Product recommendation method, electronic device and storage medium

Abstract

The present application discloses a product recommendation method, electronic device and storage medium, including: if a tracking trigger event of a product page is monitored, collecting tracking data of at least one tracking point of product n corresponding to the tracking trigger event in the product page; obtaining a target score of the user's browsing behavior based on the trigger duration corresponding to at least one tracking point; performing similarity matching between the target score and the historical scores corresponding to at least two historical users, and obtaining a similarity matching result of the browsing behavior between the user and each historical user; selecting at least one product according to the similarity matching result; in this way, the present application performs tracking setting with the product as the granularity, collects the user's trigger duration for different products in different modules of the page, and matches the target score used to characterize the browsing behavior characteristics of the current user with the historical scores of historical users who have generated valid orders in the historical user rating library one by one to perform product recommendation, thereby reducing the amount of data and complexity.

Description

Technical Field

[0001] The present application relates to the field of recommendation technology, and in particular to a product recommendation method, electronic device, and storage medium. Background Art

[0002] With the rapid development of the Internet, recommendation systems have become increasingly important to IoT platforms and e-commerce platforms in recent years. Recommendation systems can be divided into two categories: rating prediction and personalized recommendation. The former is to estimate the rating of unrated items, and the latter is to provide users with a personalized recommendation list. Recommendation algorithms in related technologies can be divided into content-based recommendation algorithms and collaborative filtering-based recommendation algorithms. Content-based recommendation algorithms are to establish feature descriptions by analyzing the user's content attributes (such as age, gender, preferences, region, income, etc.), the content attributes of items, and the user's operation behavior on items (such as comments, collections, likes, views, browsing, clicks, shopping carts, purchases, etc.), and recommend items of potential interest to users. Collaborative filtering-based recommendation algorithms use user historical behavior data to discover user preferences and predict products that users may like for recommendation.

[0003] However, the recommendation algorithms in related technologies collect various basic information of users, various types of operation behaviors and other data to build models. In the process of recommending products to users, there are at least problems of large data volume and high algorithm complexity. Summary of the Invention

[0004] The present application provides a product recommendation method, electronic device and storage medium to solve at least the problems of large data volume and high algorithm complexity in related technologies.

[0005] The technical solution of this application is achieved as follows:

[0006] The present application provides a product recommendation method, the method comprising:

[0007] If a tracking trigger event is detected on a product page, tracking data for at least one tracking point of product n corresponding to the tracking trigger event on the product page is collected. The tracking data includes product type, tracking location, and trigger duration of the tracking location.

[0008] Obtaining a target score for the user's browsing behavior based on the trigger duration corresponding to the at least one embedding point;

[0009] Perform similarity matching between the target score and the historical scores corresponding to at least two historical users, and obtain a similarity matching result of the browsing behavior between the user and each historical user; the historical score is the score of the browsing behavior of the historical user corresponding to the valid order product;

[0010] At least one product is selected from all the products according to the similarity matching result, and the at least one product is output.

[0011] The present application provides a product recommendation device, comprising:

[0012] A collection module is configured to collect, upon detecting a triggering event on a product page, tracking data for at least one tracking point of a product n corresponding to the triggering event on the product page, the tracking data including the product type, tracking location, and triggering duration of the tracking location;

[0013] a processing module, configured to obtain a target score for the user's browsing behavior based on a trigger duration corresponding to the at least one tracking point;

[0014] The processing module is further configured to perform similarity matching between the target score and the obtained historical scores corresponding to at least two historical users, thereby obtaining a similarity matching result of the browsing behavior between the user and each historical user; the historical score is a score of the browsing behavior of the historical user corresponding to the product that generated the valid order; and select at least one product from all the products based on the similarity matching result.

[0015] An output module is used to output the at least one product.

[0016] The present application provides an electronic device, comprising:

[0017] a memory for storing executable instructions;

[0018] The processor is used to execute the executable instructions stored in the memory to implement the above-mentioned product recommendation method.

[0019] The present application provides a storage medium storing one or more programs, which can be executed by one or more processors to implement the above-mentioned product recommendation method.

[0020] The present application provides a product recommendation method, electronic device and storage medium. If a buried point trigger event on a product page is monitored, buried point data of at least one buried point of product n corresponding to the buried point trigger event on the product page is collected, and the buried point data includes product type, buried point location and trigger duration of the buried point location; based on the trigger duration corresponding to at least one buried point, a target score of the user's browsing behavior is obtained; the target score is matched with the historical scores corresponding to at least two historical users to obtain a similarity matching result of the browsing behavior between the user and each historical user; the historical score is the score of the browsing behavior of the historical user corresponding to the product that generated the valid order; at least one product is selected from all products according to the similarity matching result, And output at least one product; in this way, the present application sets the tracking point with the product as the granularity, and there is no need to collect a large amount of user behavior information in multiple dimensions. It only collects the trigger duration of the user's tracking point for different products in different modules on different pages, that is, only collects time dimension data; further, the electronic device obtains a score for characterizing the user's browsing behavior characteristics based on the trigger duration corresponding to the user's browsing of different products, and matches the target score for characterizing the current user's browsing behavior characteristics with the historical score of the historical user who has generated valid order products, and selects the product for output based on the matching result; in this way, qualitative results are obtained by matching the scores, so as to recommend relevant product content, thereby reducing the amount of data and reducing the complexity of the algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A flowchart of an optional product recommendation method provided in an embodiment of the present application;

[0022] Figure 2 A schematic diagram of a method for setting product tracking points provided in an embodiment of the present application;

[0023] Figure 3 A structural block diagram of a product recommendation method provided in an embodiment of the present application;

[0024] Figure 4 A flowchart of an optional product recommendation method provided in an embodiment of the present application;

[0025] Figure 5 A flowchart of an optional product recommendation method provided in an embodiment of the present application;

[0026] Figure 6 A flowchart of an optional product recommendation method provided in an embodiment of the present application;

[0027] Figure 7 A flowchart of an optional product recommendation method provided in an embodiment of the present application;

[0028] Figure 8A flowchart of an optional product recommendation method provided in an embodiment of the present application;

[0029] Figure 9 A schematic diagram of the structure of a product recommendation device provided in an embodiment of the present application;

[0030] Figure 10 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0031] The following will clearly and completely describe the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. Here, the "another" or "yet another" mentioned in the description of the drawings does not refer to a specific embodiment. The various embodiments of the present application can be combined with each other without conflict.

