WO2017036333A1

WO2017036333A1 - Interaction event-based webpage item recommendation method and device

Info

Publication number: WO2017036333A1
Application number: PCT/CN2016/096583
Authority: WO
Inventors: 李玉龙
Original assignee: 阿里巴巴集团控股有限公司; 李玉龙
Priority date: 2015-09-02
Filing date: 2016-08-24
Publication date: 2017-03-09
Also published as: CN106484747A

Abstract

Provided in embodiments of the present application are an interaction event-based webpage item recommendation method and device. The method comprises: generating, according to interaction information of interaction events of a plurality of users with respect to a webpage item, relevance degrees among the interaction events and a target interaction event in the interaction events; for an interaction event of a current user, retrieving, according to relevance degrees among the interaction event and a plurality of target interaction events, at least one target webpage item corresponding to the target interaction events; and recommending the target webpage items to the current user. The present application avoids the problem in which webpage items cannot be recommended effectively due to lack of ratings provided by the user. Also, using interaction events with respect to a webpage item as the basis of relevance computation, webpage items a user is interested in can be objectively and accurately recommended to the user. Furthermore, only computing a relevance degree between a target interaction event and an interaction event in a relevance computation can decrease the volume of data in the relevance computation and save the computational resource and storage resource of a server.

Description

Web site project recommendation method and device based on interactive event

Technical field

The present application relates to the field of Internet technologies, and in particular, to a webpage item recommendation method based on an interaction event and a webpage item recommendation apparatus based on an interaction event.

Background technique

With the continuous development of Internet technology, more and more users browse various products on the webpage.

In order to recommend to the user a webpage item that the user may be interested in, the website provider collects a large number of user evaluation points for each item, and uses the evaluation filter to calculate the relevance matrix between the items through the collaborative filtering algorithm. For a certain user, the evaluation score of the partial merchandise that the user has made, and the degree of relevance of the partial merchandise to other merchandises are calculated, and the recommendation degree of the other merchandise is calculated and recommended to the user accordingly.

At present, there are three problems in this product recommendation method: First, in practical applications, users may not want to spend time and effort to provide evaluation points of products, resulting in the inability to conduct effective product recommendation due to lack of evaluation points; secondly, users are targeting products. The evaluation has a subjective tendency, and the product recommendation is based on subjective evaluation points. It is impossible to objectively and accurately recommend the products that meet the user's interests to the user; finally, with the increasing amount of data in the Internet, the collaborative filtering algorithm is used. Commodity recommendation will cause the data volume of the correlation matrix in the current algorithm to expand rapidly, resulting in overloading of the server, overloading of the calculation processing, and even causing the server to crash.

Summary of the invention

In view of the above problems, embodiments of the present application are proposed to provide an interactive event based webpage item recommendation method and a corresponding interactive event based webpage item recommendation apparatus that overcome the above problems or at least partially solve the above problems.

In order to solve the above problem, the present application discloses a webpage item recommendation method based on an interaction event, including:

Generating a correlation between the interaction event and a target interaction event in the interaction event according to interaction information of interaction events of the plurality of users for the webpage item;

Extracting, according to the correlation event of the interaction event and the plurality of target interaction events, the target webpage item corresponding to the at least one target interaction event;

The target webpage item is recommended to the current user.

Preferably, the step of generating the correlation between the interaction event and the target interaction event in the interaction event according to the interaction information of the interaction events of the plurality of users for the webpage item comprises:

Generating an interest score of the interaction event according to an event attribute of the interaction event;

Calculating a correlation between the interaction event and the target interaction event according to the interest score of the interaction event.

Advantageously, said interactivity event comprises a browsing event for a webpage item, and/or a favorite event, and/or adding a shopping cart event, and/or a purchase event, said generating an interest of said interactivity event based on an event attribute of the interactivity event The steps of the score are:

Taking the product of the browsing event, the favorite event, the added shopping cart event, the number of interaction events N in the purchase event, and the preset parameter as the interest degree score of the interaction event is M.

Preferably, the step of calculating the correlation between the interaction event and the target interaction event according to the interest score of the interaction event comprises:

The interaction event of the same user and the target interaction event are combined into an event pair, and the product of the interaction event value of the event pair and the interest degree score of the target interaction event is calculated;

Calculating the interest score of the interaction event in the same event pair of the plurality of users, the interest score of the target interaction event, and the product of the interest event score of the interaction event and the target interaction event, The degree to which an interaction event interacts with a target interaction event.

Preferably, the product of the interest score of the interaction event according to the same event pair of the plurality of users, the interest score of the target interaction event, and the interest score of the interaction event and the target interaction event are calculated. The correlation between the interaction event and the target interaction event in the event pair is as follows:

The same event pair for multiple users, the sum of the squares of the interest scores of the interaction events in the same event pair, the sum of the squares of the interest scores of the target interaction events, and the interaction events and target interaction events The sum of the products of the interest scores and the corresponding interactions in the event pairs The multidimensional vector value corresponding to the event and target interaction events respectively;

A cosine value between the multidimensional vector values is calculated and used as the correlation.

The interaction event of the same user and the target interaction event are combined into an event pair, and the interest scores of the interaction event and the target interaction event of the same event pair of multiple users are respectively composed of the interaction event interest degree score set and the target interaction event. Interest score set;

And calculating, as the correlation, a JCd coefficient of the set of interactivity event interest scores and the target interaction event interest score set.

Preferably, when the current user has an interaction event, the interaction event for the current user extracts the target webpage item corresponding to the at least one target interaction event according to the correlation between the interaction event and the plurality of target interaction events. The steps include:

When capturing an interaction event generated by the user, searching for a correlation between the one interaction event and multiple target interaction events;

The webpage item corresponding to the target interaction event in the first preset range is extracted as the target webpage item.

