CN103455555A - Recommendation method and device based on mobile terminal similarity - Google Patents

Recommendation method and device based on mobile terminal similarity Download PDF

Info

Publication number
CN103455555A
CN103455555A CN2013103395959A CN201310339595A CN103455555A CN 103455555 A CN103455555 A CN 103455555A CN 2013103395959 A CN2013103395959 A CN 2013103395959A CN 201310339595 A CN201310339595 A CN 201310339595A CN 103455555 A CN103455555 A CN 103455555A
Authority
CN
China
Prior art keywords
data
terminal
attribute
terminal attribute
similarity
Prior art date
Legal status (The legal status 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 status listed.)
Granted
Application number
CN2013103395959A
Other languages
Chinese (zh)
Other versions
CN103455555B (en
Inventor
雷凯
于倩
宁锐
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Peking University Shenzhen Graduate School
Original Assignee
Peking University Shenzhen Graduate School
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 Peking University Shenzhen Graduate School filed Critical Peking University Shenzhen Graduate School
Priority to CN201310339595.9A priority Critical patent/CN103455555B/en
Publication of CN103455555A publication Critical patent/CN103455555A/en
Application granted granted Critical
Publication of CN103455555B publication Critical patent/CN103455555B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Landscapes

  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a recommendation method and device based on mobile terminal similarity and relates to the field of mobile communication. The method includes: acquiring data, querying data, processing data, generating a terminal attribute data set, reducing dimensionality of the data, calculating recommendations and outputting the recommendations. The device comprises a data acquisition unit, a data query unit, a data processing unit, a terminal attribute data set generation unit, a data dimensionality reduction unit, a recommendation calculation unit, and a recommendation output unit. The data processing unit comprises a matrix query unit, a data addition unit, a similarity calculation unit and a terminal similarity matrix generation unit. The recommendation method and device has the advantages that personal recommendations are provided according to attribute difference of mobile terminals, and average errors of the traditional recommendation method are reduced.