[0032] It should be understood that the “embodiments of the present application” or “the aforementioned embodiments” mentioned throughout the specification mean that the specific features, structures or characteristics related to the embodiments are included in at least one embodiment of the present application. Therefore, “in the embodiments of the present application” or “in the aforementioned embodiments” appearing throughout the specification do not necessarily refer to the same embodiment. In addition, these specific features, structures or characteristics can be combined in one or more embodiments in any suitable manner. In the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application. The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages and disadvantages of the embodiments.

[0033] See also Figure 1 , Figure 1 : This is a flow chart of an optional product recommendation method provided in an embodiment of the present application. The product recommendation method is applied to electronic devices and includes the following steps:

[0034] Step 101: If a tracking trigger event is detected on a product page, collect tracking data of at least one tracking point of product n corresponding to the tracking trigger event on the product page. The tracking data includes product type, tracking location, and trigger duration of the tracking location.

[0035] Where n is an integer greater than or equal to 1 and less than or equal to N, and N is the total number of products viewed by users in all product pages. Here, all product pages include at least one product page, each product page contains at least one product, and the number of products contained in different product pages can be the same or different.

[0036] In the embodiment of the present application, the product page can be understood as a page that the user browses through an application installed in the electronic device. The product page can also be a page that the user browses the product through the browser that comes with the electronic device. Here, the product page includes but is not limited to the initial page, solution page, product introduction page, ordering page and help page. This application does not impose specific restrictions on this.

[0037] In an embodiment of the present application, trigger events include mouse sliding events, click events, etc. monitored when a user browses a product page.

[0038] In the embodiment of the present application, the buried point can be understood as adding a global monitoring point to the product recommendation system, that is, for any page browsed by the user, the product recommendation system can capture what operations and behaviors the user has performed at the monitoring point, and record the user's operation behavior (i.e., buried point data) at the monitoring point (i.e., buried point). Here, the buried point is also called an anchor point. The buried point data of the buried point include the product type corresponding to the buried point, the buried point location, and the triggering duration of the buried point location, such as product types including but not limited to cloud hosts, cloud hard disks, public Internet Protocol (Internet Protocol, IP), elastic load balancing, and databases; the triggering duration of the buried point location can be understood as when the triggering event of the buried point is detected, the timer is enabled to count the length of time the pointer corresponding to the mouse stays (or the browsing duration of the user); here, the product types of the buried points corresponding to different products on the same product page are different; on the same product page, due to the presence of different marketing modules such as hot selling modules, promotion modules, and product modules for displaying all products, the product types of the buried points corresponding to the buried points of the same product in different modules, i.e., different locations, are the same.

[0039] It should be noted that when a user browses a product page, for a product that the user is interested in, the user moves the pointer of the product page and keeps the pointer on the introduction module of the product for a while, but does not necessarily click in. For a product that the user is not interested in, the user will quickly swipe past the introduction module corresponding to the product.

[0040] In a feasible scenario, tracking data can be expressed as T(x, y, t), where x is the product type, y is the tracking location, and t represents the trigger duration of the tracking obtained by activating a timer when a tracking trigger event is detected when a user browses a product page. Here, the product type set composed of different product types (type) can be expressed as X = {x1, x2, x3, ..., x k}, the product tracking point set consisting of tracking points at different locations corresponding to products of different product types can be expressed as Y = {y1, y2, y3, ..., y j Here, when user m browses a product page, the tracking data of tracking point j corresponding to any product n can be expressed as T j(x,y j ,t).

[0041] In a feasible application scenario, such as Figure 2 As shown in the figure, an application starts with the root node and registers a global public component "DomRentation" with a mouse monitoring function. It defines type as the type of the public component and can pass in values ​​corresponding to different product types, such as cloud host x1, cloud disk x2, public network IP x3, elastic load balancing x4, and database x5. By reusing public components of different types in the application, a single product page containing only one product or a multi-product page M containing multiple products on a page is formed. Here, Figure 2 A single product page includes page A showing a single product, Cloud Host x1, or page B showing a single product, Cloud Disk x2, or page C showing a single product, Public IP x3. A multi-product page M includes pages showing multiple products, such as Cloud Host x1, Cloud Disk x2, Public IP x3, and Elastic Load Balancer x4.

[0042] In other embodiments of the present application, the tracking point includes a node on the document object model DOM tree corresponding to the product page.

[0043] In the embodiment of the present application, the Document Object Model (DOM) is a standard programming interface for processing Extensible Markup Language (XML) recommended by the World Wide Web Consortium. On a web page, objects of a page (or document) are organized in a tree structure, which is a standard model for representing objects in a document.

[0044] Here, in the global public component "DomRentation", the "document.addEventListener" event of the DOM element is used to monitor the user's mouse "mousemove" behavior, that is, the buried point trigger event of the product page is monitored, and the monitoring method "domReader" is bound. In the "domReader" method, the "setTimeInterVal" method is used to set the timer. When the electronic device monitors the buried point trigger event of the product page, the timer is enabled to obtain the trigger duration of the mouse pointer staying at the buried point corresponding to the product, and the trigger duration is stored. At the end of the page life cycle, the timing of all products in the product page is stopped. In this way, in the embodiment of the present application, there is no need to analyze the user portrait and create labels through active user scoring. Instead, the trigger duration of at least one buried point corresponding to the product is directly obtained by using the public component to bury the DOM module without the user's perception, and the browsing content and products that the user is interested in are determined, thereby preparing for data processing and marketing.

