a tablet (Portable Android Device, PAD for short), a portable multimedia player (Portable Media Player, PMP for short), an in-vehicle terminal (e.g., an in-vehicle navigation terminal), and the like, and a fixed terminal such as a digital TV, a desktop computer, and the like. The electronic device shown in fig. 9 is merely an example, and should not impose any limitations on the functionality and scope of use of embodiments of the present disclosure.

It should be noted that, the embodiment of the index time sequence prediction method and the embodiment of the training method of the index time sequence prediction model of the application program may be executed on the same electronic device or may be executed on different electronic devices.

As shown in fig. 9, the electronic apparatus 900 may include a processing device (e.g., a central processor, a graphics processor, or the like) 901, which may perform various appropriate actions and processes according to a program stored in a Read Only Memory (ROM) 902 or a program loaded from a storage device 908 into a random access Memory (Random Access Memory, RAM) 903. In the RAM 903, various programs and data necessary for the operation of the electronic device 900 are also stored. The processing device 901, the ROM 902, and the RAM 903 are connected to each other through a bus 904. An input/output (I/O) interface 905 is also connected to the bus 904.

In general, the following devices may be connected to the I/O interface 905: input devices 906 including, for example, a touch screen, touchpad, keyboard, mouse, camera, microphone, accelerometer, gyroscope, and the like; an output device 907 including, for example, a liquid crystal display (Liquid Crystal Display, LCD for short), a speaker, a vibrator, and the like; storage 908 including, for example, magnetic tape, hard disk, etc.; and a communication device 909. The communication means 909 may allow the electronic device 900 to communicate wirelessly or by wire with other devices to exchange data. While fig. 9 shows an electronic device 900 having various means, it is to be understood that not all illustrated means are required to be implemented or provided. More or fewer devices may be implemented or provided instead.

In particular, according to embodiments of the present disclosure, the processes described above with reference to flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable medium, the computer program comprising program code for performing the method shown in the flowcharts. In such an embodiment, the computer program may be downloaded and installed from a network via the communication device 909, or installed from the storage device 908, or installed from the ROM 902. When executed by the processing device 901, performs the above-described functions defined in the methods of the embodiments of the present disclosure.

It should be noted that the computer readable medium described in the present disclosure may be a computer readable signal medium or a computer readable storage medium, or any combination of the two. The computer readable storage medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples of the computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this disclosure, a computer-readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In the present disclosure, however, the computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, with the computer-readable program code embodied therein. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination of the foregoing. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: electrical wires, fiber optic cables, RF (radio frequency), and the like, or any suitable combination of the foregoing.

The computer readable medium may be contained in the electronic device; or may exist alone without being incorporated into the electronic device.

The computer-readable medium carries one or more programs which, when executed by the electronic device, cause the electronic device to perform the methods shown in the above-described embodiments.

Computer program code for carrying out operations of the present disclosure may be written in one or more programming languages, including an object oriented programming language such as Java, smalltalk, C ++ and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any kind of network, including a local area network (Local Area Network, LAN for short) or a wide area network (Wide Area Network, WAN for short), or it may be connected to an external computer (e.g., connected via the internet using an internet service provider).

The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

The units involved in the embodiments of the present disclosure may be implemented by means of software, or may be implemented by means of hardware. The name of the unit does not in any way constitute a limitation of the unit itself, for example the first acquisition unit may also be described as "unit acquiring at least two internet protocol addresses".

The functions described above herein may be performed, at least in part, by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: a Field Programmable Gate Array (FPGA), an Application Specific Integrated Circuit (ASIC), an Application Specific Standard Product (ASSP), a system on a chip (SOC), a Complex Programmable Logic Device (CPLD), and the like.