Description

Recommend method and the recommendation apparatus of movement-based terminal similarity
Technical field
The application relates to moving communicating field, relates in particular to a kind of recommend method and recommendation apparatus of movement-based terminal similarity.
Background technology
By now, quantity of information is very abundant in internet development, has substantially exceeded that people can accept, the scope of disposal and utilization.The information of bulk redundancy is full of network, severe jamming the selection of user to useful information, information overload has become a problem demanding prompt solution, commending system is the important means of information filtering, can effectively alleviate problem of information overload.
Mobile commending system is the commending system that is applied in the mobile Internet field, is the extension of traditional commending system, and the current research to mobile commending system is in the stage at the early-stage.The thought of mobile commending system and traditional commending system is basically identical, the method and the algorithm that adopt also can be general, but, characteristics and limitation due to mobile terminal and mobile network, the mobile recommendation also will be affected, and the first, in mobile Internet, the user whenever and wherever possible can accessing Internet, obtaining information, and different customer locations also is not quite similar for demand; The second, the mobile Internet user compares the conventional internet user, and the context environmental faced is more complicated and changeable; The 3rd, the convenience of the processing power of mobile terminal, screen size, input and output is different.Therefore, mobile commending system has the requirement of higher real-time, accuracy, personalization.
According to investigation, show, the differentiation of mobile terminal and mobile platform makes the user present different individualized features and use habit, recommended for the multipair position of the commending system of mobile Internet and environment at present, be there is no the recommend method of specially the different mobile terminal attribute being recommended.
Summary of the invention
The application provides a kind of recommend method and recommendation apparatus of movement-based terminal similarity.
According to the application's first aspect, a kind of recommend method of movement-based terminal similarity is provided, comprising:
Obtain data: obtain user data package, user data package is expressed as by user property, item attribute, terminal attribute and preference attribute and forms;
Data query: the terminal attribute data in the user data package of obtaining are inquired about in the terminal attribute data group of setting up in advance; these terminal attribute data are the set with data of an above dimension; and on each dimension, the type of data is redefined for the established data type, this data type comprises the numeric type data that there is no the classifying type of numerical values recited relation data and the numerical values recited relation is arranged;
Data are processed: the result for data query is processed, if there are inquired about terminal attribute data in terminal attribute data group, inquires about matrix, the terminal similar matrix corresponding with these terminal attribute data that this inquiry matrix is set up in advance for inquiry;
If there are not inquired about terminal attribute data in terminal attribute data group, these terminal attribute data are added into to the terminal attribute data group of setting up in advance, the similarity of all terminals in the terminal attribute data group of calculating these terminal attribute data and setting up in advance, according to terminal similarity result of calculation, generated query terminal similar matrix, and preserve;
This terminal similarity is calculated and is carried out according to following formula:
deviceSim ( a , b ) = Σ i = 0 n - 1 S ( a i , b i ) Σ i = 0 n - 1 W ( a i , b i )
The terminal similarity that wherein deviceSim (a, b) is terminal attribute example a and terminal attribute example b;
S(a i, b i) be the dimension attributes similarity, mean the similarity of terminal attribute example a and the attribute of terminal attribute example b on i+1 dimension of terminal attribute;
W(ai, bi) be terminal attribute example a and the weights of terminal attribute example b on i+1 dimension of terminal attribute;
The computing method of this dimension attributes similarity are:
If terminal attribute is the classifying type data on a certain dimension, this dimension attributes similarity is got 1 value when data are identical on this dimension at terminal attribute, gets 0 value when different;
If terminal attribute is the numeric type data on a certain dimension, the dimension attributes similarity calculates and is undertaken by following formula:
S ( a i , b i ) = 1 - | a i - b i | Max ( D i ) - Min ( D i )
S (a wherein i, b i) be the dimension attributes similarity of terminal attribute example a and terminal attribute example b;
A imean the attribute of terminal attribute example a on i+1 dimension; b imean the attribute of terminal attribute example b on i+1 dimension;
Max (D i) be the maximal value of terminal attribute attribute on i+1 dimension, Min (D i) be terminal attribute on i+1 dimension attribute minimum value;
Generate the terminal attribute data acquisition: according to the similarity threshold of setting, the inquiry terminal similar matrix is processed, extracted the terminal attribute data acquisition that similarity is greater than similarity threshold, generate the first data;
Data Dimensionality Reduction: the corresponding terminal of terminal attribute in these first data is considered as to similar terminal, extracts UAD, item attribute data and the preference attribute data of this similar terminal, generate the second data;
Recommend to calculate: to the second data, utilize proposed algorithm to generate preference and predict and store;
Output is recommended: according to preference prediction output recommendation results.
According to the application's second aspect, a kind of recommendation apparatus of movement-based terminal similarity recommend method is provided, comprising:
Obtain data cell: for obtaining user data package, user data package is expressed as by user property, item attribute, terminal attribute and preference attribute and forms;
The data query unit: the terminal attribute data for the user data package to obtaining are inquired about in the terminal attribute data group of setting up in advance; these terminal attribute data are the set with data of an above dimension; and on each dimension, the type of data is redefined for the established data type, this data type comprises the numeric type data that there is no the classifying type of numerical values recited relation data and the numerical values recited relation is arranged;
Data processing unit: processed for the result for data query, if there are inquired about terminal attribute data in terminal attribute data group, inquire about matrix, the terminal similar matrix corresponding with these terminal attribute data that this inquiry matrix is set up in advance for inquiry;