[0045] In other embodiments of the present application, if a tracking trigger event is detected on a product page, the product type and tracking location corresponding to the tracking of product n on the product page are collected. If the next tracking trigger event for the product page is not detected within the preset time, the preset time is determined to be the trigger time of the tracking location corresponding to the tracking of product n, and the trigger time, tracking location and corresponding product type are stored.

[0046] Here, the preset duration can be understood as the effective duration when a user browses any product in the product page. Exemplarily, the preset duration can be 30 seconds (second, s).

[0047] It should be noted that after monitoring the trigger event of the embedding point for the product page, the product type and embedding point location corresponding to the embedding point of product n in the product page are collected, and according to user habits and single-screen browsing speed, if the user monitors the next embedding point trigger event for the product page within a preset time, such as 30 seconds (the embedding point trigger event generated when the user operates the mouse or keyboard), the time between the monitoring of the embedding point trigger event for the product page and the next embedding point trigger event is determined to be the trigger time of the embedding point location corresponding to the embedding point of product n, which is the effective reading time of the DOM. If the user does not monitor the next embedding point trigger event for the product page within a preset time, such as 30 seconds, that is, the user does not operate the mouse or keyboard for more than the preset time, the preset time is determined to be the trigger time of the embedding point location corresponding to the embedding point of product n, and the currently obtained trigger time is stored in an array, such as the timeList array. At the end of the page life cycle, the timing of all product modules in the page is stopped, and the trigger time, embedding point location and corresponding product type are stored in the database.

[0048] In practical applications, electronic devices may be mobile terminal devices such as mobile phones, tablet computers, laptop computers, personal digital assistants (PDAs), cameras, wearable devices, smart bracelets, smart watches, vehicle-mounted devices, e-book readers, and electronic game consoles; electronic devices may also be fixed terminal devices such as desktop computers.

[0049] Step 102: Based on the trigger duration corresponding to at least one embedding point, obtain a target score for the user's browsing behavior.

[0050] In an embodiment of the present application, the electronic device can call the cloud forward generator in the cloud model based on the trigger duration corresponding to at least one buried point to obtain a target score for the user's browsing behavior. Here, the cloud model can be understood as an uncertainty conversion model between a qualitative concept represented by a language value and its quantitative representation. The cloud model can be implemented by a cloud generator, wherein the cloud generator includes a forward cloud generator and a reverse cloud generator. The forward cloud generator is a mapping from qualitative to quantitative, and evaluates the user's browsing behavior based on digital feature values; in general, the cloud model can handle the randomness and ambiguity of user ratings, and use the cloud generator to realize the conversion of quantitative and precise numerical values ​​to qualitative concepts.

[0051] In the embodiment of the present application, the target score can be understood as the quantified data of the buried points corresponding to product n when the user browses the product page to form the score information of the user's browsing behavior. The score is also called a cloud droplet. The cloud droplet is the basic unit of the cloud model and is represented by C(r, μ). Here, r represents the random number generated when the digital eigenvalue of the normal cloud satisfies the normal cloud distribution law, and u represents the quantitative certainty of the random number.

[0052] In an embodiment of the present application, when user u browses a product page through an electronic device, user u controls the mouse to move the pointer on the product page to product n or clicks on product n. The electronic device detects a movement trigger event or a click trigger event of a buried point on the product page. At this time, the electronic device obtains the product type, buried point location, and trigger duration of the pointer at the buried point location of at least one buried point corresponding to product n on the product page; further, the electronic device obtains a target score c for the browsing behavior of user u browsing the product page based on the trigger duration of at least one buried point corresponding to product n. u .

[0053] Step 103: Perform similarity matching between the target score and the historical scores corresponding to at least two historical users, and obtain similarity matching results of the browsing behaviors between the user and each historical user.

[0054] The historical rating is the rating of historical users' browsing behaviors that generated valid order products.

[0055] In an embodiment of the present application, the historical score is used to represent the potential relationship between historical users and different products. The historical score is based on the operation information between the historical trigger duration record of the historical user and the product that generates the valid order, and the historical score information is formed by quantifying this information.

[0056] In the embodiment of the present application, if there is a historical score of browsing behavior of a historical user corresponding to at least one valid order product generated within the target period, the historical score of browsing behavior of at least one historical user constitutes a historical score library. For example, the historical score library can be represented as C M ={c1,c2,…,c m ,…,c M} T Where M represents the number of historical users who have generated at least one valid order product within the target period, c m represents the historical score corresponding to the browsing behavior of historical user m, where m is an integer greater than or equal to 1 and less than or equal to M, and T represents transposition.

[0057] In the embodiment of the present application, the similarity matching result can be calculated based on the user similarity calculation algorithm to calculate the similarity of the browsing behavior between the current user and each historical user.

[0058] In the embodiment of the present application, the valid order product can be understood as the product that the user clicks to purchase, that is, the ordered product.

[0059] In the present application, refer to Figure 3 As shown, the electronic device obtains the target score c of the browsing behavior of user u based on the triggering duration corresponding to at least one buried point. u Afterwards, obtain the historical rating database C consisting of the historical ratings of the browsing behaviors of historical users corresponding to the valid order products M ={c1,c2,…,c m ,…,c M} T , perform similarity matching on the target score of the current user and the historical score corresponding to each historical user in the historical score library, and obtain the similarity matching result of the browsing behavior between the user and each historical user.

[0060] Step 104: Select at least one product from all products based on the similarity matching result, and output the at least one product.

[0061] In an embodiment of the present application, the electronic device performs similarity matching between the target score and each historical score in the obtained historical score library. After obtaining the similarity matching results of the browsing behaviors between the user and each historical user, the electronic device determines, based on the size of the similarity matching results, that the browsing behavior of the historical user corresponding to the maximum similarity is closest to the browsing behavior of the current user. Furthermore, from the browsing time of all products browsed by historical users, all products are sorted according to the browsing time. Based on the sorting results, at least one product is selected from all products and outputted, thereby promoting user consumption.