Preferably, when the current user has multiple interaction events, the interaction event for the current user extracts a target webpage item corresponding to the at least one target interaction event according to the correlation between the interaction event and the plurality of target interaction events. The steps include:

Searching for the relevance of the target interaction event to the plurality of interaction events for each target interaction event;

Calculating a recommendation degree that identifies a correlation between the target interaction event and the plurality of interaction events according to the relevance of the target interaction event to each interaction event;

The webpage item corresponding to the target interaction event in the second preset range is extracted as the target webpage item.

Preferably, the web page item comprises a transaction object, and/or a video, and/or audio, and/or an electronic book.

The application also discloses a webpage item recommendation device based on an interaction event, comprising:

a correlation generation module, configured to generate, according to interaction information of interaction events of the plurality of users for the webpage item, a correlation degree between the interaction event and the target interaction event in the interaction event;

An extraction module, configured to extract, according to the interaction event of the current user, a target webpage item corresponding to the at least one target interaction event according to the correlation between the interaction event and the plurality of target interaction events;

a recommendation module for recommending the target webpage item to the current user.

Preferably, the correlation generation module includes:

a score of interest generation sub-module, configured to generate an interest score of the interaction event according to an event attribute of the interaction event;

The correlation calculation sub-module is configured to calculate a correlation between the interaction event and the target interaction event according to the interest degree score of the interaction event.

Preferably, the interaction event comprises a browsing event for a webpage item, and/or a favorite event, and/or adding a shopping cart event, and/or a purchase event, the interest score generation sub-module comprising:

a score of interest M to generate a subunit for using the product of the browsing event, the favorite event, the added shopping cart event, the number of interaction events N in the purchase event, and a preset parameter as The interest score of the interactive event is M.

Preferably, the correlation calculation submodule comprises:

The event pair component unit is configured to form an event pair of the interaction event and the target interaction event of the same user, and calculate a product of the interaction event value of the event pair and the interest degree score of the target interaction event;

a correlation calculation sub-unit for multiplying the interest score of the interaction event in the same event pair of the plurality of users, the interest score of the target interaction event, and the interest score of the interaction event and the target interaction event And calculating a correlation between the interaction event in the event pair and the target interaction event.

Preferably, the correlation calculation subunit is specifically configured to:

Preferably, the correlation calculation submodule comprises:

The interest score group constitutes a sub-unit, which is used to form an event pair of the same user's interaction event and the target interaction event, and respectively adopts the same event pair of the multiple user pairs and the interest degree score of the target interaction event respectively. An interactive event interest score set and a target interaction event interest score set;

a Jaccard coefficient calculation subunit, configured to calculate and select the Jaccard coefficient of the interactivity event interest score set and the target interactivity interest score set.

Preferably, when the current user has an interaction event, the extraction module includes:

a first correlation finding submodule, configured to: when capturing an interaction event generated by the user, searching for a correlation between the one interaction event and multiple target interaction events;

The first extraction sub-module is configured to extract a webpage item corresponding to the target interaction event in the first preset range as the target webpage item.

Preferably, when the current user has multiple interaction events, the extraction module includes:

a second relevance search sub-module, configured to search for a correlation between the target interaction event and the multiple interaction events for each target interaction event;

a recommendation calculation sub-module, configured to calculate a recommendation degree that identifies a correlation between the target interaction event and the plurality of interaction events according to the correlation between the target interaction event and each interaction event;

The second extraction sub-module is configured to extract, as the target webpage item, a webpage item corresponding to the target interaction event in the second preset range.

Embodiments of the present application include the following advantages:

The embodiment of the present application calculates the relevance of the interaction event of the webpage item by the user, and avoids the problem that the effective webpage item recommendation cannot be performed because the user does not provide the evaluation score; and, the interaction event for the webpage item is used as the correlation calculation. Based on the basis, the website items that meet the user's interests can be objectively and accurately recommended to the user; further, relevant In the degree calculation, only the correlation between the target interaction event and the interaction event is calculated, the amount of data in the correlation calculation is reduced, and the computing resources and storage resources of the server are saved.

DRAWINGS

1 is a flow chart of steps of Embodiment 1 of a webpage item recommendation method based on an interaction event according to the present application;

2 is a flow chart of steps of Embodiment 2 of a webpage item recommendation method based on an interaction event according to the present application;

3 is a structural block diagram of an embodiment of a webpage item recommending apparatus based on an interactive event according to the present application;

4 is a flow chart of calculating correlation between a target interaction event and an interaction event by using Map Reduce in Embodiment 2 of the present application.

detailed description

The above described objects, features and advantages of the present application will become more apparent and understood.

Referring to FIG. 1 , a flow chart of a step 1 of a method for recommending a webpage item based on an interactive event according to the present application is shown, which may specifically include the following steps:

Step 101: Generate, according to interaction information of interaction events of the plurality of users for the webpage item, a correlation degree between the interaction event and the target interaction event in the interaction event.

It should be noted that the above webpage items may include various transaction objects, videos, audio or electronic books in the webpage. Users can get the corresponding information or services through different webpage projects.

In the actual application scenario, the user will generate different interaction events for different webpage projects according to the degree of interest. For example, if a user is interested in a transaction object in a webpage, an interactive event that browses the transaction object may be, after multiple browsing, the user's interest in the transaction object is increased, and a transaction object is generated. The interactive event, or the interactive event placed in the virtual shopping cart provided by the webpage, the end user is quite interested in the transaction object, and may even The interactive event for the purchase. It can be seen from the actual scenario that the different interaction events of the user for the same webpage item can reflect the degree of interest of the user on the webpage item. Therefore, the webpage item can be combined with the interaction behavior of the user for the webpage item to form a webpage item. Interactive event.

