CN106952130A - Common user item based on collaborative filtering recommends method - Google Patents

Abstract

The invention discloses a method for recommending general items based on collaborative filtering. The method utilizes a user-based collaborative filtering method to calculate the similarity between users through Euclidean distance, obtain similar user sets, and further obtain different user profiles. Candidate recommendation set; and then calculate the recommendation score of the candidate recommendation set through the classification and feature assignment of the user's initial interest item attributes, so as to obtain a recommendation result with strong applicability and robustness, and help users more conveniently Get interesting content. Compared with the traditional recommendation methods based on demographics and content, the present invention pays more attention to the individual differences between users, and recommends by analyzing the user's historical behavior data and tapping the user's interest; therefore, recommending The results vary from person to person, more consideration is given to the user's personal preferences, and different method parameters can be customized for different recommendation scenarios.

Description

A general item recommendation method based on collaborative filtering

technical field

The invention relates to the technical field of information recommendation, in particular to a method for recommending general items based on collaborative filtering.

Background technique

Now that we have entered an era of data explosion, with the development of Web 2.0, the Web has become a platform for data sharing, so it will become more and more difficult for people to find the information they need in the massive data .

Under such circumstances, search engines (Google, Bing, Baidu, etc.) become the best way for everyone to quickly find target information. When users are relatively clear about their needs, it is very convenient to use search engines to quickly find the information they need through keyword searches. However, search engines cannot fully meet users' needs for information discovery, because in many cases, users do not know their needs clearly, or their needs are difficult to express with simple keywords. Or they need results that are more in line with their personal tastes and preferences, so there is a recommendation system that corresponds to a search engine, and everyone is used to calling it a recommendation engine.

With the emergence of recommendation engines, the way for users to obtain information has changed from simple data searches with specific goals to more advanced information discovery that is more in line with people's usage habits.

Today, with the continuous development of recommendation technology, recommendation engines have been used in e-commerce (E-commerce, such as Amazon, Dangdang) and some social-based social sites (including music, movie and book sharing, such as Douban, Mtime, etc.) All achieved great success. This further shows that in the Web2.0 environment, facing massive amounts of data, users need this kind of information discovery mechanism that is more intelligent and better understands their needs, tastes and preferences. Collaborative filtering recommendation algorithm is the earliest and well-known recommendation algorithm. The main functions are prediction and recommendation. The algorithm discovers user preferences through mining historical user behavior data, divides users into groups based on different preferences, and recommends products with similar tastes. Collaborative filtering recommendation algorithms are divided into two categories, namely, user-based collaborative filtering algorithm (user-based collaborative filtering) and item-based collaborative filtering algorithm (item-based collaborative filtering). To put it simply: people of a feather flock together and things of a feather flock together.

Contents of the invention

The purpose of the present invention is to provide a method for recommending general items based on collaborative filtering in order to solve the above-mentioned defects in the prior art.

The purpose of the present invention can be achieved by taking the following technical solutions:

A method for recommending general items based on collaborative filtering, the method for recommending general items comprises the following steps:

S1. The business side initializes and defines the attributes and feature values of recommended items and users, classifies user behaviors and sets the conditions of the initial search engine, and initiates a recommendation request to the recommendation engine APK after completing the initialization settings and accessing the recommendation engine APK , and send the initialized data set to the recommendation engine APK;

S2. The recommendation engine APK first summarizes the data that meets the search conditions according to the initial search engine of the business party to form a recommendable candidate set A, and obtains a similar user set through the similarity judgment for the recommendable candidate set A, and based on user-based collaboration Filter ideas and filter data to get recommendable result set B;

S3. Divide the recommendable result set B according to the attributes of the items, and divide the feature values of the attributes, and according to the proportion of each feature value in the user's interest set, obtain the proportion of each attribute feature value in the users of the recommendable result set B According to the weight of each most significant feature set to obtain the most sensitive attributes of user perception, according to the different reference weights of the most sensitive attributes, the user's interest set is sorted, and the preliminary recommendation list C is obtained;

S4. Send the preliminary recommendation result list C to the business party, and the business party rearranges it according to requirements, and finally obtains the recommendation result list D.