[0062] The present application provides a product recommendation method, which includes the following steps: if a buried point triggering event on a product page is detected, the buried point data of at least one buried point of product n corresponding to the buried point triggering event on the product page is collected, and the buried point data includes the product type, buried point location and triggering duration of the buried point location; based on the triggering duration corresponding to at least one buried point, a target score of the user's browsing behavior is obtained; the target score is matched with the historical scores corresponding to at least two historical users to obtain a similarity matching result of the browsing behavior between the user and each historical user; the historical score is the score of the browsing behavior of the historical user corresponding to the valid order product; at least one product is selected from all products according to the similarity matching result, and at least one product is output. a product; in this way, the present application sets tracking points at the product granularity, and there is no need to collect a large amount of user behavior information in multiple dimensions. It only collects the trigger duration of the user's tracking points for different products in different modules on different pages, that is, it only collects time dimension data; further, the electronic device obtains a score for characterizing the user's browsing behavior characteristics based on the trigger duration corresponding to the user's browsing of different products, and matches the target score for characterizing the current user's browsing behavior characteristics with the historical scores of historical users who have generated valid order products, and selects products for output based on the matching results; in this way, qualitative results are obtained through matching scores, so as to recommend relevant product content, thereby reducing the amount of data and reducing the complexity of the algorithm.

[0063] See also Figure 4 , Figure 4 : This is a flow chart of an optional product recommendation method provided in an embodiment of the present application. The product recommendation method is applied to electronic devices and includes the following steps:

[0064] Step 301: If a buried point triggering event is detected on a product page, the buried point data of at least one buried point of product n corresponding to the buried point triggering event on the product page is collected. The buried point data includes the product type, buried point location, and the triggering duration of the buried point location.

[0065] Step 302: Filter out the trigger duration of the buried point that is greater than or equal to the duration threshold from at least one trigger duration corresponding to product n.

[0066] Here, the duration threshold is used to filter the trigger duration. Exemplarily, the duration threshold may be 1 second.

[0067] In an embodiment of the present application, the electronic device filters out the trigger duration of the buried points that is greater than or equal to the duration threshold from at least one trigger duration corresponding to product n for all products, and then filters the trigger duration corresponding to all products. In this way, by filtering the trigger duration, that is, removing the user's shorter browsing time, the trigger duration caused by the user's accidental click or erroneous operation is avoided, thereby achieving data cleaning.

[0068] In the embodiment of the present application, the trigger duration of the tracking point that is greater than or equal to the duration threshold is screened out from at least one trigger duration corresponding to product n. This can be obtained by the following (Formula 1) and (Formula 2):

[0069] ΩT=T j (x,y j , t)×δ(t) (Formula 1)

[0070] ΩT represents the trigger duration of the selected tracking points that are greater than or equal to the duration threshold, T j (x,y j , t) represents the embedding data of embedding point j corresponding to any product n when the user browses the product page, x represents the product type corresponding to product n, y j It represents the location of the tracking point corresponding to the product, t represents the triggering duration of the tracking point location; δ(t) represents the function corresponding to the duration threshold.

[0071] Here, the function corresponding to the duration threshold can be expressed as follows (Formula 2):

[0072]

[0073] Step 303: Based on the trigger durations of the filtered embedding points, the total trigger durations of all embedding points corresponding to product n are obtained.

[0074] In the embodiment of the present application, the electronic device determines the trigger duration of all the filtered buried points corresponding to product n based on the trigger duration of the filtered buried points, thereby obtaining the total trigger duration of all the filtered buried points corresponding to product n; further, for the N products browsed by the user, the N total trigger durations corresponding to the N products browsed by the user are obtained. Here, the N total trigger durations corresponding to the N products browsed by user u can be represented by a one-dimensional matrix T u =[T u1 T u2 ... T uN ].

[0075] In the embodiment of the present application, step 303 can be combined with the process of obtaining the total triggering duration of all the tracking points corresponding to product n based on the triggering duration of the filtered tracking points. Figure 5 For further explanation,

[0076] Step 3031: Obtain the weight corresponding to the burial point location.

[0077] In the embodiments of the present application, different embedding locations have different corresponding weights. For example, on the same product page, due to the presence of different marketing modules such as a hot selling module, a promotional module, and a product module for displaying all products, embedding points can be located in the hot selling module, the promotional module, and the product module. The embedding points in different modules (i.e., different embedding locations) have different corresponding weights, but the sum of the weights of the embedding points in all embedding locations is equal to 1.

[0078] Here, the sum of the weights of all buried points is equal to 1, which can be expressed as follows (Formula 3):

[0079]

[0080] Among them, w nj It represents the weight of the buried point j corresponding to product n at different buried point locations, and p represents the number of buried points j corresponding to product n.

[0081] Step 3032: Based on the weight corresponding to the embedding location, the trigger duration corresponding to the embedding location, and the memory forgetting function corresponding to the embedding location, the total trigger duration corresponding to product n is obtained.

[0082] In an embodiment of the present application, the electronic device obtains, for product n, the weight corresponding to the burying point position of at least one burying point corresponding to product n, the trigger duration corresponding to the burying point position, and the memory forgetting function corresponding to the trigger duration. Based on the weights corresponding to the burying points corresponding to product n at different burying point positions, the trigger durations corresponding to different burying point positions, and the memory forgetting function corresponding to the trigger duration, the total trigger duration corresponding to product n is obtained, and then the N total trigger durations corresponding to the N products browsed by the user are obtained.

[0083] In other embodiments of the present application, the memory forgetting function is a function that determines how the product n corresponding to the buried point location is gradually forgotten based on the time from the time when the buried point trigger event corresponding to the buried point location is monitored to the current time, and a pre-set memory correction factor. Exemplarily, the memory forgetting function f(l) can be expressed as follows (Formula 4):

[0084] f(l)=e -λ1 (Formula 4)

[0085] Where λ is the memory correction factor, and l is the time elapsed between the time the corresponding tracking point trigger event was detected and the current time. It should be noted that the memory forgetting function is the user's memory forgetting curve. The multiplication of the trigger duration of each tracking point by the memory forgetting function represents the weakening effect of the trigger duration of the tracking point location over time. In other words, on a product page, the earlier the product is viewed, the more likely it is to be forgotten by the user. This function is used to correct the data, and the larger λ is, the faster the user forgets the product.