At the same time, there is a correlation between the user's various interaction events for different webpage projects. For example, an interaction event in which the same user browses the transaction object A and an interaction event in which the transaction object B is purchased may be understood as a user who has a certain interest in the transaction object A and browses, and may also be the transaction object B. Generated a very strong interest and eventually made an interactive event for the purchase. As can be seen from this example, there is an implicit correlation between the interactive events of the web page project. This implicit correlation can reflect the user's interest in different web page items.

The degree of interest reflected by the user's interaction event of the webpage item can be quantified into a specific interest score M by using the interaction information of the interaction event. The interaction information described above may include at least one of an interaction event category of an interaction event, an interaction event execution count, an interaction event occurrence time, and an interaction event duration. For example, if the number of times the user views the webpage item A is three times, the interest score M of the browsing event of the webpage item A is three. Of course, those skilled in the art can set weights for different interaction events according to actual needs to distinguish the importance of different interaction events.

Correlation between interaction events can be quantified to relevance using a specific interest score. In a specific operation, the interest degree scores of the same two interactive events of a large number of users may respectively form a multi-dimensional vector of two interest degree scores, and the cosine value between the two multi-dimensional vectors is calculated by the cosine value formula, the cosine value This is the correlation between two interactive events.

Of course, those skilled in the art can also calculate the correlation between interaction events according to actual needs. For example, if a user has performed an interaction event, the interest score can be defined as 1, if not executed. However, the interest score can be defined as 0, and the interest scores of the same two interactive events of a large number of users respectively form a multi-dimensional vector of two interest scores, and calculate the Jacques between the two multi-dimensional vectors. The German coefficient, which is the correlation between two interactions.

However, not all correlations between interaction events are worthy of attention and calculation, because only part of the interaction events in the user's interaction events with web projects represent strong user interest. Interestingly, for example, the purchased interactive event represents the user's most interest in the web project, and the interactive event may be simply that the user simply understands that the browsed web project does not meet the user's interest. The purpose of the webpage project recommendation is to recommend the webpage item that is most likely to cause user interest and is most in line with the user's interest. Therefore, it is only necessary to pay attention to the target interaction event in the interaction event, and calculate the correlation between the target interaction event and other interaction events. Just fine. In a specific implementation, the interaction information of the interaction event may be used, the target interaction event and the non-target interaction event in the interaction event are distinguished, and the correlation between the target interaction event and other interaction events is calculated, thereby obtaining the target event of the webpage project and another The relevance of the target interactive event or non-target interactive event of the web project. As a preferred example of the present invention, the calculated correlation can be saved in a database.

A person skilled in the art may use one or more interaction events in the interaction event as the target interaction event according to actual needs, which is not limited by the embodiment of the present invention.

Step 102: For the interaction event of the current user, extract the target webpage item corresponding to the at least one target interaction event according to the correlation between the interaction event and the multiple target interaction events.

Step 103: Recommend the target webpage item to the current user.

In an actual application, the interaction event of the current user to a webpage item has multiple correlations with target interaction events of other webpage items, and may target one or more related target interaction events within a preset range. The corresponding landing page item is extracted and recommended to the current user.

Compared with the current method for calculating the correlation between webpage items based on the user's evaluation of the webpage item, the present application combines the webpage item with the interaction behavior for the webpage item to form an interaction event, and performs correlation calculation based on the interaction event. The problem that the effective webpage item recommendation cannot be performed because the user does not provide the evaluation score is avoided; and the interactive event for the webpage item is used as the basis of the correlation calculation, and the webpage item that meets the user interest can be objectively and accurately recommended to the webpage item. Further, in the correlation calculation, only the correlation between the target interaction event and the interaction event is calculated, the amount of data in the correlation calculation is reduced, and the computing resources and storage resources of the server are saved.

Referring to FIG. 2, a flow chart of the steps of the second embodiment of the webpage item recommendation method based on the interaction event is shown in the following.

Step 201: Generate an interest score of the interaction event according to an event attribute of the interaction event.

The event attribute described above may include at least one of an interaction event execution number of an interaction event, an interaction event occurrence time, and an interaction event duration. A specific interest score for the interaction event can be generated based on various event attributes of the interaction event. In an actual application, when the user does not have an interaction event, the interest score of the interaction event of the user may be 0.

As a preferred example of an embodiment of the present invention, the interaction event includes a browsing event for a webpage item, and/or a favorite event, and/or adding a shopping cart event, and/or a purchase event, the step 201 may include the following sub-steps :

Sub-step S11, the product of the browsing event, the favorite event, the added shopping cart event, the number of interaction events N in the purchase event, and the preset parameter is used as the interest score of the interaction event. For M.

User interaction events for webpage items typically include interactive events such as browsing, collecting, adding shopping carts, purchasing, and the like. On the one hand, the number of times a user performs an interaction event typically reflects the level of interest in the web project. On the other hand, different interaction events also have different importance levels. For example, the purchase event is the most important event. Adding a shopping cart is the second, and browsing is a relatively common event. The importance of the event is given as a preset parameter as the weight of the event. The interest score can be made to more reasonably reflect the user's interest in the web project.

Step 202: Calculate a correlation between the interaction event and the target interaction event according to the interest degree score of the interaction event.

As a preferred example 1 of the embodiment of the present invention, the step 202 may include the following sub-steps:

Sub-step S21, the interaction event of the same user and the target interaction event are combined into an event pair, and the product of the interaction event of the event pair and the interest score of the target interaction event is calculated.

The interaction event of the same user and the target interaction event may be combined into an event pair, and the interest score of the interaction event in the event pair is multiplied by the interest score of the target interaction event. When there is no target interaction event, that is, the interest degree score of the target interaction event is 0, the result of the multiplication is also 0. Because only the degree of relevance of the target interaction event and the interaction event needs to be concerned, when there is no target interaction event, the multiplication result of the event pair 0 can save subsequent correlation calculation.