If there are not inquired about terminal attribute data in terminal attribute data group, these terminal attribute data are added into to the terminal attribute data group of setting up in advance, the similarity of all terminals in the terminal attribute data group of calculating these terminal attribute data and setting up in advance, according to terminal similarity result of calculation, generated query terminal similar matrix, and preserve;
This terminal similarity is calculated and is carried out according to following formula:
deviceSim ( a , b ) = Σ i = 0 n - 1 S ( a i , b i ) Σ i = 0 n - 1 W ( a i , b i )
The terminal similarity that wherein deviceSim (a, b) is terminal attribute example a and terminal attribute example b;
S(a i, b i) be the dimension attributes similarity, mean the similarity of terminal attribute example a and the attribute of terminal attribute example b on i+1 dimension of terminal attribute;
W(ai, bi) be terminal attribute example a and the weights of terminal attribute example b on i+1 dimension of terminal attribute;
The computing method of this dimension attributes similarity are:
If terminal attribute is the classifying type data on a certain dimension, this dimension attributes similarity is got 1 value when data are identical on this dimension at terminal attribute, gets 0 value when different;
If terminal attribute is the numeric type data on a certain dimension, the dimension attributes similarity calculates and is undertaken by following formula:
S ( a i , b i ) = 1 - | a i - b i | Max ( D i ) - Min ( D i )
S (a wherein i, b i) be the dimension attributes similarity of terminal attribute example a and terminal attribute example b;
A imean the attribute of terminal attribute example a on i+1 dimension; b imean the attribute of terminal attribute example b on i+1 dimension;
Max (D i) be the maximal value of terminal attribute attribute on i+1 dimension, Min (D i) be terminal attribute on i+1 dimension attribute minimum value;
Generate the terminal attribute data aggregation unit: for the similarity threshold according to setting, the inquiry terminal similar matrix is processed, extracted the terminal attribute data acquisition that similarity is greater than similarity threshold, generate the first data;
Data Dimensionality Reduction unit: be considered as similar terminal for the corresponding terminal of terminal attribute by these first data, extract UAD, item attribute data and the preference attribute data of this similar terminal, generate the second data;
Recommend computing unit: for utilize proposed algorithm to generate preference to the second data, predict and store;
Output recommendation unit: for predict the output recommendation results according to preference;
Wherein data processing unit comprises:
The inquiry matrix unit: for when there are inquired about terminal attribute data in terminal attribute data group, the terminal similar matrix corresponding with these terminal attribute data that inquiry is set up in advance;
Add data cell: for when there are not inquired about terminal attribute data in terminal attribute data group, these terminal attribute data are added into to the terminal attribute data group of setting up in advance;
Similarity calculated: for the similarity of the terminal attribute data of calculating this interpolation and all terminals of terminal attribute data group of setting up in advance;
Generated query terminal similar matrix unit: for according to terminal similarity result of calculation, generated query terminal similar matrix, and preserve.
The application's beneficial effect is recommend method and the recommendation apparatus that has proposed a kind of movement-based terminal similarity, has realized carrying out personalized recommendation for the difference of mobile terminal attribute, has reduced the average error of traditional recommend method.
The accompanying drawing explanation
The data Establishing process figure of the recommend method that Fig. 1 is movement-based terminal similarity;
The process flow diagram of the recommend method embodiment that Fig. 2 is movement-based terminal similarity;
The structured flowchart of the recommendation apparatus embodiment that Fig. 3 is movement-based terminal similarity recommend method;
The structured flowchart of the recommendation apparatus data processing unit that Fig. 4 is movement-based terminal similarity recommend method.
Embodiment
Below by embodiment, by reference to the accompanying drawings the present invention is described in further detail.
The application provides a kind of recommend method and recommendation apparatus of movement-based terminal similarity.
Embodiment mono-:
The recommend method of the application's movement-based terminal similarity carries out based on collaborative filtering, because generally there is the cold start-up problem in collaborative filtering, if do not exist raw data or raw data less while being recommended in system, the recommendation results precision that can't export recommendation results or output is lower, in order to solve the cold start-up problem, need carry out to system the foundation of raw data, in the present embodiment, the foundation of raw data is adopted the mode of typing user data, simultaneously in order to facilitate subsequent user to use the method to be recommended, also the user is generated and stored in follow-up required data, the data Establishing process figure of the recommend method of movement-based terminal similarity as shown in Figure 1, comprise:
Obtain data 10: obtain user data package, user data package is expressed as by user property, item attribute, terminal attribute and preference attribute and forms;
In carrying out system during the setting up of raw data, obtaining user data package can be a document, the form of document includes but not limited to txt, word and xls, the document comprises user property, item attribute, terminal attribute and four data rows of preference attribute, is called UAD group, item attribute data group, terminal attribute data group and preference attribute data group; If preference attribute disappearance, available specific symbol meaned, as null etc.Every data line of document is mutual corresponding relation, is mutually corresponding user property, item attribute, terminal attribute and preference attribute.User property, item attribute, terminal attribute and preference attribute data may be various structures, can, by creating a rational protocol rule, to data, carry out standardization.In this application, user property is used for identifying different user, can directly be standardized as digital data, as 1,2,3 etc.; The preference attribute turns to the numeral of certain limit, the scoring as 1 to 5,1 to 100 satisfaction etc.; Terminal attribute has various structures, the data structure of setting terminal attribute according to actual needs, and item attribute is generally limited due to data scale, can not process.Data are carried out to standardization, be conducive to the execution of follow-up flow process.