[0086] Here, the total trigger duration corresponding to when the user browses product n can be expressed as follows (Formula 5):

[0087]

[0088] Among them, t n Indicates the total triggering duration corresponding to when the user browses product n, w nj Indicates the weight of the buried point j corresponding to product n at different buried point locations, t j represents the triggering duration of the embedding point j corresponding to the user browsing product n, f(l) represents the memory forgetting function, and p represents the number of embedding points j corresponding to product n.

[0089] In other embodiments of the present application, after obtaining the N total trigger durations corresponding to the N products browsed by the user, the electronic device normalizes the N total trigger durations to obtain the normalized N total trigger durations.

[0090] Here, normalizing the N total trigger durations can be achieved by the following (Formula 6):

[0091]

[0092] Where t′ n is the normalized total trigger duration corresponding to product n, t n The unnormalized total trigger duration corresponding to product n, It is the sum of all total trigger durations corresponding to N products.

[0093] From the above, it can be seen that the embodiment of the present application sets the tracking point based on the product granularity, and there is no need to collect a large amount of user behavior information in multiple dimensions. It only collects the trigger duration of the user's tracking point for different products in different modules on different pages, that is, only collects time dimension data; further, the electronic device processes the trigger duration based on the trigger duration corresponding to the user browsing different products, removes the user's erroneous operation data, and performs memory correction according to the user's memory forgetting function, and normalizes the corrected trigger duration. In this way, the amount of data is reduced and the calculation is convenient and quick.

[0094] Step 304: Obtain a target score for the user's browsing behavior based on the N total trigger durations corresponding to the N products browsed by the user.

[0095] In the embodiment of the present application, step 304 is based on the N total trigger durations corresponding to the N products browsed by the user, and the process of obtaining the target score of the user's browsing behavior is combined with Figure 6 For further explanation,

[0096] Step 3041: Based on the N total trigger durations, obtain the digital characteristic value of the user's browsing behavior.

[0097] Among them, digital eigenvalues ​​include expected value, entropy value and super entropy value.

[0098] In the embodiment of the present application, the digital characteristic values ​​include expected value (Ex), entropy value (En) and hyper entropy value (He).

[0099] In the embodiment of the present application, the expected value reflects the average triggering time for users to browse each product, also known as the browsing time, and is represented by Ex. The entropy value can reflect the degree of uncertainty in the browsing time of users browsing different products, and is represented by En. The super entropy value measures the uncertainty of entropy, and the super entropy value is represented by He, which reflects the degree of discreteness of uncertainty. The uncertainty of the browsing time of users browsing different products becomes increasingly discrete as the super entropy increases, and vice versa, it exhibits a stable characteristic.

[0100] In a feasible application scenario, refer to Figure 3 As shown, the electronic device obtains N total trigger durations corresponding to the user browsing N products, and determines the average of the N total trigger durations as the expected value Ex; based on the expected value Ex and the N trigger durations, the variance S corresponding to the N total trigger durations is calculated; further, based on the N total trigger durations and the expected value Ex, the entropy value En corresponding to the N total trigger durations is calculated as follows (Formula 7); finally, based on the variance S and the entropy value En, the super entropy value En corresponding to the N total trigger durations is calculated as follows (Formula 8). In this way, the digital feature value of the user's total trigger duration for N products is obtained.

[0101]

[0102]

[0103] Where N represents the total number of products, t n Indicates the total trigger duration corresponding to when the user browses product n.

[0104] Step 3042: Based on the expected value, entropy value, and super entropy value, a target score for the user's browsing behavior is obtained.

[0105] In an embodiment of the present application, the electronic device inputs the expected value, entropy value, and super entropy value into the forward cloud generator of the cloud model to obtain a target score of the user's browsing behavior output by the forward cloud generator. Here, the forward cloud generator is a mapping from qualitative to quantitative. Here, the score (or cloud droplet) obtained by mapping the digital feature values, i.e., the expected value, entropy value, and super entropy value, through the forward cloud generator is defined as a predicted score.

[0106] In the embodiment of the present application, step 3042 is based on the expected value, entropy value and super entropy value to obtain the target score of the user's browsing behavior. Figure 7 For further explanation,

[0107] Step S1: Taking the entropy value as the expected value of the first normal distribution function and the super entropy value as the variance of the first normal distribution function, a first random number is obtained through the first normal distribution function.

[0108] Step S2: Taking the expected value as the expected value of the second normal distribution function and the first random number as the variance of the second normal distribution function, obtain a second random number through the second normal distribution function.

[0109] Step S3: Determine the quantitative certainty of the second random number to obtain a target score.

[0110] The target score includes a second random number and a quantitative certainty.

[0111] In an embodiment of the present application, the electronic device uses the entropy value as the expected value of the first normal distribution function and the super entropy value as the variance of the first normal distribution function to obtain a first random number through the first normal distribution function. Using the expected value as the expected value of the second normal distribution function and the first random number as the variance of the second normal distribution function, a second random number is obtained through the second normal distribution function. Based on the second random number, the expected value and the entropy value, the quantitative certainty of the second random number is determined as follows (Formula 9), and a target score containing the quantitative certainty of the second random number and the second random number is obtained.

[0112]

[0113] Wherein, u represents the quantitative certainty of the second random number, r represents the second random number, Ex represents the expected value, and En represents the entropy value.

[0114] Step 305: Perform similarity matching between the target score and the historical scores corresponding to at least two historical users, and obtain a similarity matching result of the browsing behavior between the user and each historical user.

[0115] The historical rating is the rating of historical users' browsing behaviors that generated valid order products.

[0116] In the embodiment of the present application, the process of obtaining the historical scores corresponding to at least two historical users in step 305 is combined with Figure 8 For further explanation,

[0117] Step 3051: Obtain the products that generate valid orders within the target period. When the corresponding historical user browses the historical product page, obtain the historical total trigger duration of all buried points corresponding to each of the at least two historical products browsed, and obtain the historical total trigger duration of at least two historical products browsed by the historical user.