Sub-step S22, according to the interest score of the interaction event in the same event pair of the plurality of users, the interest score of the target interaction event, and the interest score of the interaction event and the target interaction event The product is used to calculate the correlation between the interaction event in the event pair and the target interaction event.

The correlation between the interaction event and the target interaction event may be a cosine value of the interest scores of the two, or a JCD coefficient of the interest scores of the two. A person skilled in the art can calculate the correlation by using any algorithm that can calculate the degree of similarity between the interaction event and the interest score of the target interaction event, as the case may be.

As a preferred example of the embodiment of the present invention, the sub-step S22 may specifically include:

Sub-step S22-1, calculating the sum of the squares of the interest scores of the interactive events in the same event pair, the sum of the squares of the interest scores of the target interaction events, and the interaction for the same event pair of the plurality of users The sum of the products of the interest scores of the event and the target interaction event, and correspondingly forming a multi-dimensional vector value corresponding to the interaction event and the target interaction event in the event pair respectively;

Sub-step S22-2, calculating a cosine value between the multi-dimensional vector values as the correlation.

The same event pair for multiple users, multiple interest scores for the same target interaction event, and multiple interest scores for the same interaction event, forming a multi-dimensional vector corresponding to the target interaction event and the interaction event respectively Value, the cosine value formula is used to calculate the cosine between the multi-dimensional vector values, and the cosine value is used as the correlation between the target interaction event and the interaction event in the event pair.

In practical applications, Map Reduce can be used to calculate the correlation degree in combination with the cosine degree formula. Referring to the flowchart of calculating the correlation between the target interaction event and the interaction event by using Map Reduce, the correlation calculation is implemented by five Map Reduce processes.

The cosine value formula can be:

Among them, S _ij is the correlation between the target interaction event and the interaction event.

Map Reduce 1, collect all the interactive events of the same user, form an event pair with the target interaction event, and calculate the product of the interest score of the event pair and the target interaction event. Where <e _i , e _j > is the event pair formed by the interaction event and the target interaction event; r _ui is the interest degree score of the interaction event; r _uj is the interest degree score of the target interaction event.

Map Reduce 2, which calculates the sum of the product of the target interaction event and the interest score of the interaction event for the same event pair of multiple users. among them,

The sum of the product of the target interaction event and the interest score of the interaction event for the same event pair of multiple users.

Map Reduce 3, which calculates the sum of the squares of the interest scores of the interactive events in the same event pair for multiple users. among them,

The sum of the squares of interest scores for the interaction events of the same event pair for multiple users.

Map Reduce 4, which calculates the sum of the squares of the interest scores of the target interaction events for the same event pairs of different users. among them,

The sum of the squares of interest scores for the target interaction events for the same event pair for multiple users.

Map Reduce 5, which calculates the cosine of the interest scores of the target interaction events and interaction events for the same event pairs of multiple users. among them,

The opening of the sum of the squares of the interest scores of the interactive events for the same event pairs of different users;

The square root of the sum of the squares of the interest scores of the target interaction events for the same event pairs of different users.

As a preferred example 2 of the embodiment of the present invention, the step 202 may include the following sub-steps:

Sub-step S31, the interaction event and the target interaction event of the same user are combined into an event pair, and the interest event scores of the interaction event and the target interaction event of the same event pair of the multiple users are respectively composed of the interaction event interest degree score set. Interacting with the target event scores.

Sub-step S32, calculating, as the correlation degree, the Jaccard coefficient of the interaction event interest score set and the target interaction event interest score set.

In an actual application, when a user does not generate an interaction event for a webpage item, the interaction The interest score of the event may be 0. If there is an interaction event, the interest score of the interaction event may be 1. The multi-dimensional vector value of all the dimensions corresponding to the interaction event and the target interaction event is 1 or 0, and the calculation dimension is calculated by using the same interest event of the plurality of users and the interest score of the target interaction event and the target interaction event is 1 or 0. The Jaccard coefficients between two multi-dimensional vector values of 1 or 0, and the calculated JCD coefficient is used as the correlation between the interaction event and the target interaction event. Calculating the correlation using the Jaccard coefficient is suitable for calculating the correlation between interaction events that only need to pay attention to the existence or not.

Step 203: For the interaction event of the current user, extract the target webpage item corresponding to the at least one target interaction event according to the correlation between the interaction event and the plurality of target interaction events.

The user may generate one or more interaction events for the webpage items in the webpage, and according to the interaction events generated by the user, the target webpage items corresponding to the target interaction events related to the interaction events may be extracted. In actual applications, there may be a large number of related target interaction events, and some of the target interaction events are less correlated with user-generated interaction events, so it is not necessary to extract the webpage items corresponding to the target interaction events in full. The target interaction event is sorted according to the degree of relevance of the interaction event, and the webpage item corresponding to the previous target interaction event is extracted, and a range of relevance degrees may be preset, and the target interaction event corresponding to the range is matched. Web page item extraction. A person skilled in the art may also extract the target webpage item by using other methods, for example, by calculating the recommendation degree of the target interactive event, and recommending the webpage item corresponding to the target interactive event as the target webpage item to the user when the recommendation degree meets the preset range. .

As a preferred example 1 of the embodiment of the present invention, when an interaction event occurs in the current user, the step 203 may include the following sub-steps:

Sub-step S41, when an interaction event generated by the user is captured, the correlation between the one interaction event and the multiple target interaction events is searched for.

Sub-step S42, extracting a webpage item corresponding to the target interaction event in the first preset range as the target webpage item.

When it is captured that the user generates an interaction event for a webpage item, and has a plurality of correlations with the target interaction events of other webpage items, the target corresponding to the one or more relevance target interaction events within the preset range may be Web page item extraction for subsequent web page project recommendation. Pass pre The scope can be targeted to recommend a webpage item corresponding to a highly relevant target interaction event.