Further, the specific process of the step S2 is as follows:

S201. Divide user behaviors into K categories T ₁ to T _K , and assign weights w ₁ to w _k to these K categories of behaviors, and classify them into positive, negative, high, medium, and low according to different user behaviors. dimension, the value of the assignment vector w is w=[-2,-1,0,1,2,3];

S202. Obtain the accumulated value of the user's behavior and operation on the item to obtain the user's preference for the item H=∑w; when H>3, it is considered that the user is interested in the item;

S203. Using the Euclidean distance according to the preferences H of different users for each item Calculate the similarity between users:

When the similarity sim(x,y)>k between two users, where k is determined by the business side, that is, the two are considered similar to obtain a similar user set, and according to the idea of user-based collaborative filtering, each user’s Result set B can be recommended.

Further, the specific process of step S3 is as follows:

S301. According to the recommendable result set B, divide the attribute and characteristic value of the item, and set the attribute vector of the item as:

The eigenvalue vector of attribute S _i is:

S302. Construct an attribute feature matrix through the discrimination degree vector of each attribute and the candidate recommendation set;

For user A, the eigenvalue v _k of an item attribute S _i in the interested item set Q _A is The proportion of Then in the candidate recommendation set _QT , set the weight of the feature value v _k of the item attribute:

Also, when (i, k take arbitrary values), then it is considered that the discrimination degree of attribute S _x is the strongest; the discrimination degree of each attribute is taken as:

Therefore, when Q _A is given, the simple discrimination vector can be obtained:

After obtaining Q _T , the attribute characteristic matrix of the item can be obtained:

S303. According to the attribute feature matrix and the attribute discrimination degree vector, the item recommendation score vector of the candidate recommendation set Q _T can be obtained:

S304. Determine the preliminary recommendation result list C by sorting the recommendable result set B according to the obtained recommendation score vector.

Further, in the step S2, if the number of people who can recommend the candidate set A does not meet the requirement of the minimum recommended number of people, request the business party to expand the search conditions.

Further, the user's attributes and feature values are defined in the form of [attribute-value] key-value pairs.

Furthermore, the user's initial attributes are filled in by the user during registration, and the recommendation engine can perform feature division for the initial user based on the attribute values filled in by the user.

Compared with the prior art, the present invention has the following advantages and effects:

Compared with traditional demographic-based and content-based recommendation methods, the present invention draws on the ideas of collaborative filtering and content-based recommendation, and pays more attention to the individual differences among users. By analyzing the historical behavior data of users , mining the interests of users to make recommendations; therefore, the recommendation results vary from person to person, more consideration is given to the user's personal preferences, and different method parameters can be customized for different recommendation scenarios.

Description of drawings

Fig. 1 is a flow chart of a method for recommending general items based on collaborative filtering disclosed in the present invention.

detailed description

In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments It is a part of embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

Example

As shown in FIG. 1 , this embodiment discloses a method for recommending general items based on coordinated filtering. The method for recommending general items draws on ideas of collaborative filtering and content-based recommendation to design a recommendation engine for general items. Among them, general-purpose items include but are not limited to: books, music, and movies. The difference between different general-purpose items is that the measurement standard of preference and the difference in attribute characteristics of items can be adjusted by the business side according to different recommendation scenarios.

The item recommendation method specifically includes the following steps:

S1. The business side initializes and defines the attributes and feature values of recommended items and users. The definition method is [attribute-value] key-value pairs, such as [height-170CM]; classify user behavior and set initial search engine conditions , after completing the initialization settings and accessing the recommendation engine APK, initiate a recommendation request to the recommendation engine APK, and send the initialized data set to the recommendation engine APK.

In a specific application, the user's initial attributes are filled in by the user during registration, and the recommendation engine can classify the characteristics of the initial user according to the attribute values filled in by the user.

S2. The recommendation engine APK first summarizes the data that meets the search conditions according to the initial search engine of the business party to form a recommendable candidate set A, and obtains a similar user set through the similarity judgment for the recommendable candidate set A, and based on user-based collaboration Filter ideas and perform data screening to obtain recommendable result set B; if the number of recommendable candidate set A does not meet the minimum number of recommenders, request the business side to expand the search conditions.