[0118] In the embodiment of the present application, the target period can be understood as a fixed period, which can be one day, one week or one month. This application does not impose any specific restrictions on this.

[0119] In the embodiment of the present application, the total historical trigger duration can be a normalized duration, or a non-normalized duration, and the present application does not impose any specific restrictions on this. It should be noted that before obtaining the total historical trigger duration of each historical user for each product, the electronic device will filter the trigger duration of each product browsed by the historical user obtained through the above (Formula 2) to avoid the trigger duration caused by the user's accidental click or misoperation, thereby achieving data cleaning. The electronic device can also use the weights and memory forgetting functions corresponding to the buried point positions through the above (Formula 3)-(Formula 6) to perform memory correction and normalization on the cleaned trigger duration to achieve the conversion from trigger duration to the generation of valid orders.

[0120] In an embodiment of the present application, the electronic device determines the relationship between the triggering time generated when the historical user who generated the valid order product browsed each product and the final successful order result based on the user data of the valid order product generated after browsing the product recommendation system. When obtaining the browsing history product page of the historical user who generated the valid order product within the target time period, for each of the at least two historical products browsed, the historical total triggering time of all the buried points corresponding to each of the historical products can be expressed as follows: for example, after the historical user 1 who generated the valid order product browsed N products, the historical total triggering time corresponding to the N products browsed by the historical user 1 can be expressed as follows: 11 T 12 … T 1N ]express.

[0121] Step 3052: For at least two valid order products, obtain at least two historical total trigger durations corresponding to at least two products browsed by all historical users corresponding to the at least two valid order products.

[0122] In the embodiment of the present application, after obtaining the historical total trigger duration corresponding to all products browsed by historical users corresponding to a valid order product in step 3051, for at least two valid order products, the at least two historical total trigger durations corresponding to at least two products browsed by all historical users corresponding to the at least two valid order products are obtained. For example, when there are M historical users and N products corresponding to M valid order products, the total trigger duration corresponding to user m browsing product n can be expressed as T mn Indicates that the total historical triggering time corresponding to M historical users browsing N products is T mn Can be used for a two-dimensional matrix T MN express,

[0123]

[0124] Here, T MN It can also be expressed as a historical user's overall rating matrix for all products.

[0125] In this way, the embodiment of the present application sets tracking points based on product granularity, without the need to collect a large amount of user behavior information in multiple dimensions. It only collects the trigger duration of historical users' tracking points in different modules on different pages for different products, that is, only collects time dimension data. Furthermore, the electronic device processes the trigger duration based on the trigger duration corresponding to the historical users' browsing of different products, removes user error operation data, and performs memory correction according to the user's memory forgetting function, and normalizes the corrected trigger duration, thereby obtaining the overall user rating matrix T of historical users for all products. MN , in order to achieve the conversion from triggering time to generating effective consumption orders.

[0126] Step 3053: Obtain historical scores of all historical users based on at least two historical total trigger durations corresponding to at least two products browsed by all historical users.

[0127] In a feasible application scenario, refer to Figure 3 As shown, the electronic device converts at least two historical total trigger durations corresponding to at least two products browsed by all historical users into a two-dimensional matrix T MN , and input the two-dimensional matrix into the forward cloud generator to obtain the digital characteristic value of the browsing behavior of each historical user; and then obtain the digital characteristic value of the browsing behavior of all historical users; then, the electronic device determines the historical quantitative certainty based on the historical user for the rated product through the forward cloud generator based on the digital characteristic value corresponding to each historical user; finally, based on the historical quantitative certainty of each historical user, obtain the historical score of the browsing behavior of each historical user, and then obtain the historical score C corresponding to all historical users. M ={c1,c2,…,cm ,…,c M} T It should be noted that the method of calculating the historical score of each historical user's browsing behavior based on the numerical feature value corresponding to each historical user is the same as the method of obtaining the historical score of the current user based on the numerical feature value corresponding to the current user as described above. This will not be explained in detail here.

[0128] Step 306: Select at least one product from all products based on the similarity matching result, and output the at least one product.

[0129] As can be seen from the above, in the embodiment of the present application, the electronic device obtains the historical scores corresponding to historical users who have generated valid order products based on the triggering time when the user browses the product in the product recommendation system and a large amount of data on the order results, and performs similarity matching with each historical score for the target score of the current user used to characterize the user's browsing behavior characteristics to obtain quantitative products that the user may consume. Afterwards, recommendations can be made based on the quantitative products to promote product consumption.

[0130] It should be noted that, for the description of the same steps and contents in this embodiment as those in other embodiments, reference can be made to the description in other embodiments and will not be repeated here.

[0131] Based on the above embodiments, the present application provides a product recommendation device, which can be used to implement Figure 1 、 Figures 4 to 8 Corresponding to a product recommendation method provided, refer to Figure 9 As shown, the product recommendation device 9 includes:

[0132] The collection module 901 is configured to collect tracking data of at least one tracking point of product n corresponding to the tracking trigger event on the product page upon detecting the tracking trigger event on the product page. The tracking data includes the product type, tracking location, and trigger duration of the tracking location.

[0133] Processing module 902, configured to obtain a target score for the user's browsing behavior based on the trigger duration corresponding to at least one tracking point;

[0134] Processing module 902 is further configured to perform similarity matching between the target score and the obtained historical scores corresponding to at least two historical users, thereby obtaining a similarity matching result of the browsing behavior between the user and each historical user; the historical score is the score of the browsing behavior of the historical user corresponding to the product that generated the valid order; and selecting at least one product from all products based on the similarity matching result.

[0135] The output module 903 is configured to output at least one product.

[0136] In other embodiments of the present application, the tracking point includes a node on the document object model DOM tree corresponding to the product page.

[0137] In other embodiments of the present application, the processing module 902 is also used to filter out the trigger duration of the buried points that is greater than or equal to the duration threshold from at least one trigger duration corresponding to product n; obtain the total trigger duration of all buried points corresponding to product n based on the triggered duration of the filtered buried points; and obtain the target score of the user's browsing behavior based on the N total trigger durations corresponding to the N products browsed by the user.