The preset range can be set by a person skilled in the art according to actual conditions. For example, the preset range can be adjusted for the number of webpage items corresponding to the extracted target interaction event, and the number of webpage items corresponding to the target interactive event that can be extracted is too small. , you can lower the lower limit of the preset range to ensure that there are enough webpage items to recommend to the current user.

In practical applications, a large cluster of distributed file systems (Hadoop Distributed File System, Hadoop) can be used in advance to calculate the correlation between interaction events and target interaction events. When the user is captured for a web page project, an interaction event can be generated in the database. Searching for the target interaction event related to the interaction event and its relevance, and extracting the target webpage item corresponding to the target interaction event whose relevance is in accordance with the preset scope.

As a preferred example 2 of the embodiment of the present invention, when multiple interaction events occur in the current user, the step 203 may include the following sub-steps:

Sub-step S51, searching for the relevance of the target interaction event and the plurality of interaction events for each target interaction event.

Sub-step S52, calculating a recommendation degree that identifies a correlation between the target interaction event and the plurality of interaction events according to the relevance of the target interaction event to each interaction event.

Sub-step S53, extracting a webpage item corresponding to the target interaction event in the second preset range as the target webpage item.

In practical applications, a user may generate multiple interaction events, which may be related to the same target interaction event, use interest scores of multiple interaction events, and interact with multiple interaction events. The relevance of the event can be used to calculate the recommendation degree of the target interaction event. When the recommendation degree meets the preset range, the webpage item corresponding to the target interaction event can be used as the target webpage item. By using the relevance as the weight to calculate the recommendation degree of the target interaction event, and using the recommendation degree as the basis of the recommendation of the webpage item, the webpage item recommended to the user is more in line with the user's interest.

Step 204: Recommend the target webpage item to the current user.

The web page items described above may include transaction objects, and/or video, and/or audio, and/or electronic books.

In the embodiment of the present application, the interest degree score of the interaction event is generated according to the event attribute of the interaction event, and the user's interest level of the webpage item can be objectively quantized into the interest degree score, and the webpage item recommendation is performed based on the interest degree score. The webpage project is more in line with the user's interest.

Moreover, the target webpage item corresponding to the target interaction event whose relevance is within the preset range is recommended to the user, which improves the flexibility of the webpage item recommendation.

In order to enable a person skilled in the art to better understand the present application, the following describes, by way of a specific example, the interactive event in the present application according to the interaction information of the interaction events of the plurality of users for the webpage item, and the interaction event and the target interaction event in the interaction event. The method of relevance:

In a three users A, B and C on page two scenarios project x and y, browse event for the web project x and y of web projects, collection events, placing shopping cart events, shopping events were E ₁ , E ₂ , E ₃ , E ₄ , E ₅ , E ₆ , E ₇ and E ₈ . In this example, the interactive shopping as a target event, so E ₄ and E ₈ for examples of target interaction events.

The product of the event, the favorite event, the add shopping cart event, the number of interaction events N in the purchase event, and the preset parameter may be browsed as the interest score of the interaction event is M.

The number of executions of user A, B, and C for interactive events such as browsing events, favorite events, shopping cart events, shopping events, etc. for webpage items x and y are E _{Ax Browse} = 3, E _{Ax Collection} = 2, E _Ax _{placed Shopping Cart} =1, E _{Ax Shopping} =2, E _{Ay Browse} =2, E _{Ay Collection} =1, E _{Ay Place Shopping Cart} =1, E _{Ay Shopping} =0, E _{Bx Browse} =5, E _{Bx Favorites} =1, E _{Bx Place Shopping Cart} =1, E _{Bx Shopping} =1, E _{By Browse} =2,E _{By Favorites} =2,E _{By Place Shopping Cart} =2,E _{By Shopping} =1,E _{Cx Browse} =2,E _{Cx Favorites} =1 , E _{Cx Place Shopping Cart} =1, E _{Cx Shopping} =1, E _{Cy Browse} =3, E _{Cy Collection} =2, E _Cy _{Place Shopping Cart} =1 and E _{Cy Shopping} =1.

The preset parameters of the browsing event, the favorite event, the shopping cart event, and the shopping event are 1, 2, 3, and 4, respectively, and the product of the number of executions of the interactive event and the preset parameter of the interactive event is used as the interest score of the interactive event. The values, and thus the scores of interest for each interaction event are shown in the following table:

The interaction event of the same user and the target interaction event are combined into an event pair, and the product of the interaction event of the event pair and the interest score of the target interaction event is calculated. For the sake of simplicity, it is only necessary to calculate the correlation between E ₄ and E ₅ , E ₂ and E ₈ , E ₁ and E ₄ , and S ₁₄ , S ₅₄ and S ₂₈ can be calculated using the cosine formula, wherein the specific calculation process is as follows :

User A's interaction event E ₅ and target interaction event E _{4 are} composed of event pairs <E ₅ , E ₄ >, and event pairs <E ₅ , E ₄ >, and the product of interest scores of E ₅ and E ₄ is r _A5 *r _A4 = 2*8=16. The same calculation method can obtain the user A's event pair <E ₂ , E ₈ >, the product of the E ₂ and E ₈ interest scores is r _A2 * r _A8 = 4 * 0 = 0; also get user A's In the event pair <E ₁ , E ₄ >, the product of the interest scores of E ₁ and E ₄ is r _A1 *r _A4 =3*8=24.

According to the above method, the interest scores of user B's event pairs <E ₅ , E ₄ >, event pairs <E ₂ , E ₈ >, and event pairs <E ₁ , E ₄ >, E ₅ and E _{4 can be obtained} . The product of the product is 8, the product of the interest scores of E ₂ and E ₈ is 8, and the product of the interest scores of E ₁ and E ₄ is 20.