In the context of this disclosure, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

In a first aspect, according to one or more embodiments of the present disclosure, there is provided an index timing prediction method of an application program, including:

and inputting each input vector into a preset index time sequence prediction model according to time sequence, and predicting the index to be predicted of the application program in a target application scene through the preset index time sequence prediction model, wherein the preset index time sequence prediction model is an index time sequence prediction model capable of performing time sequence prediction on various indexes of the application program in different application scenes, and the preset index time sequence prediction model is an index time sequence prediction model capable of performing time sequence prediction on various indexes of the application program in different application scenes.

According to one or more embodiments of the present disclosure, the fusing and processing the scene information of the target application scene and the historical data of the to-be-predicted index at any moment, to construct an input vector corresponding to the moment, includes:

According to one or more embodiments of the present disclosure, the performing one-hot encoding on the scene information of the target application scene, after obtaining the initial vector of the scene information part, further includes:

According to one or more embodiments of the present disclosure, an initial vector normalization process for the index history data portion includes:

The predicting the to-be-predicted index of the application program in the target application scene through the preset index time sequence prediction model includes:

In a second aspect, according to one or more embodiments of the present disclosure, there is provided a training method of an index timing prediction model, including:

According to one or more embodiments of the present disclosure, the training an initial indicator timing prediction model according to input vector sets corresponding to different indicators of an application program in different application scenarios includes:

According to one or more embodiments of the present disclosure, the obtaining, according to scene information of different application scenes and historical data of different indexes of an application program in the different application scenes, an input vector set corresponding to the different indexes of the application program in the different application scenes includes:

According to one or more embodiments of the present disclosure, after performing one-hot encoding on the scene information to obtain the initial vector of the scene information part, the method further includes:

In a third aspect, according to one or more embodiments of the present disclosure, there is provided an index timing prediction apparatus of an application program, including:

According to one or more embodiments of the present disclosure, when the preprocessing unit fuses and processes the scene information of the target application scene and the historical data of the to-be-predicted index at any moment, the preprocessing unit is configured to:

According to one or more embodiments of the present disclosure, the preprocessing unit is further configured to, before merging the initial vector of the scene information part and the initial vector of the index history data part into one vector:

The preprocessing unit is further configured to, after obtaining the initial vector of the indicator history data portion:

According to one or more embodiments of the present disclosure, the preprocessing unit is configured, in performing an initial vector normalization processing on the index history data portion, to:

the prediction unit is configured to, when predicting the to-be-predicted index of the application program in the target application scenario through the preset index timing prediction model:

In a fourth aspect, according to one or more embodiments of the present disclosure, there is provided a training apparatus of an index timing prediction model, including:

According to one or more embodiments of the present disclosure, when the training unit trains an initial index timing prediction model according to input vector sets corresponding to different indexes of an application program in different application scenarios, the training unit is configured to:

According to one or more embodiments of the present disclosure, when acquiring input vector sets corresponding to different indexes of an application program in different application scenarios according to scenario information of the different application scenarios and historical data of the different indexes of the application program in the different application scenarios, the preprocessing unit is configured to:

In a fifth aspect, according to one or more embodiments of the present disclosure, there is provided an electronic device comprising: at least one processor and memory;

the memory stores computer-executable instructions;

In a sixth aspect, according to one or more embodiments of the present disclosure, there is provided an electronic device comprising: at least one processor and memory;

the memory stores computer-executable instructions;

In a seventh aspect, according to one or more embodiments of the present disclosure, there is provided a computer-readable storage medium having stored therein computer-executable instructions which, when executed by a processor, implement the index timing prediction method of the application program as described above in the first aspect and the various possible designs of the first aspect and/or the training method of the index timing prediction model as described above in the second aspect and the various possible designs of the second aspect.

According to an eighth aspect, according to one or more embodiments of the present disclosure, there is provided a computer program product comprising computer-executable instructions which, when executed by a processor, implement the index timing prediction method of an application program as described above in the first aspect and the various possible designs of the first aspect and/or the training method of the index timing prediction model as described above in the second aspect and the various possible designs of the second aspect.