Wherein, the specific process of performing data screening to obtain the recommendable result set B is as follows:

S201. Divide user behaviors into K categories T ₁ to T _K , and assign weights w ₁ to w _k to these K categories of behaviors, and classify them into positive, negative, high, medium, and low according to different user behaviors. dimension, the value of the assignment vector w is w=[-2,-1,0,1,2,3];

In a specific application, the characteristics of different behaviors are defined according to the user's online behavior data.

S202. Obtain the accumulated value of the user's behavior and operation on the item to obtain the user's preference for the item H=∑w. When H>3, it is considered that the user is interested in the item;

S203. Using the Euclidean distance according to the preferences H of different users for each item

Calculate the similarity between users:

When the similarity sim(x,y)>k between two users, where k is determined by the business side, that is, the two are considered similar to obtain a similar user set, and according to the idea of user-based collaborative filtering, each user’s Result set B can be recommended.

S3. Divide the recommendable result set B according to the attributes of the items, and divide the feature values of the attributes, and according to the proportion of each feature value in the user's interest set, obtain the proportion of each attribute feature value in the users of the recommendable result set B According to the most significant feature sets to obtain the most sensitive attribute of user perception, according to the different reference weights of the most sensitive attribute, the user's interest set is sorted, and the preliminary recommendation list C is obtained.

The specific process of this step is as follows:

S301. According to the recommendable result set B, divide the attribute and characteristic value of the item, and set the attribute vector of the item as:

The eigenvalue vector of attribute S _i is:

S302. Construct an attribute feature matrix through the discrimination degree vector of each attribute and the candidate recommendation set;

For user A, the feature value v _k of an item attribute S _i in the candidate recommendation set Q _T is The proportion of Then in the candidate recommendation set _QT , set the weight of the feature value v _k of the item attribute:

Also, when (i, k take any possible value)

Then it is considered that the distinguishing degree of attribute S _x is the strongest; the distinguishing degree of each attribute is taken as:

Therefore, when Q _A is given, the simple discrimination vector can be obtained:

After obtaining Q _T , the attribute characteristic matrix of the item can be obtained:

S303. According to the attribute feature matrix and the attribute discrimination degree vector, the item recommendation score vector of the candidate recommendation set Q _T can be obtained:

S304. Determine the preliminary recommendation result list C by sorting the recommendable result set B according to the obtained recommendation score vector.

S4. The preliminary recommendation result list C is sent to the business party, and the business party performs appropriate rearrangement according to requirements, and finally obtains the recommendation result list D, which is the final result.

In summary, the present invention draws on the ideas of collaborative filtering and content-based recommendation, pays attention to the individual differences among users, analyzes the user's historical behavior data, and taps the user's interests to make recommendations; therefore, the recommendation results vary from person to person , more consideration is given to the user's personal preferences, and different method parameters can be customized for different recommendation scenarios.

The above-mentioned embodiment is a preferred embodiment of the present invention, but the embodiment of the present invention is not limited by the above-mentioned embodiment, and any other changes, modifications, substitutions, combinations, Simplifications should be equivalent replacement methods, and all are included in the protection scope of the present invention.

Claims (8)