According to the above method, the interest scores of the user C events <E ₅ , E ₄ >, the event pairs <E ₂ , E ₈ >, and the event pairs <E ₁ , E ₄ >, E ₄ and E _{5 can be obtained} . The product of the product is 12, the product of the interest scores of E ₂ and E ₈ is 8, and the product of the interest scores of E ₁ and E ₄ is 8.

Wherein, in the event pair <E ₂ , E ₈ > of the user A, the number of executions of the purchase event of the webpage item y is 0, that is, the interaction event has not occurred, and the interest score of the purchase event is 0, so the product is also 0. , the data can be eliminated during calculation, saving data and calculation.

The sum of the products of the same event pair of the plurality of users and the interest score of the target interaction event is calculated.

Calculates the sum of the squares of the interest scores of the interaction events in the same event pair for multiple users.

Calculates the sum of the squares of the interest scores of the target interaction events in the same event pair for multiple users.

Calculate the correlation between target interaction events and interaction events in the same event pair of multiple users.

Events can be obtained <E _5, E _4> interactivity events in _{E. 5} and target affinity interaction events E ₄ is approximately 0.88, in the event of <E _2, E _8> interactivity events certain interactivity events of E ₂ and E ₈ the correlation is about 0.58, events <E _1, E _4> in the interactive event E ₁ and E target interaction event correlation ₄ is about 0.86.

In actual application, there is a situation in which the user only has one interaction event. In this case, when an interaction event generated by the user is captured, the correlation between the interaction event and the multiple target interaction events is searched, and the correlation degree is extracted. A webpage item corresponding to the target interaction event within the first preset range is used as the target webpage item.

For example, when the user D performs the favorite event E ₅ for the webpage item y, it finds that there is a correlation S ₅₄ between E ₄ and E ₅ , and S ₅₄ =0.88, which belongs to the preset relevance of S _ij ∈[0.8,1]. Within the scope, the E ₄ corresponding web page item x is extracted and recommended to the current user.

When user E performs the favorite event E ₂ for the webpage item x, it finds that there is a correlation S ₂₈ between E ₈ and E ₂ , but S ₂₈ =0.58, which does not belong to the correlation range of the preset S _ij ∈[0.8,1]. The web page item y corresponding to E ₈ may not be recommended to the current user.

In another case, when a plurality of interaction events occur in the current user, the correlation between the target interaction events and the multiple interaction events may be searched for each target interaction event; and the target interaction events are respectively associated with the respective interaction events. Correlation degree, a recommendation degree that identifies a correlation between the target interaction event and the plurality of interaction events is calculated; and the webpage item corresponding to the target interaction event in the second preset range is extracted as the target webpage item.

For example, a user D performed for the web project x 1 time browsing event E _1, on page item y was once viewing the event E _5; find E ₁ and E ₄ correlation of S _14, and S ₁₄ = 0.86 When it is found that there is a correlation degree S ₅₄ between E ₅ and E ₄ , and S ₅₄ =0.88, the recommendation degree of the web page item x (the webpage item corresponding to the interaction event E ₄ ) is recommended to the user D as 1*0.86+1*0.88= 1.74, which falls within the recommended range of the preset R _uj ∈[0.8,+∞], so the webpage item x is recommended as the target webpage item to the current user.

In addition to the manner of setting the recommendation threshold, the recommendation may also be performed according to the order of recommendation. For example, when there are 10 target interaction events that can be recommended to the user D, the ranking may be ranked according to the recommendation degree from high to low. The webpage items corresponding to the three target interaction events are recommended to the user D as the target webpage item.

It should be noted that, for the method embodiments, for the sake of simple description, they are all expressed as a series of action combinations, but those skilled in the art should understand that the embodiments of the present application are not limited by the described action sequence, because According to embodiments of the present application, certain steps may be in other orders or At the same time. In the following, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required in the embodiments of the present application.

Referring to FIG. 3, a structural block diagram of an embodiment of an interactive event-based webpage item recommendation apparatus of the present application is shown, which may specifically include the following modules:

The relevance generation module 301 is configured to generate, according to the interaction information of the interaction events of the plurality of users for the webpage item, the correlation between the interaction event and the target interaction event in the interaction event.

As a preferred example of the embodiment of the present invention, the relevance generation module 401 may include the following sub-modules:

The interest score generation sub-module is configured to generate an interest score of the interaction event according to an event attribute of the interaction event.

As a preferred example of an embodiment of the present invention, the interaction event includes a browsing event for a webpage item, and/or a favorite event, and/or adding a shopping cart event, and/or a purchase event, the interest score generation sub-module Can include the following subunits:

As a preferred example 1 of the embodiment of the present invention, the relevance calculation submodule may include the following subunits:

The event pair component unit is configured to form an event pair of the interaction event and the target interaction event of the same user, and calculate a product of the interaction event value of the event pair and the interest degree score of the target interaction event.

As a preferred example of the embodiment of the present invention, the correlation calculation subunit may be specifically configured to:

The same event pair for multiple users, the sum of the squares of the interest scores of the interaction events in the same event pair, the sum of the squares of the interest scores of the target interaction events, and the interaction events and target interaction events a sum of products of interest scores, and correspondingly forming multi-dimensional vector values corresponding to the interaction events in the event pair and the target interaction events; calculating a cosine value between the multi-dimensional vector values as the correlation.

As a preferred example 2 of the embodiment of the present invention, the correlation calculation submodule may include the following subunits:

The interest score group constitutes a sub-unit, which is used to form an event pair of the same user's interaction event and the target interaction event, and respectively adopts the same event pair of the multiple user pairs and the interest degree score of the target interaction event respectively. The set of interactive event interest scores and the target interaction event interest scores.

The extracting module 302 is configured to extract, according to the interaction event of the current user, the target webpage item corresponding to the at least one target interaction event according to the correlation between the interaction event and the plurality of target interaction events.

As a preferred example 1 of the embodiment of the present invention, when an interaction event occurs in the current user, the extraction module 302 may include the following sub-modules:

The first relevance search sub-module is configured to search for an interaction event generated by the user to find a correlation between the one interaction event and the multiple target interaction events.

As a preferred example 2 of the embodiment of the present invention, when multiple interaction events occur in the current user, the extraction module 302 may include the following sub-modules:

The second relevance search sub-module is configured to search for a correlation between the target interaction event and the multiple interaction events for each target interaction event.

The recommendation degree calculation sub-module is configured to calculate a recommendation degree that identifies a correlation between the target interaction event and the plurality of interaction events according to the relevance of the target interaction event to each interaction event.

The recommendation module 303 is configured to recommend the target webpage item to the current user.

The web page item can include a transaction object, and/or video, and/or audio, and/or an electronic book.

The device of the present application combines a webpage item with an interaction behavior for a webpage item to form an interaction event, and performs correlation calculation based on the interaction event, thereby avoiding the problem that the effective webpage item recommendation cannot be performed because the user does not provide the evaluation score; The interaction event for the webpage project is used as the basis of the correlation calculation, and the webpage item that meets the user's interest can be objectively and accurately recommended to the user; further, only the correlation degree of the target interaction event and the interaction event is calculated in the correlation calculation, The amount of data in the correlation calculation is reduced, and the computing resources and storage resources of the server are saved.

For the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the description of the method embodiment.

The various embodiments in the present specification are described in a progressive manner, and each embodiment focuses on differences from other embodiments, and the same similar parts between the various embodiments can be referred to each other.

Those skilled in the art will appreciate that embodiments of the embodiments of the present application can be provided as a method, apparatus, or computer program product. Therefore, the embodiments of the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Moreover, embodiments of the present application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) including computer usable program code.

In a typical configuration, the computer device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory. The memory may include non-persistent memory, random access memory (RAM), and/or non-volatile memory in a computer readable medium, such as read only memory (ROM) or flash memory. Memory is computer readable medium Example. Computer readable media includes both permanent and non-persistent, removable and non-removable media. Information storage can be implemented by any method or technology. The information can be computer readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read only memory. (ROM), electrically erasable programmable read only memory (EEPROM), flash memory or other memory technology, compact disk read only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, Magnetic tape cartridges, magnetic tape storage or other magnetic storage devices or any other non-transportable media can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include non-persistent computer readable media, such as modulated data signals and carrier waves.

Embodiments of the present application are described with reference to flowcharts and/or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the present application. It will be understood that each flow and/or block of the flowchart illustrations and/or FIG. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processor or other programmable data processing terminal device to produce a machine such that instructions are executed by a processor of a computer or other programmable data processing terminal device Means are provided for implementing the functions specified in one or more of the flow or in one or more blocks of the flow chart.

The computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture comprising the instruction device. The instruction device implements the functions specified in one or more blocks of the flowchart or in a flow or block of the flowchart.

These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device such that a series of operational steps are performed on the computer or other programmable terminal device to produce computer-implemented processing, such that the computer or other programmable terminal device The instructions executed above provide steps for implementing the functions specified in one or more blocks of the flowchart or in a block or blocks of the flowchart.

While a preferred embodiment of the embodiments of the present application has been described, those skilled in the art can make further changes and modifications to the embodiments once they are aware of the basic inventive concept. Therefore, the appended claims are intended to be interpreted as including all the modifications and the modifications

Finally, it should also be noted that in this context, relational terms such as first and second are used merely to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply these entities. There is any such actual relationship or order between operations. Furthermore, the terms "comprises" or "comprising" or "comprising" or any other variations are intended to encompass a non-exclusive inclusion, such that a process, method, article, or terminal device that includes a plurality of elements includes not only those elements but also Other elements that are included, or include elements inherent to such a process, method, article, or terminal device. An element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or terminal device that comprises the element, without further limitation.

The above is an interactive event-based webpage project recommendation method and a cross-event-based webpage project recommendation device provided by the present application, and a specific example is applied to illustrate the principle and implementation manner of the present application. The description of the above embodiments is only for helping to understand the method of the present application and its core ideas; at the same time, for those of ordinary skill in the art, depending on the idea of the present application, there will be changes in specific implementations and applications. In summary, the contents of this specification are not to be construed as limiting the application.

Claims

A method for recommending webpage items based on interactive events, characterized in that it comprises:

Generating a correlation between the interaction event and a target interaction event in the interaction event according to interaction information of interaction events of the plurality of users for the webpage item;

Extracting, according to the correlation event of the interaction event and the plurality of target interaction events, the target webpage item corresponding to the at least one target interaction event;

The target webpage item is recommended to the current user.
The method according to claim 1, wherein the step of generating the correlation between the interaction event and the target interaction event in the interaction event according to the interaction information of the interaction events of the plurality of users for the webpage item comprises: :

Generating an interest score of the interaction event according to an event attribute of the interaction event;

Calculating a correlation between the interaction event and the target interaction event according to the interest score of the interaction event.
The method of claim 2, wherein the interactivity event comprises a browsing event for a webpage item, and/or a favorite event, and/or adding a shopping cart event, and/or a purchase event, the interactivity event The step of generating an interest score of the interactive event by the event attribute includes:

Taking the product of the browsing event, the favorite event, the added shopping cart event, the number of interaction events N in the purchase event, and the preset parameter as the interest degree score of the interaction event is M.
The method according to claim 2, wherein the step of calculating the correlation between the interaction event and the target interaction event according to the interest score of the interaction event comprises:

The interaction event of the same user and the target interaction event are combined into an event pair, and the product of the interaction event value of the event pair and the interest degree score of the target interaction event is calculated;

Calculating the interest score of the interaction event in the same event pair of the plurality of users, the interest score of the target interaction event, and the product of the interest event score of the interaction event and the target interaction event, The degree to which an interaction event interacts with a target interaction event.
The method according to claim 4, wherein the interest score according to the interaction event in the same event pair of the plurality of users, the interest score of the target interaction event, and The product of the interaction event and the interest score of the target interaction event, and the correlation between the interaction event and the target interaction event in the event pair is calculated as follows:

The same event pair for multiple users, the sum of the squares of the interest scores of the interaction events in the same event pair, the sum of the squares of the interest scores of the target interaction events, and the interaction events and target interaction events a sum of products of interest scores, and correspondingly forming multi-dimensional vector values corresponding to the interaction events and the target interaction events in the event pair;

A cosine value between the multidimensional vector values is calculated and used as the correlation.
The method according to claim 2, wherein the step of calculating the correlation between the interaction event and the target interaction event according to the interest score of the interaction event comprises:

The interaction event of the same user and the target interaction event are combined into an event pair, and the interest scores of the interaction event and the target interaction event of the same event pair of multiple users are respectively composed of the interaction event interest degree score set and the target interaction event. Interest score set;

And calculating, as the correlation, a JCd coefficient of the set of interactivity event interest scores and the target interaction event interest score set.
The method according to claim 1, wherein when the current user has an interaction event, the interaction event for the current user is extracted according to the correlation between the interaction event and the plurality of target interaction events. The steps of a target webpage project corresponding to a target interaction event include:

When capturing an interaction event generated by the user, searching for a correlation between the one interaction event and multiple target interaction events;

The webpage item corresponding to the target interaction event in the first preset range is extracted as the target webpage item.
The method according to claim 1, wherein when the current user has a plurality of interaction events, the interaction event for the current user is extracted according to the correlation between the interaction event and the plurality of target interaction events. The steps of the target webpage item corresponding to the at least one target interaction event include:

Searching for the relevance of the target interaction event to the plurality of interaction events for each target interaction event;

Calculating a recommendation degree that identifies a correlation between the target interaction event and the plurality of interaction events according to the relevance of the target interaction event to each interaction event;

The webpage item corresponding to the target interaction event in the second preset range is extracted as the target webpage item.
The method of claim 1 wherein the web page item comprises a transaction object, and/or a video, and/or audio, and/or an electronic book.
A webpage item recommendation device based on an interaction event, comprising:

a correlation generation module, configured to generate, according to interaction information of interaction events of the plurality of users for the webpage item, a correlation degree between the interaction event and the target interaction event in the interaction event;

An extraction module, configured to extract, according to the interaction event of the current user, a target webpage item corresponding to the at least one target interaction event according to the correlation between the interaction event and the plurality of target interaction events;

a recommendation module for recommending the target webpage item to the current user.
The apparatus according to claim 10, wherein the correlation generation module comprises:

a score of interest generation sub-module, configured to generate an interest score of the interaction event according to an event attribute of the interaction event;

The correlation calculation sub-module is configured to calculate a correlation between the interaction event and the target interaction event according to the interest degree score of the interaction event.
The device of claim 11, wherein the interaction event comprises a browsing event for a webpage item, and/or a favorite event, and/or adding a shopping cart event, and/or a purchase event, the interest score The value generation submodule includes:

a score of interest M to generate a subunit for using the product of the browsing event, the favorite event, the added shopping cart event, the number of interaction events N in the purchase event, and a preset parameter as The interest score of the interactive event is M.
The apparatus according to claim 11, wherein the correlation calculation sub-module comprises:

The event pair constitutes a subunit for organizing the interaction event of the same user with the target interaction event. a pair and calculate a product of the interest scores of the interaction events in the event pair and the target interaction events;

a correlation calculation sub-unit for multiplying the interest score of the interaction event in the same event pair of the plurality of users, the interest score of the target interaction event, and the interest score of the interaction event and the target interaction event And calculating a correlation between the interaction event in the event pair and the target interaction event.
The apparatus according to claim 13, wherein the correlation calculation subunit is specifically configured to:

The same event pair for multiple users, the sum of the squares of the interest scores of the interaction events in the same event pair, the sum of the squares of the interest scores of the target interaction events, and the interaction events and target interaction events a sum of products of interest scores, and correspondingly forming multi-dimensional vector values corresponding to the interaction events and the target interaction events in the event pair;

A cosine value between the multidimensional vector values is calculated and used as the correlation.
The apparatus according to claim 11, wherein the correlation calculation sub-module comprises:

The interest score group constitutes a sub-unit, which is used to form an event pair of the same user's interaction event and the target interaction event, and respectively adopts the same event pair of the multiple user pairs and the interest degree score of the target interaction event respectively. An interactive event interest score set and a target interaction event interest score set;

a Jaccard coefficient calculation subunit, configured to calculate and select the Jaccard coefficient of the interactivity event interest score set and the target interactivity interest score set.
The apparatus according to claim 10, wherein when the current user has an interaction event, the extraction module comprises:

a first correlation finding submodule, configured to: when capturing an interaction event generated by the user, searching for a correlation between the one interaction event and multiple target interaction events;

The first extraction sub-module is configured to extract a webpage item corresponding to the target interaction event in the first preset range as the target webpage item.
The device according to claim 10, wherein when said current user occurs When a plurality of interaction events are performed, the extraction module includes:

a second relevance search sub-module, configured to search for a correlation between the target interaction event and the multiple interaction events for each target interaction event;

a recommendation calculation sub-module, configured to calculate a recommendation degree that identifies a correlation between the target interaction event and the plurality of interaction events according to the correlation between the target interaction event and each interaction event;

The second extraction sub-module is configured to extract, as the target webpage item, a webpage item corresponding to the target interaction event in the second preset range.
The device of claim 10, wherein the web page item comprises a transaction object, and/or a video, and/or audio, and/or an electronic book.