The foregoing description is only of the preferred embodiments of the present disclosure and description of the principles of the technology being employed. It will be appreciated by persons skilled in the art that the scope of the disclosure referred to in this disclosure is not limited to the specific combinations of features described above, but also covers other embodiments which may be formed by any combination of features described above or equivalents thereof without departing from the spirit of the disclosure. Such as those described above, are mutually substituted with the technical features having similar functions disclosed in the present disclosure (but not limited thereto).

Moreover, although operations are depicted in a particular order, this should not be understood as requiring that such operations be performed in the particular order shown or in sequential order. In certain circumstances, multitasking and parallel processing may be advantageous. Likewise, while several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of the present disclosure. Certain features that are described in the context of separate embodiments can also be implemented in combination in a single embodiment. Conversely, various features that are described in the context of a single embodiment can also be implemented in multiple embodiments separately or in any suitable subcombination.

Although the subject matter has been described in language specific to structural features and/or methodological acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are example forms of implementing the claims.

Claims

1. An index timing prediction method for an application program, comprising:

fusing and processing scene information of the target application scene and historical data of the index to be predicted at any moment to construct an input vector corresponding to the moment, wherein the input vector comprises the scene information part and the index historical data part;

2. The method of claim 1, wherein the fusing and processing the scene information of the target application scene and the historical data of the to-be-predicted index at any moment to construct the input vector corresponding to the moment includes:

performing one-hot encoding on the index to be predicted, and setting the dimension corresponding to the index to be predicted in the one-hot encoding as the historical data of the index to be predicted at any moment to obtain an initial vector of the index historical data part;

3. The method according to claim 2, wherein the performing one-hot uni-thermal encoding on the scene information of the target application scene, after obtaining the initial vector of the scene information part, further comprises:

4. A method according to claim 3, wherein the initial vector normalization of the indicator history data portion comprises:

5. The method according to any one of claims 1-4, wherein the predetermined index timing prediction model is a two-way long and short memory network BiLSTM model, including a forward LSTM network and a backward LSTM network; the forward LSTM network comprises a plurality of LSTM units which are sequentially connected according to a time sequence, and the backward LSTM network comprises a plurality of LSTM units which are sequentially connected according to a reverse sequence;

6. The training method of the index time sequence prediction model is characterized by comprising the following steps of:

7. The method of claim 6, wherein training an initial indicator timing prediction model according to the input vector set corresponding to different indicators of the application program in different application scenarios comprises:

8. The method according to claim 6 or 7, wherein the obtaining the input vector set corresponding to the different indexes of the application program in the different application scenes according to the scene information of the different application scenes and the history data of the different indexes of the application program in the different application scenes includes:

9. The method of claim 8, wherein the performing one-hot encoding on the scene information, after obtaining the initial vector of the scene information part, further comprises:

10. An index timing prediction apparatus of an application program, comprising:

11. A training apparatus for an index timing prediction model, comprising:

The training unit is used for training an initial index time sequence prediction model according to input vector sets corresponding to different indexes of the application program in different application scenes to obtain a universal index time sequence prediction model, so that the index time sequence prediction model can conduct time sequence prediction on various indexes of the application program in different application scenes.

12. An electronic device, comprising: at least one processor and memory;

the memory stores computer-executable instructions;

the at least one processor executing computer-executable instructions stored in the memory causes the at least one processor to perform the method of any one of claims 1-5.

13. An electronic device, comprising: at least one processor and memory;

the memory stores computer-executable instructions;

the at least one processor executing computer-executable instructions stored in the memory causes the at least one processor to perform the method of any one of claims 6-9.

14. A computer readable storage medium having stored therein computer executable instructions which, when executed by a processor, implement the method of any of claims 1-5 or 6-9.

15. A computer program product comprising computer-executable instructions which, when executed by a processor, implement the method of any of claims 1-5 or 6-9.