[0138] In other embodiments of the present application, the acquisition module 901 is also used to obtain the weight corresponding to the burial point location; the processing module 902 is also used to obtain the total trigger duration corresponding to the product n based on the weight corresponding to the burial point location, the trigger duration corresponding to the burial point location, and the memory forgetting function corresponding to the burial point location.

[0139] In other embodiments of the present application, the memory forgetting function is a function that determines that the product n corresponding to the burial point location is gradually forgotten based on the time from the time when the burial point trigger event corresponding to the burial point location is monitored to the current time, and a pre-set memory correction factor.

[0140] In other embodiments of the present application, the processing module 902 is also used to obtain the digital characteristic value of the user's browsing behavior based on N total trigger durations, and the digital characteristic value includes an expected value, an entropy value and a super entropy value; based on the expected value, the entropy value and the super entropy value, a target score of the user's browsing behavior is obtained.

[0141] In other embodiments of the present application, the processing module 902 is also used to obtain a first random number through the first normal distribution function with the entropy value as the expected value of the first normal distribution function and the super entropy value as the variance of the first normal distribution function; obtain a second random number through the second normal distribution function with the expected value as the expected value of the second normal distribution function and the first random number as the variance of the second normal distribution function; determine the quantitative certainty of the second random number to obtain a target score; wherein the target score includes the second random number and the quantitative certainty.

[0142] In other embodiments of the present application, the processing module 902 is also used to obtain the historical total trigger duration of all buried points corresponding to at least two historical products browsed by the corresponding historical user when the historical user browses the historical product page and generates valid order products within the target period, and obtain the at least two historical total trigger durations corresponding to the at least two historical products browsed by the historical user; for at least two valid order products, obtain the at least two historical total trigger durations corresponding to the at least two products browsed by all historical users corresponding to the at least two valid order products; based on the at least two historical total trigger durations corresponding to the at least two products browsed by all historical users, obtain the historical scores of all historical users.

[0143] Based on the above embodiments, the present application provides an electronic device that can be used to implement Figure 1 、 Figures 4 to 8 Corresponding to a product recommendation method provided, refer to Figure 10 As shown, the electronic device 10 ( Figure 10 The electronic device 10 corresponds to Figure 9 The product recommendation device 9) includes: a memory 1001 and a processor 1002, wherein the processor 1002 is used to execute the product recommendation program stored in the memory 1001, and the electronic device 10 implements the following steps through the processor 1002:

[0144] If a tracking trigger event is detected on a product page, tracking data for at least one tracking point of product n corresponding to the tracking trigger event on the product page is collected. The tracking data includes the product type, tracking location, and trigger duration of the tracking location.

[0145] Based on the trigger duration of at least one tracking point, a target score for the user's browsing behavior is obtained.

[0146] Perform similarity matching between the target score and the historical scores of at least two historical users to obtain a similarity matching result of the browsing behavior between the user and each historical user; the historical score is the score of the browsing behavior of the historical user corresponding to the valid order product;

[0147] At least one product is selected from all products according to the similarity matching result, and the at least one product is output.

[0148] In other embodiments of the present application, the tracking point includes a node on the document object model DOM tree corresponding to the product page.

[0149] In other embodiments of the present application, the processor 1002 is configured to execute the product recommendation program stored in the memory 1001 to implement the following steps:

[0150] From at least one trigger duration corresponding to product n, filter out the trigger durations of the buried points that are greater than or equal to the duration threshold; based on the trigger durations of the filtered buried points, obtain the total trigger durations of all buried points corresponding to product n; based on the N total trigger durations corresponding to the N products browsed by the user, obtain the target score of the user's browsing behavior.

[0151] In other embodiments of the present application, the processor 1002 is configured to execute the product recommendation program stored in the memory 1001 to implement the following steps:

[0152] Obtain the weight corresponding to the burial point location; based on the weight corresponding to the burial point location, the trigger duration corresponding to the burial point location, and the memory forgetting function corresponding to the burial point location, obtain the total trigger duration corresponding to product n.

[0153] In other embodiments of the present application, the memory forgetting function is a function that determines that the product n corresponding to the burial point location is gradually forgotten based on the time from the time when the burial point trigger event corresponding to the burial point location is monitored to the current time, and a pre-set memory correction factor.

[0154] In other embodiments of the present application, the processor 1002 is configured to execute the product recommendation program stored in the memory 1001 to implement the following steps:

[0155] Based on N total triggering times, digital characteristic values ​​of the user's browsing behavior are obtained, which include expected value, entropy value and super entropy value; based on the expected value, entropy value and super entropy value, a target score of the user's browsing behavior is obtained.

[0156] In other embodiments of the present application, the processor 1002 is configured to execute the product recommendation program stored in the memory 1001 to implement the following steps:

[0157] Taking the entropy value as the expected value of the first normal distribution function and the super entropy value as the variance of the first normal distribution function, a first random number is obtained through the first normal distribution function; taking the expected value as the expected value of the second normal distribution function and the first random number as the variance of the second normal distribution function, a second random number is obtained through the second normal distribution function; determining the quantitative certainty of the second random number to obtain a target score; wherein, the target score includes the second random number and the quantitative certainty.

[0158] In other embodiments of the present application, the processor 1002 is configured to execute the product recommendation program stored in the memory 1001 to implement the following steps:

[0159] Obtain valid order products generated within the target period. When the corresponding historical users browse the historical product page, obtain the historical total trigger duration of all buried points corresponding to at least two historical products browsed, and obtain the at least two historical total trigger durations corresponding to at least two historical products browsed by the historical users; for at least two valid order products, obtain at least two historical total trigger durations corresponding to at least two products browsed by all historical users corresponding to the at least two valid order products; based on the at least two historical total trigger durations corresponding to at least two products browsed by all historical users, obtain the historical scores of all historical users.