For different user is recommended, must understand the preference value of a certain project of different user, for the user, the hobby of project is recommended, be only the effective way of recommendation.The user is for the preference data of a certain project, and generally comment, third party evaluation mechanism or the alternate manner by user under a certain project obtains.The data of obtaining by such mode are not the complete data that comprise user property, item attribute, terminal attribute and preference attribute usually, usually excalation preference attribute data.The purpose of the recommend method of the application's movement-based terminal similarity is exactly the preference attribute data that obtains more accurately this excalation.
The user data package that the application obtains can be expressed as by user property, item attribute, terminal attribute and preference attribute and form, had more terminal attribute than traditional data that formed by user property, item attribute and preference attribute that are expressed as, the recommend method that is a kind of movement-based terminal of the application similarity is not only considered the preference attribute based on user property and item attribute, also based on terminal attribute.
The terminal similarity calculates 11: the terminal attribute data that comprise all terminals in user data package are carried out to the calculating of terminal similarity;
The terminal similarity calculates 11 for for all terminals of terminal attribute data group in user data package, to carry out the calculating of terminal similarity.
The terminal similarity is calculated according to the terminal similarity algorithm of setting and is carried out, and the calculating of terminal similarity will be set forth later.
Generate terminal similar matrix 12: calculate 11 the result generation terminal similar matrix corresponding with terminal attribute according to the terminal similarity, and storage;
The step that generates terminal similar matrix 12 also comprises: the data to the terminal attribute data group of carrying out the calculating of terminal similarity are carried out deduplication processing renewal;
Data to the terminal similar matrix are carried out deduplication processing renewal.
The purpose of data being carried out to deduplication processing renewal is in order to reduce calculated amount, saves storage space.If n different terminals arranged, total n after calculating by the terminal similarity so 2individual data, consider some characteristics of terminal similar matrix, can carry out data compression to it.Be specially: the restriction of terminal sequencing while due to the similarity of two terminal rooms, not carried out the calculating of terminal similarity, regardless of carrying out terminal similarity terminal sequencing while calculating, drawn terminal similarity is same value, therefore by all terminal attributes being carried out to the terminal similar matrix that the terminal similarity calculates, be symmetric matrix, can remove symmetric data, retain upper triangle or the lower triangular matrix of symmetric matrix; The terminal similar matrix generated in addition must comprise the similarity of same terminal, and the terminal similarity is 1, can, by this record deletion, remove the cornerwise data of terminal similar matrix.Not only can carry out above-mentioned deduplication to the data of terminal similar matrix and process, can also remove the data of complete different terminals, the data that the terminal similarity is 0.The data of the terminal similar matrix after above processing are stored, covered original terminal similar matrix.After above processing, although the order of magnitude of data does not reduce, saved storage space.The data of the terminal similar matrix after above processing are stored, covered original terminal similar matrix.
The terminal similar matrix corresponding with terminal attribute generated comprised the similarity between terminal, this terminal similar matrix is stored, can be first by the mode of inquiry when the user uses, so that simple flow, the inquiry mode when user uses will be set forth later.
By performing step 10～12, not only in system typing raw data, solved the cold start-up problem, and can in the system of the recommend method of the application's movement-based terminal similarity, set up original user data bag and the terminal similar matrix corresponding with terminal attribute, when the user uses the recommend method of the application's movement-based terminal similarity, can first utilize terminal attribute data group in the user data package of setting up in advance in system to be inquired about with the terminal similar matrix corresponding with terminal attribute, and the data basis that utilizes in system the user data package set up in advance to calculate as collaborative filtering, as shown in Figure 2, the process flow diagram of the recommend method embodiment that Fig. 2 is the application's movement-based terminal similarity, comprise:
Obtain data 10, with obtaining in Fig. 1, data 10 are identical, and this repeats no more.
The user data package got in Fig. 2 is comprised of user property, item attribute, terminal attribute and preference attribute, the data that user property, item attribute, terminal attribute and the preference attribute of correspondence form mutually are designated as to the preference record, it in this user data package, can be a preference record, can be also many preference records, utilize the purpose of the recommend method of the application's movement-based terminal similarity to be to obtain unknown preference attribute data value in the preference record.
Data query 21: the terminal attribute data in the user data package of obtaining are inquired about in the terminal attribute data group of setting up in advance; these terminal attribute data are the set with data of an above dimension; and on each dimension, the type of data is redefined for the established data type, this data type comprises the numeric type data that there is no the classifying type of numerical values recited relation data and the numerical values recited relation is arranged;
To obtaining terminal attribute data in user data package, in the terminal attribute data group of setting up in advance, inquired about, in the raw data process that this terminal attribute data group of setting up has in advance comprised system as shown in Figure 1 and the terminal attribute data that accumulate in subsequent user use procedure as shown in Figure 2.If comprise many preference records in the user data package of obtaining, the mode that the terminal attribute that adopts every preference to record is inquired about respectively, the terminal attribute data group of setting up is in advance inquired about, if there is the terminal attribute data message of required inquiry, can directly utilize the terminal similar matrix of setting up in advance, be to have comprised similarity between the terminal attribute of this inquiry and other terminal attribute in system, do not need the terminal similarity is calculated.
User's terminal attribute has multiple, as brand, system, version etc., can mean a terminal that has n attribute with a n-dimensional vector model, as follows:
Device={D 0,D 1,···,D i,···,D n-1}
D wherein imean the attribute of terminal attribute on i+1 dimension.
Data process 22: the result for data query is processed, if there are inquired about terminal attribute data in terminal attribute data group, inquire about matrix 221, the terminal similar matrix corresponding with these terminal attribute data that this inquiry matrix 221 is set up in advance for inquiry;