1. A general article recommendation method based on collaborative filtering, characterized in that, the general article recommendation method comprises the following steps: S1. The business side initializes and defines the attributes and feature values of recommended items and users, classifies user behaviors and sets the conditions of the initial search engine, and initiates a recommendation request to the recommendation engine APK after completing the initialization settings and accessing the recommendation engine APK , and send the initialized data set to the recommendation engine APK; S2. The recommendation engine APK first summarizes the data that meets the search conditions according to the initial search engine of the business party to form a recommendable candidate set A, and obtains a similar user set through the similarity judgment for the recommendable candidate set A, and based on user-based collaboration Filter ideas and filter data to get recommendable result set B; S3. Divide the recommendable result set B according to the attributes of the items, and divide the feature values of the attributes, and according to the proportion of each feature value in the user's interest set, obtain the proportion of each attribute feature value in the users of the recommendable result set B According to the weight of each most significant feature set to obtain the most sensitive attributes of user perception, according to the different reference weights of the most sensitive attributes, the user's interest set is sorted, and the preliminary recommendation list C is obtained; S4. Send the preliminary recommendation result list C to the business party, and the business party rearranges it according to requirements, and finally obtains the recommendation result list D. 2. The general article recommendation method based on collaborative filtering according to claim 1, wherein the specific process of the step S2 is as follows: S201. Divide user behaviors into K categories T ₁ to T _K , and assign weights w ₁ to w _k to these K categories of behaviors, and classify them into positive, negative, high, medium, and low according to different user behaviors. dimension, the value of the assignment vector w is w=[-2,-1,0,1,2,3]; S202. Obtain the accumulated value of the user's behavior and operation on the item to obtain the user's preference for the item H=∑w; S203. Using the Euclidean distance according to the preferences H of different users for each item Calculate the similarity between users: When the similarity sim(x,y)>k between two users, where k is determined by the business side, that is, the two are considered similar to obtain a similar user set, and according to the idea of user-based collaborative filtering, each user’s Result set B can be recommended. 3. The general article recommendation method based on collaborative filtering according to claim 1, wherein the specific process of the step S3 is as follows: S301. According to the recommendable result set B, divide the attribute and characteristic value of the item, and set the attribute vector of the item as: The eigenvalue vector of attribute S _i is: S302. Construct an attribute feature matrix through the discrimination degree vector of each attribute and the candidate recommendation set; For user A, the eigenvalue v _k of an item attribute S _i in the interested item set Q _A is The proportion of Then in the candidate recommendation set _QT , set the weight of the feature value v _k of the item attribute: $\mathrm{<span\; class="notranslate">\; q</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; q</span>}}_{{\mathrm{<span\; class="notranslate">\; v</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}}^{\mathrm{<span\; class="notranslate">\; i</span>}}$ Also, when (i, k take arbitrary values), then it is considered that the discrimination degree of attribute S _x is the strongest; the discrimination degree of each attribute is taken as: ${\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; max</span>}{\mathrm{<span\; class="notranslate">\; q</span>}}_{\mathrm{<span\; class="notranslate">\; v</span>}}^{\mathrm{<span\; class="notranslate">\; w</span>}}$ Therefore, when Q _A is given, the simple discrimination vector can be obtained: After obtaining Q _T , the attribute characteristic matrix of the item can be obtained: ${\mathrm{<span\; class="notranslate">\; A</span>}}_{{\mathrm{<span\; class="notranslate">\; Q</span>}}_{\mathrm{<span\; class="notranslate">\; e</span>}}}<span\; class="notranslate">\; =</span>\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}\mathrm{<span\; class="notranslate">\; .........</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}\\ {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}^{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}\mathrm{<span\; class="notranslate">\; .........</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}\\ \mathrm{<span\; class="notranslate">\; ...................................</span>}\\ \mathrm{<span\; class="notranslate">\; ...................................</span>}\\ {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}\mathrm{<span\; class="notranslate">\; .........</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}\end{array}\right]<span\; class="notranslate">\; ;</span>$ S303. According to the attribute feature matrix and the attribute discrimination degree vector, the item recommendation score vector of the candidate recommendation set Q _T can be obtained: S304. Determine the preliminary recommendation result list C by sorting the recommendable result set B according to the obtained recommendation score vector. 4. the general item recommendation method based on collaborative filtering according to claim 1, characterized in that, If the number of people who can recommend the candidate set A in the step S2 does not meet the requirement of the minimum number of recommended people, request the business party to expand the search conditions. 5. The method for recommending general items based on collaborative filtering according to claim 1, wherein the user's attributes and feature values are defined in the form of [attribute-value] key-value pairs. 6. The general item recommendation method based on collaborative filtering according to claim 1, wherein the user's initial attributes are filled in by the user when registering, and the recommendation engine can perform feature division for the initial user according to the attribute values filled in by the user. 7. the method for recommending general items based on collaborative filtering according to claim 2, characterized in that, When the user's preference for an item is H>3, it is considered that the user is interested in the item. 8. The general item recommendation method based on collaborative filtering according to any one of claims 1 to 7, characterized in that, The general-purpose items include: books, music, and movies. The difference between different general-purpose items is that the measurement standard of preference and the difference in attribute characteristics of the items can be adjusted by the business side according to different recommendation scenarios.