[0160] The present application provides a storage medium storing one or more programs, which can be executed by one or more processors to implement the following Figure 1 、 Figures 4 to 8 A corresponding product recommendation method is provided.

[0161] The present application provides a storage medium, which, if a point-triggering event on a product page is monitored, collects the point-triggering data of at least one point of the product n corresponding to the point-triggering event on the product page, the point-triggering data including the product type, point-triggering location and triggering duration of the point-triggering location; obtains a target score of the user's browsing behavior based on the triggering duration corresponding to at least one point; performs similarity matching on the target score with the historical scores corresponding to at least two historical users, and obtains a similarity matching result of the browsing behavior between the user and each historical user; the historical score is the score of the browsing behavior of the historical user corresponding to the valid order product; selects at least one product from all products according to the similarity matching result, and outputs at least one product; in this way, the present application sets tracking points at the product granularity, and there is no need to collect a large amount of user behavior information in multiple dimensions. It only collects the trigger duration of the user's tracking points for different products in different modules on different pages, that is, only collects time dimension data; further, the electronic device obtains a target score for characterizing the user's browsing behavior characteristics based on the trigger duration corresponding to the user's browsing of different products, and matches the target score for characterizing the current user's browsing behavior characteristics with the historical score of the historical user who generated valid order products, and selects the product for output based on the matching result; in this way, qualitative results are obtained through matching scores, so as to recommend relevant product content, thereby reducing the amount of data and reducing the complexity of the algorithm.

[0162] It should be noted that the above-mentioned computer storage medium / memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a magnetic random access memory (FRAM), a flash memory (Flash Memory), a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM) and other memories; it can also be various electronic devices that include one or any combination of the above-mentioned memories, such as mobile phones, computers, tablet devices, personal digital assistants, etc.

[0163] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0164] In addition, the functional units in the embodiments of the present application can all be integrated into one processing module, or each unit can be a separate unit, or two or more units can be integrated into one unit; the above-mentioned integrated unit can be implemented in the form of hardware or in the form of hardware plus software functional units. It can be understood by ordinary technicians in this field that all or part of the steps of the above-mentioned method embodiments can be completed by hardware related to program instructions, and the above-mentioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above-mentioned method embodiments; and the above-mentioned storage medium includes: mobile storage devices, read-only memory (ROM), random access memory (RAM), disks or optical disks, etc. Various media that can store program codes.

[0165] The methods disclosed in the several method embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments.

[0166] The features disclosed in the several method or device embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments or device embodiments.

[0167] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A product recommendation method, characterized in that: The method comprises: If a tracking trigger event is detected on a product page, tracking data for at least one tracking point of product n corresponding to the tracking trigger event on the product page is collected. The tracking data includes product type, tracking location, and trigger duration of the tracking location. Obtaining a target score for the user's browsing behavior based on the trigger duration corresponding to the at least one tracking point; Perform similarity matching between the target score and the historical scores corresponding to at least two historical users, and obtain a similarity matching result of the browsing behavior between the user and each historical user; the historical score is the score of the browsing behavior of the historical user corresponding to the valid order product; selecting at least one product from all products according to the similarity matching result, and outputting the at least one product; The step of obtaining a target score for the user's browsing behavior based on the trigger duration corresponding to the at least one tracking point includes: Filter out the trigger duration of the tracking point that is greater than or equal to the duration threshold from at least one trigger duration corresponding to the product n; Based on the trigger duration of the filtered tracking points, the total trigger duration of all tracking points corresponding to the product n is obtained, including: obtaining the weight corresponding to the tracking point at the tracking point position; Based on the weight corresponding to the embedding location, the triggering duration corresponding to the embedding location, and the memory forgetting function corresponding to the embedding location, the total triggering duration corresponding to the product n is obtained; Based on the N total trigger durations corresponding to the N products browsed by the user, a target score for the user's browsing behavior is obtained, including: Based on the N total trigger durations, a digital characteristic value of the user's browsing behavior is obtained, where the digital characteristic value includes an expected value, an entropy value, and a super entropy value; Obtaining a target score for the user's browsing behavior based on the expected value, the entropy value, and the super entropy value; Wherein, obtaining a target score for the user's browsing behavior based on the expected value, the entropy value, and the super entropy value includes: Taking the entropy value as the expected value of a first normal distribution function and the super entropy value as the variance of the first normal distribution function, a first random number is obtained through the first normal distribution function; Taking the expected value as the expected value of a second normal distribution function and the first random number as the variance of the second normal distribution function, and obtaining a second random number through the second normal distribution function; Determine a quantitative certainty of the second random number to obtain the target score; wherein the target score includes the second random number and the quantitative certainty.

2. The method according to claim 1, characterized in that The embedding point includes a node on the document object model DOM tree corresponding to the product page.

3. The method according to claim 1, characterized in that The memory forgetting function is a function that determines how the product n corresponding to the burial point location is gradually forgotten based on the time from the time when the burial point trigger event corresponding to the burial point location is monitored to the current time, and a pre-set memory correction factor.

4. The method according to claim 1, wherein Before performing similarity matching between the target score and the obtained historical scores corresponding to at least two historical users to obtain similarity matching results of browsing behaviors between the user and each historical user, the method includes: Obtain the historical total trigger duration of all tracking points corresponding to at least two historical products browsed by the corresponding historical user when the product generated a valid order within the target time period, and obtain the historical total trigger duration of at least two historical products browsed by the historical user; For at least two valid order products, obtaining at least two historical total trigger durations corresponding to at least two products browsed by all historical users corresponding to the at least two valid order products; Based on at least two historical total trigger durations corresponding to at least two products browsed by all historical users, historical scores of all historical users are obtained.

5. An electronic device, characterized in that: The electronic device comprises: A memory for storing executable instructions; a processor for executing the executable instructions stored in the memory to implement the product recommendation method according to any one of claims 1 to 4.

6. A storage medium, characterized in that The storage medium stores one or more programs, and the one or more programs are executed by one or more processors to implement the product recommendation method according to any one of claims 1 to 4.

Images