If there are not inquired about terminal attribute data in terminal attribute data group, these terminal attribute data are added into to the terminal attribute data group of setting up in advance, the similarity of all terminals in the terminal attribute data group of calculating these terminal attribute data and setting up in advance, according to terminal similarity result of calculation, generated query terminal similar matrix, and preserve;
If there are inquired about terminal attribute data in terminal attribute data group, inquire about matrix 221, the similarity that the inquiry matrix is this terminal of inquiry and other all terminals extracts Query Result from matrix.The preserving type of Query Result can, for multiple, be preserved as adopted matrix or the form identical with the flesh and blood of matrix representation.
As shown in Figure 2, if do not exist inquired about terminal attribute data to carry out in terminal attribute data group: add data 222, similarity calculating 223 and generate terminal similar matrix 224;
Add data 222: the terminal attribute data are added into to the terminal attribute data group of setting up in advance;
In the present embodiment, can adopt the user data package will got to be added into the mode in the user data package document of setting up in advance in system, the user property in the user data package that is about to get, item attribute, terminal attribute and preference attribute are added into respectively in UAD group, item attribute data group, terminal attribute data group and the preference attribute data group of setting up in advance.So, both the terminal attribute data had been added into to the terminal attribute data group of setting up in advance, and had preserved again the user data package of newly obtaining, the data in this user data package can be used as the accumulation of data in system, the basic data of using as subsequent user.
Similarity calculates 223: the similarity of calculating all terminals in the terminal attribute data of newly adding and the terminal attribute data group of setting up in advance;
The terminal similarity is calculated and is carried out according to following formula:
deviceSim ( a , b ) = Σ i = 0 n - 1 S ( a i , b i ) Σ i = 0 n - 1 W ( a i , b i )
The terminal similarity that wherein deviceSim (a, b) is terminal attribute example a and terminal attribute example b;
S(a i, b i) be the dimension attributes similarity, mean the similarity of terminal attribute example a and the attribute of terminal attribute example b on i+1 dimension of terminal attribute;
W(ai, bi) be terminal attribute example a and the weights of terminal attribute example b on i+1 dimension of terminal attribute;
The computing method of dimension attributes similarity are:
If terminal attribute is not for there is no the classifying type data of numerical values recited relation on a certain dimension, the dimension attributes similarity is got 1 value when data are identical on this dimension at terminal attribute, gets 0 value when different;
If terminal attribute is for there being the numeric type data of numerical values recited relation on a certain dimension, the dimension attributes similarity calculates and is undertaken by following formula:
S ( a i , b i ) = 1 - | a i - b i | Max ( D i ) - Min ( D i )
S (a wherein i, b i) be the dimension attributes similarity of terminal attribute example a and terminal attribute example b;
A imean the attribute of terminal attribute example a on i+1 dimension; b imean the attribute of terminal attribute example b on i+1 dimension;
Max (D i) be the maximal value of terminal attribute attribute on i+1 dimension, Min (D i) be terminal attribute on i+1 dimension attribute minimum value.
Generate terminal similar matrix 224: calculate 223 results according to similarity, generated query terminal similar matrix, and preserve;
The similarity of all terminals in the terminal attribute data of newly interpolation of similarity calculating 223 calculating and the terminal attribute data group of foundation in advance, the inquiry terminal similar matrix of the similarity of all terminals in the terminal attribute data group of the terminal attribute data that generation comprises this new interpolation and foundation in advance.
This inquiry terminal similar matrix can be added into to the terminal similar matrix of setting up in advance, and preserve, the similarity data that set up having comprised after realizing adding in advance and the new terminal attribute added and the terminal similar matrix of other all terminal similarity data.
Generate terminal attribute data acquisition 23: according to the similarity threshold of setting, the inquiry terminal similar matrix is processed, extracted the terminal attribute set that similarity is greater than similarity threshold, generate the first data;
These first data can adopt the form of matrix to mean, comprised similarity and be greater than the end message of similarity threshold and the similarity information between terminal in the first data.
Data Dimensionality Reduction 24: the terminal attribute in the first data is considered as to similar terminal, extracts UAD, item attribute data and the preference attribute data of described similar terminal, generate the second data;
The terminal that similarity in the inquiry terminal similar matrix is greater than to similarity threshold is considered as similar terminal, because the terminal attributive information possibility of similar terminal is identical, therefore can't extract according to terminal attribute the user property of this terminal attribute, item attribute and preference attribute, can adopt according to terminal attribute and search corresponding user property, then search the mode of corresponding item attribute and preference attribute according to user property.Search corresponding user property according to terminal attribute, the principle of following is: if the first the data matrix or the form identical with the flesh and blood of matrix representation mean, the position of this terminal attribute in former participation terminal similarity computing terminal attribute data group can be found by the dimensional information at terminal attribute place in matrix, according to this position, corresponding user property can be found; And at first search user property but not search item attribute and the reason of preference attribute is, only there is user property not have repetition, may there be polyisomenism in the item attribute that certain user property is corresponding and preference attribute, can't be corresponding one by one.
Terminal attribute in the first data is considered as to similar terminal, extract UAD, item attribute data and the preference attribute data of described similar terminal, generate the second data, these second data be regenerate only comprise UAD group, the document of item attribute data group and preference attribute data group.
Generate terminal attribute data acquisition 23 steps and Data Dimensionality Reduction 24 steps by execution, realized the user property that has will comprised in the user data package of input, the data of item attribute, terminal attribute and preference attribute are converted to the second data that comprise user property, item attribute and preference attribute, so both consider the terminal attribute of equipment, and be convenient to again utilize traditional proposed algorithm to recommend to calculate to data.
Recommend to calculate 25: to described the second data, utilize proposed algorithm to generate the preference prediction;
The proposed algorithm adopted is slope one proposed algorithm, and slope one proposed algorithm is project-based collaborative filtering, adopts this algorithm, can calculate according to the existing subscriber preference prediction of the project to not providing the preference property value to the preference property value of project.
Output recommends 26: according to preference prediction output recommendation results.
In the present embodiment, the recommendation results of output is the preference attribute data value of output user to project.
Known by step 21～26, this step is only calculated for a preference record in the user data package of obtaining, if comprise M bar preference record in this user data package, step 21～26 need be carried out M time.How initialization system automatically performs repeatedly flow process, is known to the skilled person general knowledge, repeats no more herein.
Embodiment bis-:
According to the application's second aspect, a kind of recommendation apparatus of movement-based terminal similarity recommend method is provided, the structured flowchart of the recommendation apparatus embodiment of movement-based terminal similarity recommend method as shown in Figure 3, comprising:
Obtain data cell 30: for obtaining user data package, user data package is expressed as by user property, item attribute, terminal attribute and preference attribute and forms;
Data query unit 31: the terminal attribute data for the user data package to obtaining are inquired about in the terminal attribute data group of setting up in advance; these terminal attribute data are the set with data of an above dimension; and on each dimension, the type of data is redefined for the established data type, this data type comprises the numeric type data that there is no the classifying type of numerical values recited relation data and the numerical values recited relation is arranged;
Data processing unit 32: processed for the result for data query, if there are inquired about terminal attribute data in terminal attribute data group, inquire about matrix, the terminal similar matrix corresponding with these terminal attribute data that this inquiry matrix is set up in advance for inquiry;
If there are not inquired about terminal attribute data in terminal attribute data group, these terminal attribute data are added into to the terminal attribute data group of setting up in advance, the similarity of all terminals in the terminal attribute data group of calculating these terminal attribute data and setting up in advance, according to terminal similarity result of calculation, generated query terminal similar matrix, and preserve;
This terminal similarity is calculated and is carried out according to following formula:
deviceSim ( a , b ) = Σ i = 0 n - 1 S ( a i , b i ) Σ i = 0 n - 1 W ( a i , b i )
The terminal similarity that wherein deviceSim (a, b) is terminal attribute example a and terminal attribute example b;
S(a i, b i) be the dimension attributes similarity, mean the similarity of terminal attribute example a and the attribute of terminal attribute example b on i+1 dimension of terminal attribute;
W(ai, bi) be terminal attribute example a and the weights of terminal attribute example b on i+1 dimension of terminal attribute;
The computing method of this dimension attributes similarity are:
If terminal attribute is the classifying type data on a certain dimension, this dimension attributes similarity is got 1 value when data are identical on this dimension at terminal attribute, gets 0 value when different;
If terminal attribute is the numeric type data on a certain dimension, the dimension attributes similarity calculates and is undertaken by following formula:
S ( a i , b i ) = 1 - | a i - b i | Max ( D i ) - Min ( D i )
S (a wherein i, b i) be the dimension attributes similarity of terminal attribute example a and terminal attribute example b;
A imean the attribute of terminal attribute example a on i+1 dimension; b imean the attribute of terminal attribute example b on i+1 dimension;
Max (D i) be the maximal value of terminal attribute attribute on i+1 dimension, Min (D i) be terminal attribute on i+1 dimension attribute minimum value;
Generate terminal attribute data aggregation unit 33: for the similarity threshold according to setting, the inquiry terminal similar matrix is processed, extracted the terminal attribute data acquisition that similarity is greater than similarity threshold, generate the first data;
Data Dimensionality Reduction unit 34: be considered as similar terminal for the corresponding terminal of terminal attribute by these first data, extract UAD, item attribute data and the preference attribute data of this similar terminal, generate the second data;
Recommend computing unit 35: for utilize proposed algorithm to generate preference to the second data, predict and store;
Output recommendation unit 36: for predict the output recommendation results according to preference;
The structured flowchart of the recommendation apparatus data processing unit 32 of movement-based terminal similarity recommend method as shown in Figure 4, comprising:
Inquiry matrix unit 321: for when there are inquired about terminal attribute data in terminal attribute data group, the terminal similar matrix corresponding with these terminal attribute data that inquiry is set up in advance;
Add data cell 322: for when there are not inquired about terminal attribute data in terminal attribute data group, these terminal attribute data are added into to the terminal attribute data group of setting up in advance;
Similarity calculated 323: for the similarity of the terminal attribute data of calculating this interpolation and all terminals of terminal attribute data group of setting up in advance;
Generate terminal similar matrix unit 324: for according to terminal similarity result of calculation, generated query terminal similar matrix, and preserve.
In sum, the application's beneficial effect is recommend method and the recommendation apparatus that has proposed a kind of movement-based terminal similarity, realized carrying out personalized recommendation for the difference of mobile terminal attribute, owing to having considered terminal attribute, therefore reduced the average error of traditional recommend method.
It will be appreciated by those skilled in the art that, in above-mentioned embodiment, all or part of step of the whole bag of tricks can come the instruction related hardware to complete by program, this program can be stored in a computer-readable recording medium, and storage medium can comprise: ROM (read-only memory), random access memory, disk or CD etc.
Above content is in conjunction with concrete embodiment further description made for the present invention, can not assert that specific embodiment of the invention is confined to these explanations.For the general technical staff of the technical field of the invention, without departing from the inventive concept of the premise, can also make some simple deduction or replace.

Claims (4)

1. the recommend method of a movement-based terminal similarity, is characterized in that, comprising:
Obtain data: obtain user data package, user data package is expressed as by user property, item attribute, terminal attribute and preference attribute and forms;
Data query: the terminal attribute data in the user data package of obtaining are inquired about in the terminal attribute data group of setting up in advance; described terminal attribute data are the set with data of an above dimension; and on each dimension, the type of data is redefined for the established data type, described data type comprises the numeric type data that there is no the classifying type of numerical values recited relation data and the numerical values recited relation is arranged;
Data are processed: the result for data query is processed, if there are inquired about terminal attribute data in terminal attribute data group, inquires about matrix, the terminal similar matrix corresponding with these terminal attribute data that described inquiry matrix is set up in advance for inquiry;
If there are not inquired about terminal attribute data in terminal attribute data group, described terminal attribute data are added into to the terminal attribute data group of setting up in advance, the similarity of all terminals in the terminal attribute data group of calculating described terminal attribute data and setting up in advance, according to terminal similarity result of calculation, generated query terminal similar matrix, and preserve;
Described terminal similarity is calculated and is carried out according to following formula:
deviceSim ( a , b ) = Σ i = 0 n - 1 S ( a i , b i ) Σ i = 0 n - 1 W ( a i , b i )
The terminal similarity that wherein deviceSim (a, b) is terminal attribute example a and terminal attribute example b;
S(a i, b i) be the dimension attributes similarity, mean the similarity of terminal attribute example a and the attribute of terminal attribute example b on i+1 dimension of terminal attribute;
W(ai, bi) be terminal attribute example a and the weights of terminal attribute example b on i+1 dimension of terminal attribute;
The computing method of described dimension attributes similarity are:
If terminal attribute is the classifying type data on a certain dimension, described dimension attributes similarity is got 1 value when data are identical on this dimension at terminal attribute, gets 0 value when different;
If terminal attribute is the numeric type data on a certain dimension, described dimension attributes similarity calculates and is undertaken by following formula:
S ( a i , b i ) = 1 - | a i - b i | Max ( D i ) - Min ( D i )
S (a wherein i, b i) be the dimension attributes similarity of terminal attribute example a and terminal attribute example b;
A imean the attribute of terminal attribute example a on i+1 dimension; b imean the attribute of terminal attribute example b on i+1 dimension;
Max (D i) be the maximal value of terminal attribute attribute on i+1 dimension, Min (D i) be terminal attribute on i+1 dimension attribute minimum value;
Generate the terminal attribute data acquisition: according to the similarity threshold of setting, the inquiry terminal similar matrix is processed, extracted the terminal attribute data acquisition that similarity is greater than similarity threshold, generate the first data;
Data Dimensionality Reduction: the corresponding terminal of terminal attribute in described the first data is considered as to similar terminal, extracts UAD, item attribute data and the preference attribute data of described similar terminal, generate the second data;
Recommend to calculate: to described the second data, utilize proposed algorithm to generate preference and predict and store;
Output is recommended: according to preference prediction output recommendation results.
2. the recommend method of movement-based terminal similarity as claimed in claim 1, is characterized in that, described proposed algorithm is Slope One proposed algorithm.
3. the recommendation apparatus of a movement-based terminal similarity recommend method, is characterized in that, comprising:
Obtain data cell: for obtaining user data package, user data package is expressed as by user property, item attribute, terminal attribute and preference attribute and forms;
The data query unit: the terminal attribute data for the user data package to obtaining are inquired about in the terminal attribute data group of setting up in advance; described terminal attribute data are the set with data of an above dimension; and on each dimension, the type of data is redefined for the established data type, described data type comprises the numeric type data that there is no the classifying type of numerical values recited relation data and the numerical values recited relation is arranged;
Data processing unit: processed for the result for data query, if there are inquired about terminal attribute data in terminal attribute data group, inquire about matrix, the terminal similar matrix corresponding with these terminal attribute data that described inquiry matrix is set up in advance for inquiry;
If there are not inquired about terminal attribute data in terminal attribute data group, these terminal attribute data are added into to the terminal attribute data group of setting up in advance, the similarity of all terminals in the terminal attribute data group of calculating these terminal attribute data and setting up in advance, according to terminal similarity result of calculation, generated query terminal similar matrix, and preserve;
Described terminal similarity is calculated and is carried out according to following formula:
deviceSim ( a , b ) = Σ i = 0 n - 1 S ( a i , b i ) Σ i = 0 n - 1 W ( a i , b i )
The terminal similarity that wherein deviceSim (a, b) is terminal attribute example a and terminal attribute example b;
S(a i, b i) be the dimension attributes similarity, mean the similarity of terminal attribute example a and the attribute of terminal attribute example b on i+1 dimension of terminal attribute;
W(ai, bi) be terminal attribute example a and the weights of terminal attribute example b on i+1 dimension of terminal attribute;
The computing method of described dimension attributes similarity are:
If terminal attribute is the classifying type data on a certain dimension, described dimension attributes similarity is got 1 value when data are identical on this dimension at terminal attribute, gets 0 value when different;
If terminal attribute is the numeric type data on a certain dimension, the dimension attributes similarity calculates and is undertaken by following formula:
S ( a i , b i ) = 1 - | a i - b i | Max ( D i ) - Min ( D i )
S (a wherein i, b i) be the dimension attributes similarity of terminal attribute example a and terminal attribute example b;
A imean the attribute of terminal attribute example a on i+1 dimension; b imean the attribute of terminal attribute example b on i+1 dimension;
Max (D i) be the maximal value of terminal attribute attribute on i+1 dimension, Min (D i) be terminal attribute on i+1 dimension attribute minimum value;
Generate the terminal attribute data aggregation unit: for the similarity threshold according to setting, the inquiry terminal similar matrix is processed, extracted the terminal attribute data acquisition that similarity is greater than similarity threshold, generate the first data;
Data Dimensionality Reduction unit: be considered as similar terminal for the corresponding terminal of terminal attribute by described the first data, extract UAD, item attribute data and the preference attribute data of described similar terminal, generate the second data;
Recommend computing unit: for utilize proposed algorithm to generate preference to described the second data, predict and store;
Output recommendation unit: for predict the output recommendation results according to preference.
4. the recommendation apparatus of a kind of movement-based terminal similarity recommend method as claimed in claim 3, is characterized in that, described data processing unit comprises:
The inquiry matrix unit: for when there are inquired about terminal attribute data in terminal attribute data group, the terminal similar matrix corresponding with these terminal attribute data that inquiry is set up in advance;
Add data cell: for when there are not inquired about terminal attribute data in terminal attribute data group, these terminal attribute data are added into to the terminal attribute data group of setting up in advance;
Similarity calculated: for the similarity of the terminal attribute data of calculating described interpolation and all terminals of terminal attribute data group of setting up in advance;
Generated query terminal similar matrix unit: for according to terminal similarity result of calculation, generated query terminal similar matrix, and preserve.
CN201310339595.9A 2013-08-06 2013-08-06 Recommendation method and recommendation apparatus based on mobile terminal similarity Active CN103455555B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201310339595.9A CN103455555B (en) 2013-08-06 2013-08-06 Recommendation method and recommendation apparatus based on mobile terminal similarity

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201310339595.9A CN103455555B (en) 2013-08-06 2013-08-06 Recommendation method and recommendation apparatus based on mobile terminal similarity

Publications (2)

Publication Number Publication Date
CN103455555A true CN103455555A (en) 2013-12-18
CN103455555B CN103455555B (en) 2016-06-08

Family

ID=49737918

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201310339595.9A Active CN103455555B (en) 2013-08-06 2013-08-06 Recommendation method and recommendation apparatus based on mobile terminal similarity

Country Status (1)

Country Link
CN (1) CN103455555B (en)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104731866A (en) * 2015-02-27 2015-06-24 湖南大学 Individual gourmet recommending method based on position
CN105205107A (en) * 2015-08-27 2015-12-30 湖南人文科技学院 Internet of Things data similarity processing method
CN105488213A (en) * 2015-12-11 2016-04-13 哈尔滨工业大学深圳研究生院 LBS-oriented individual recommendation method based on Markov prediction algorithm
CN105989131A (en) * 2014-10-21 2016-10-05 株式会社日立制作所 Information search/presentation device and information search/presentation method
CN106846082A (en) * 2016-12-10 2017-06-13 江苏途致信息科技有限公司 Tourism cold start-up consumer products commending system and method based on hardware information
CN108121803A (en) * 2017-12-22 2018-06-05 维沃移动通信有限公司 A kind of method and server of definite page layout
CN108629609A (en) * 2017-03-22 2018-10-09 ***通信集团河北有限公司 The method and apparatus of reflexless terminal

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120130848A1 (en) * 2010-11-24 2012-05-24 JVC Kenwood Corporation Apparatus, Method, And Computer Program For Selecting Items
CN102622390A (en) * 2011-10-11 2012-08-01 北京掌汇天下科技有限公司 Application recommending method and application recommending server in mobile terminal
CN103106259A (en) * 2013-01-25 2013-05-15 西北工业大学 Mobile webpage content recommending method based on situation
CN103106208A (en) * 2011-11-11 2013-05-15 ***通信集团公司 Streaming media content recommendation method and system in mobile internet

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120130848A1 (en) * 2010-11-24 2012-05-24 JVC Kenwood Corporation Apparatus, Method, And Computer Program For Selecting Items
CN102622390A (en) * 2011-10-11 2012-08-01 北京掌汇天下科技有限公司 Application recommending method and application recommending server in mobile terminal
CN103106208A (en) * 2011-11-11 2013-05-15 ***通信集团公司 Streaming media content recommendation method and system in mobile internet
CN103106259A (en) * 2013-01-25 2013-05-15 西北工业大学 Mobile webpage content recommending method based on situation

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
王立才等: ""上下文感知推荐***"", 《软件学报》, vol. 23, no. 1, 15 May 2012 (2012-05-15) *

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105989131A (en) * 2014-10-21 2016-10-05 株式会社日立制作所 Information search/presentation device and information search/presentation method
CN104731866A (en) * 2015-02-27 2015-06-24 湖南大学 Individual gourmet recommending method based on position
CN104731866B (en) * 2015-02-27 2020-05-19 湖南松桂坊电子商务有限公司 Personalized food recommendation method based on position
CN105205107A (en) * 2015-08-27 2015-12-30 湖南人文科技学院 Internet of Things data similarity processing method
CN105488213A (en) * 2015-12-11 2016-04-13 哈尔滨工业大学深圳研究生院 LBS-oriented individual recommendation method based on Markov prediction algorithm
CN105488213B (en) * 2015-12-11 2019-05-24 哈尔滨工业大学深圳研究生院 The personalized recommendation method based on Markov forecast techniques algorithm towards LBS
CN106846082A (en) * 2016-12-10 2017-06-13 江苏途致信息科技有限公司 Tourism cold start-up consumer products commending system and method based on hardware information
CN106846082B (en) * 2016-12-10 2021-07-30 江苏途致信息科技有限公司 Travel cold start user product recommendation system and method based on hardware information
CN108629609A (en) * 2017-03-22 2018-10-09 ***通信集团河北有限公司 The method and apparatus of reflexless terminal
CN108121803A (en) * 2017-12-22 2018-06-05 维沃移动通信有限公司 A kind of method and server of definite page layout

Also Published As

Publication number Publication date
CN103455555B (en) 2016-06-08

Similar Documents

Publication Publication Date Title
CN103455555A (en) Recommendation method and device based on mobile terminal similarity
CN112508118B (en) Target object behavior prediction method aiming at data offset and related equipment thereof
CN103970873A (en) Music recommending method and system
CN103678647A (en) Method and system for recommending information
WO2022142001A1 (en) Target object evaluation method based on multi-score card fusion, and related device therefor
CN104239373A (en) Document tag adding method and document tag adding device
CN104598539A (en) Internet event hot degree calculation method and terminal
CN112801425B (en) Method and device for determining information click rate, computer equipment and storage medium
CN104992348A (en) Method and device for displaying information
CN106227881A (en) A kind of information processing method and server
CN106708871A (en) Method and device for identifying social service characteristics user
CN106528896A (en) Database optimization method and apparatus
CN106257449B (en) A kind of information determines method and apparatus
CN112199374B (en) Data feature mining method for data missing and related equipment thereof
CN113256395B (en) Product recommendation method, device, equipment and storage medium based on recommendation graph network
CN113327154B (en) E-commerce user message pushing method and system based on big data
CN115481767A (en) Operation data processing method and device for power distribution network maintenance and computer equipment
CN114118858A (en) Power security intelligent evaluation method, apparatus, device, medium, and program product
CN108460475A (en) Poor student's prediction technique and device based on network playing by students behavior
CN112685516A (en) Multi-channel recall recommendation method and device, electronic equipment and medium
CN106909665A (en) A kind of patent retrieval method based on body
CN105912727A (en) Quick recommendation method in online social network labeling system
CN110019499A (en) The treating method and apparatus and electronic equipment of fast resampling
CN113901333B (en) Friend recommendation method integrating graph structure and text characteristics
CN118115025A (en) Business target data adjustment method, device, computer equipment, medium and product

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant