Prop recommendation method, device, equipment and medium

By combining the purchase time and similarity of the users' recent purchase of props, the recommendation weight is calculated, and the problem of insufficient accuracy of props recommendations in the existing technology is solved, and higher recommendation accuracy is achieved.

CN114504829BActive Publication Date: 2024-11-26TENCENT TECHNOLOGY (SHENZHEN) CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210108003.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-01-28
Publication Date
2024-11-26
Estimated Expiration
2042-01-28

AI Technical Summary

Technical Problem

The prior art cannot effectively capture changes in user interests, resulting in insufficient accuracy of prop recommendations.

Method used

By combining the purchase time of the user's recent purchase of props and the similarity of similar props, the recommendation weight of historical purchases is calculated, and the weighted recommendation degree of candidate recommendation props is determined, and the props are recommended to users.

Benefits of technology

It improves the accuracy of prop recommendations, can capture changes in user interests more accurately, and improves the accuracy of recommendations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114504829B_ABST
    Figure CN114504829B_ABST
Patent Text Reader

Abstract

The present application discloses a prop recommendation method, device, equipment and medium, which relates to the field of machine learning. The method includes: counting similar props of each historically purchased prop in the historical purchase record of the user account to obtain a candidate recommendation set; calculating the recommendation weight of each historically purchased prop according to the prop purchase time and the prop similarity, the prop purchase time is used to measure the influence of the purchase time of the historically purchased prop on the recommended prop, and the prop similarity is used to measure the influence of the similar prop on the recommended prop; according to the recommendation weight of each historically purchased prop and each candidate recommended prop in the candidate recommendation set, determining the weighted recommendation degree of each candidate recommended prop; according to the ranking of the weighted recommendation degree of each candidate recommended prop, determining the recommended prop recommended to the user account. The present application can improve the accuracy of recommended props.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of machine learning, and in particular to a prop recommendation method, device, equipment and medium. Background Art

[0002] Providing recommended items to users allows them to purchase their favorite items conveniently and quickly, saving time and improving user experience.

[0003] The related technology will first obtain the user's historical purchase records, and determine the candidate recommendation set based on the similar items of each historically purchased item in the historical purchase records. Calculate the similarity between the candidate recommended items in the candidate recommendation set and each historically purchased item, and then obtain the recommendation degree of the candidate recommended items. Take the candidate recommended items with the largest recommendation degrees and recommend them to the user.

[0004] However, users' interests change dynamically, and related technologies cannot capture changes in users' interests. Summary of the invention

[0005] The embodiment of the present application provides a prop recommendation method, device, equipment and medium. The method takes into account the props recently purchased by the user and the user's preferences to improve the accuracy of the prop recommendation. The technical solution is as follows:

[0006] According to one aspect of the present application, a prop recommendation method is provided, the method comprising:

[0007] Count the similar items of each historically purchased item in the historical purchase records of the user account to obtain a candidate recommendation set;

[0008] Calculate the recommendation weight of each of the historically purchased items according to the item purchase time and item similarity, wherein the item purchase time is used to measure the influence of the purchase time of the historically purchased item on the recommended item, and the item similarity is used to measure the influence of the similar item on the recommended item;

[0009] Determining a weighted recommendation degree of each candidate recommended item according to the recommendation weight of each historically purchased item and each candidate recommended item in the candidate recommendation set;

[0010] The recommended items to be recommended to the user account are determined according to the ranking of the weighted recommendation degrees of the candidate recommended items.

[0011] According to one aspect of the present application, a prop recommendation device is provided, the device comprising:

[0012] A statistical module is used to count similar items of each historically purchased item in the historical purchase records of the user account to obtain a candidate recommendation set;

[0013] A calculation module, used to calculate the recommendation weight of each of the historically purchased props according to the prop purchase time and the prop similarity, wherein the prop purchase time is used to measure the influence of the purchase time of the historically purchased prop on the recommended prop, and the prop similarity is used to measure the influence of the similar prop on the recommended prop;

[0014] The calculation module is further used to determine the weighted recommendation degree of each candidate recommended item according to the recommendation weight of each historically purchased item and each candidate recommended item in the candidate recommendation set;

[0015] The calculation module is further used to determine the recommended props to be recommended to the user account according to the ranking of the weighted recommendation degrees of the candidate recommended props.

[0016] According to another aspect of the present application, a computer device is provided, comprising: a processor and a memory, wherein at least one instruction, at least one program, a code set or an instruction set is stored in the memory, and the at least one instruction, at least one program, a code set or an instruction set is loaded and executed by the processor to implement the prop recommendation method as described above.

[0017] According to another aspect of the present application, a computer storage medium is provided, in which at least one program code is stored, and the program code is loaded and executed by a processor to implement the prop recommendation method as described above.

[0018] According to another aspect of the present application, a computer program product or a computer program is provided, wherein the computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the item recommendation method described in the above aspect.

[0019] The beneficial effects brought by the technical solution provided by the embodiment of the present application include at least:

[0020] Since the purchase time of historically purchased items can reflect the impact of the user's recent purchases of items on recommended items, and similar items can reflect the impact of the user's interests on recommended items, this method takes into account the impact of the purchase time of historically purchased items and similar items on recommended items, and can accurately recommend items to users with a high accuracy rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0022] Figure 1 is a schematic diagram of the structure of a computer system provided by an exemplary embodiment of the present application;

[0023] Figure 2 is a flowchart of a prop recommendation method provided by an exemplary embodiment of the present application;

[0024] Figure 3 is a flowchart of a prop recommendation method provided by an exemplary embodiment of the present application;

[0025] Figure 4 is a flowchart of determining time weight provided by an exemplary embodiment of the present application;

[0026] Figure 5 is a schematic diagram of a structure for determining similarity weights provided by an exemplary embodiment of the present application;

[0027] Figure 6 is a flowchart of game props recommendation provided by an exemplary embodiment of the present application;

[0028] Figure 7 is a schematic diagram of an interface of a method for recommending game props provided by an exemplary embodiment of the present application;

[0029] Figure 8 is a schematic diagram of a method for recommending game props provided by an exemplary embodiment of the present application;

[0030] Fig. 9 is a schematic diagram of an interface of a method for recommending game props provided by an exemplary embodiment of the present application;

[0031] Fig.10 is a structural schematic diagram of a prop recommendation device provided by an exemplary embodiment of the present application;

[0032] Fig.11 It is a structural diagram of a computer device provided by an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0033] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below with reference to the accompanying drawings.

[0034] First, the nouns involved in the embodiments of the present application are introduced:

[0035] Artificial Intelligence (AI): It is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence. Artificial intelligence is to study the design principles and implementation methods of various intelligent machines, so that machines have the functions of perception, reasoning and decision-making.

[0036] Artificial intelligence technology is a comprehensive discipline that covers a wide range of fields, including both hardware-level and software-level technologies. The basic technologies of artificial intelligence generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics and other technologies. Artificial intelligence software technology mainly includes computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0037] Machine Learning (ML): It is a multi-disciplinary interdisciplinary subject, involving probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and other disciplines. It specializes in studying how computers simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent. Its applications are spread across all areas of artificial intelligence. Machine learning and deep learning usually include artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and teaching learning.

[0038] Collaborative filtering: It is an algorithm that predicts the content that users may be interested in and recommends the props to users by analyzing the similarities between users or props.

[0039] With the research and advancement of artificial intelligence technology, artificial intelligence technology has been studied and applied in many fields, such as common smart homes, smart wearable devices, virtual assistants, smart speakers, smart marketing, unmanned driving, automatic driving, drones, robots, smart medical care, smart customer service, etc. I believe that with the development of technology, artificial intelligence technology will be applied in more fields and play an increasingly important role.

[0040] Figure 1 The schematic diagram of the structure of a computer system provided by an exemplary embodiment of the present application is shown. The computer system 100 includes: a terminal 120 and a server 140 .

[0041] The terminal 120 is installed with an application program related to item recommendation. The application program may be a small program in an app (application), a dedicated application program, or a web client. The terminal 120 is at least one of a smart phone, a tablet computer, an e-book reader, an MP3 player, an MP4 player, a laptop computer, and a desktop computer.

[0042] The terminal 120 is connected to the server 140 via a wireless network or a wired network.

[0043] The server 140 may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms. The server 140 is used to provide background services and send the results to the terminal 120. Optionally, the server 140 undertakes the main computing work and the terminal 120 undertakes the secondary computing work; or, the server 140 undertakes the secondary computing work and the terminal 120 undertakes the main computing work; or, the server 140 and the terminal 120 both adopt a distributed computing architecture for collaborative computing.

[0044] Figure 2 FIG. 1 shows a flow chart of a prop recommendation method provided by an embodiment of the present application. The method can be Figure 1 The terminal 120 or the server 140 shown in the figure performs the method, which includes the following steps:

[0045] Step 202: Count similar items of each historically purchased item in the historical purchase records of the user account to obtain a candidate recommendation set.

[0046] In the embodiment of the present application, props refer to at least one of commodities, services, and virtual props in games.

[0047] Optionally, historical purchase records of the user account are obtained in compliance with laws and regulations and with user authorization.

[0048] Similar items refer to items whose similarity to historically purchased items is less than a similarity threshold. For example, taking a historically purchased item A as an example, the cosine similarity between the candidate item and the historically purchased item A is calculated, and items whose cosine similarity is less than the similarity threshold are determined as similar items to the historically purchased item A.

[0049] Optionally, the item neighbor model is called to process the data of each historically purchased item to obtain similar items of each historically purchased item.

[0050] The candidate recommendation set is a set of candidate recommendation items. Optionally, the candidate recommendation items in the candidate recommendation set are selected from similar items of each historically purchased item. For example, the candidate recommendation set is obtained by randomly selecting candidate recommendation items from similar items of each historically purchased item. Alternatively, at least two similar items with the largest number of purchases are positioned as candidate recommendation items in the candidate recommendation set based on the number of purchases of similar items of each historically purchased item.

[0051] Optionally, similar props of each historically purchased prop in the historical purchase record are determined. Since different historically purchased props may have the same similar props, for example, the similar props of historically purchased prop A are props X and prop Y, and the similar props of historically purchased prop B are props X and prop Z, where prop X is the same prop. At this time, the aforementioned same similar props are redundant. Therefore, the repeated props in the similar props of each historically purchased prop are deleted to obtain a candidate recommendation set, and the repeated props refer to the props whose number of props in the similar props of each historically purchased prop is greater than 1.

[0052] Step 204: Calculate the recommendation weight of each historically purchased item based on the item purchase time and item similarity. The item purchase time is used to measure the influence of the purchase time of the historically purchased item on the recommended item, and the item similarity is used to measure the influence of similar items on the recommended item.

[0053] Optionally, the recommendation weight includes a time weight and a similarity weight. The time weight is a weight related to the time of item purchase, and the similarity weight is a weight related to the similarity of the item.

[0054] The applicant has found that the interests of users change over time, but are relatively stable over a short period of time. Therefore, the props that a user is interested in are most likely similar to the props that the user has recently purchased. Optionally, the time weight of each historically purchased prop is calculated based on the purchase time of the props. The time weight is negatively correlated with the difference between the purchase time of the historically purchased props and the current time, that is, the closer the purchase time of the historically purchased props is to the current time, the greater the time weight.

[0055] On the other hand, the applicant found that in reality, the speed and regularity of interest changes of different users are different. In addition, users' interests often change repeatedly, so the props purchased by users in the early stage are also important for determining the recommended props. Simply using only the time weight may have a negative impact on the recommendation effect. Therefore, similarity weights based on prop similarity are introduced to optimize the recommendation weights. Optionally, based on the prop similarity, the similarity weights of each historically purchased prop are calculated.

[0056] Step 206: Determine the weighted recommendation degree of each candidate recommended item according to the recommendation weight of each historically purchased item and each candidate recommended item in the candidate recommendation set.

[0057] In one implementation of the present application, the similarity between the p-th historical purchase item and the q-th candidate recommended item is calculated, where p and q are positive integers and their initial values ​​are 1; the product of the recommendation weight and the similarity of the p-th historical purchase item is calculated to obtain the weighted similarity of the p-th historical purchase item; p is updated to p+1, and the above two steps are repeated k times until k weighted similarities of the q-th candidate recommended item are obtained, where k represents the number of historical purchase items; the sum of the k weighted similarities of the q-th candidate recommended item is calculated to obtain the weighted recommendation degree of the q-th candidate recommended item; q is updated to q+1, p is initialized, and the above four steps are repeated until the weighted recommendation degree of each candidate recommended item is obtained.

[0058] In an embodiment of the present application, a prop recommendation model can also be used to determine the weighted recommendation degree. Exemplarily, the prop recommendation model is called to process the historical purchase records and the candidate recommendation set, and the weighted recommendation degree of each candidate recommended prop is output. Among them, the prop recommendation model includes a time weight calculation network and a similarity weight calculation network. The time weight calculation network is used to calculate the time weight of each historically purchased prop, and the similarity weight calculation network is used to calculate the similarity weight of each historically purchased prop. Optionally, the time weight calculation network and the similarity weight calculation network are both fully connected feedforward networks.

[0059] In the embodiment of the present application, different user accounts correspond to different item recommendation models, for example, user account A corresponds to item recommendation model 1, and user account B corresponds to item recommendation model 2. That is, each item recommendation model is specifically used for a certain user account.

[0060] Optionally, the above-mentioned prop recommendation model will be updated in real time. For example, when a similar account of a user account makes a selection of a recommended prop, the prop recommendation model will be trained for a second time. Optionally, similar user accounts of the user account are determined; and the prop recommendation model is trained for a second time based on the purchase of recommended props by similar user accounts. Exemplarily, a similar user account of user account A is user account B. After user account B obtains the recommended props, it chooses to purchase recommended props 1 and recommended props 2, but does not purchase recommended props 3. At this time, the purchase of recommended props by user account B is recorded. The prop recommendation model belonging to user account A is trained for a second time based on the purchase of recommended props by user account B. Optionally, user accounts whose similarity with the user account is less than a similarity threshold are determined as similar accounts.

[0061] Step 208: Determine the recommended items to be recommended to the user account according to the ranking of the weighted recommendation degrees of each candidate recommended item.

[0062] Optionally, each candidate recommended item is sorted in descending order according to the weighted recommendation degree, and the top N candidate recommended items are determined as recommended items, where N is a positive integer and can be set by the technician.

[0063] Optionally, the candidate recommended item corresponding to the maximum weighted recommendation degree is determined as the recommended item of the user account.

[0064] To sum up, in this embodiment, since the purchase time of historically purchased props can reflect the impact of the user's recent purchase of props on the recommended props, and similar props can reflect the impact of the user's interests on the recommended props, this method takes into account the impact of the purchase time of the above-mentioned historically purchased props and similar props on the recommended props, and can accurately recommend props to users with a high accuracy rate.

[0065] Figure 3 FIG. 1 shows a flow chart of a prop recommendation method provided by an embodiment of the present application. The method can be Figure 1 The terminal 120 or the server 140 shown in the figure performs the method, which includes the following steps:

[0066] Step 301: Count similar items of each historically purchased item in the historical purchase records of the user account to obtain a candidate recommendation set.

[0067] In the embodiment of the present application, props refer to at least one of commodities, services, and virtual props in games.

[0068] Optionally, historical purchase records of the user account are obtained in compliance with laws and regulations and with user authorization.

[0069] The candidate recommendation set is a set of candidate recommendation items. Optionally, the candidate recommendation items in the candidate recommendation set are selected from similar items of each historically purchased item. For example, the candidate recommendation set is obtained by randomly selecting candidate recommendation items from similar items of each historically purchased item. Alternatively, at least two similar items with the largest number of purchases are positioned as candidate recommendation items in the candidate recommendation set based on the number of purchases of similar items of each historically purchased item.

[0070] Similar items refer to items whose similarity to historically purchased items is less than the similarity threshold. For example, similar items to historically purchased item A are item X and item Y, and similar items to historically purchased item B are item X and item Z, where item X is the same item.

[0071] Step 302: Calculate the time weight of the i-th historically purchased item in the historical purchase record according to the item purchase time.

[0072] Wherein, i is a positive integer, and the initial value of i is 1. The i-th historical purchase item is any item in the historical purchase record.

[0073] Optionally, this step includes the following sub-steps:

[0074] Step 41: Determine the time interval between the purchase time of the i-th historically purchased item and the purchase time of the first purchased item.

[0075] The first purchase item refers to the earliest purchase item in the historical purchase record of the user account. For example, the historical purchase record will record the items purchased by the user account in the historical period. For example, the historical purchase record will record the items purchased by the user account in the past 10 days, and the earliest purchase item in the historical purchase record will be determined as the first purchase item.

[0076] Exemplarily, the first item purchased by the user after registering the user account is determined as the first-time purchased item.

[0077] For example, the time interval is recorded as D ui , where u represents the user account.

[0078] Step 42: Determine the time span from the current time to the last time the recommended item was obtained.

[0079] The prop recommendation method provided in the embodiment of the present application can be reused. For example, in a game scenario, as long as the user logs in to the game, the game system will recommend props to the user. When the user logs in to the game on January 2, the game system detects that the user last logged in to the game on January 1, and the period from January 2 to January 1 is used as the time span.

[0080] For example, in an online shopping scenario, in response to a product recommendation operation, the online shopping system provides a product recommendation service product to the user account, wherein the online shopping system determines the recommended product by the method provided in the embodiment of the present application. When the user uses the product recommendation service of the online shopping system on January 20, the online shopping system detects that the user last used the product recommendation service on January 10, and then uses the period from January 20 to January 10 as the time span.

[0081] For example, the time span is denoted as L u , where u represents the user account.

[0082] Step 43: According to the ratio of the time interval and the time span, the time weight of the i-th historical purchase item is obtained.

[0083] For example, the time weight of the i-th historical purchase item:

[0084]

[0085] Among them, α is a positive coefficient, and the value of α can be set by the technician according to actual needs. ui Indicates the time interval, L u Indicates a time span.

[0086] Therefore, WT(u,i) is about D ui A non-decreasing function of ui >D uj WT(u,i)>=WT(u,j).

[0087] Step 303: Calculate the similarity weight of the i-th historically purchased item based on the item similarity.

[0088] Optionally, this step includes the following sub-steps:

[0089] Step 51: extract a subset of historical purchase records from the historical purchase records according to the time window.

[0090] The length of the time window can be set by the technician, for example, the length of the time window is 7 days, 10 days, 20 days, 30 days, etc. The embodiment of the present application does not limit the length of the time window.

[0091] Optionally, the time window is set with the current time as the right end point of the time window. For example, if the current time is January 30 and the length of the time window is 7 days, then after the time window is set, the time in the time window is from January 24 to January 30.

[0092] Optionally, the historical purchase records within the time window are used to determine the historical purchase record subset. For example, the historical purchase records include purchase records from January 1 to January 30, and the time window is from January 20 to January 30, then the historical purchase subset refers to the purchase records from January 20 to January 30.

[0093] For example, the historical purchase subset is recorded as I uT , where u represents the user account and T represents the time window.

[0094] Step 52: Calculate the overall similarity between the i-th historical purchase item and the historical purchase record subset.

[0095] Optionally, the similarity between the ith historical purchase item and each historical purchase item in the historical purchase record subset is calculated first. Then the average of the similarities is calculated to obtain the overall similarity. For example, the overall similarity is Among them, j represents the jth historical purchase item in the historical purchase record subset, and sim(i, j) represents the similarity between the ith historical purchase item and the jth historical purchase item.

[0096] Step 53: According to the ratio between the overall similarity and the number of items in the subset of historical purchase records, the similarity weight of the i-th historical purchase item is obtained.

[0097] Exemplary, similarity weights:

[0098]

[0099] Among them, size(I uT ) represents the historical purchase record subset I uT The number of props.

[0100] It should be noted that the order of step 302 and step 303 can be swapped, and step 302 can be performed first and then step 303. Step 303 can also be performed and then step 302. The embodiment of the present application does not limit this.

[0101] Step 304: weighted calculation of the time weight of the i-th historically purchased item and the similarity weight of the i-th historically purchased item to obtain the recommendation weight of the i-th historically purchased item.

[0102] Time weight can reflect the importance of recently purchased items, so as to capture the user's current interests in a timely manner, which is suitable for handling situations where user interests change frequently. Similarity weight can reflect the relevance of items that users have purchased in the past and their current interests, avoiding the neglect of valuable early purchase data, which is suitable for handling situations where user interests are repeated. Therefore, consider combining the two weight strategies with a certain proportional factor β and define the recommendation weight as:

[0103] WTS(u,i)=β×WT(u,i)+(1-β)×WS(u,i);

[0104] Among them, the proportional factor β∈[0,1], β and 1-β respectively represent the proportions of the two weights. The value of β can be set by the technician according to actual needs.

[0105] Step 305: Update i to i+1, and repeat the above three steps until the recommendation weight of each historically purchased item is obtained.

[0106] In the process from step 302 to step 304, only the i-th historically purchased item is obtained, so it is necessary to repeat steps 302 to step 304 to obtain the recommendation weight of each historically purchased item.

[0107] Step 306: Calculate the similarity between the pth historically purchased item and the qth candidate recommended item.

[0108] Wherein, p and q are positive integers, and the initial values ​​of p and q are 1.

[0109] Optionally, the cosine similarity between the pth historically purchased item and the qth candidate recommended item is calculated. For example, the cosine similarity is:

[0110]

[0111] Among them, p represents the p-th historical purchase item, q represents the q-th candidate recommended item, ‖p‖ represents the size of the vector corresponding to the p-th historical purchase item, and ‖q‖ represents the size of the vector corresponding to the q-th candidate recommended item.

[0112] Step 307: Calculate the product of the recommendation weight and the similarity of the p-th historically purchased item to obtain the weighted similarity of the p-th historically purchased item.

[0113] Exemplarily, the weighted similarity of the pth historically purchased item is WTS(u, p)×sim(p, q), where WTS(u, p) represents the recommendation weight of the pth historically purchased item, u represents the user account, and sim(p, q) represents the similarity.

[0114] Step 308: Update p to p+1, and repeat the above two steps k times until k weighted similarities of the qth candidate recommended item are obtained.

[0115] k represents the number of historically purchased items.

[0116] In the embodiment of the present application, it is necessary to calculate the weighted similarity between the qth candidate recommended item and each historically purchased item, and determine the weighted recommendation degree of the qth candidate recommended item through the weighted similarity.

[0117] Step 309: Calculate the sum of the k weighted similarities of the qth candidate recommended item to obtain the weighted recommendation degree of the qth candidate recommended item.

[0118] For example, the weighted recommendation degree of the qth candidate recommendation item is:

[0119]

[0120] Among them, u represents the user account, I u represents the historical purchase record, WTS(u, p) represents the recommendation weight of the qth candidate recommended item, and sim(p, q) represents the similarity between the qth candidate recommended item and the pth historical purchase item.

[0121] Step 310: Update q to q+1, initialize p, and repeat the above four steps until the weighted recommendation degree of each candidate recommended item is obtained.

[0122] Since steps 306 to 309 obtain the weighted recommendation degree of the qth candidate recommendation item, and the candidate recommendation set also includes other candidate recommendation items, steps 306 to 309 need to be repeated to obtain the weighted recommendation degrees of all candidate recommendation items in the candidate recommendation set.

[0123] Step 311: Determine the recommended items to be recommended to the user account according to the ranking of the weighted recommendation degrees of each candidate recommended item.

[0124] Optionally, each candidate recommended item is sorted in descending order according to the weighted recommendation degree, and the top N candidate recommended items are determined as recommended items, where N is a positive integer and can be set by the technician.

[0125] Optionally, the candidate recommended item corresponding to the maximum weighted recommendation degree is determined as the recommended item of the user account.

[0126] In summary, in this embodiment, since the purchase time of historically purchased items can reflect the impact of the user's recent purchase of items on the recommended items, and similar items can reflect the impact of the user's interests on the recommended items, this method takes into account the impact of the purchase time of historically purchased items and similar items on the recommended items, and can accurately recommend items to users with a high accuracy rate. Moreover, the method for calculating the weighted recommendation degree is relatively simple, the amount of calculation is small, and the recommended items of the user account can be quickly calculated, with good real-time performance.

[0127] In the following embodiments, the prop recommendation method is applied in a game as an example for explanation, as shown below:

[0128] Figure 6 FIG. 1 shows a flow chart of a prop recommendation method provided by an embodiment of the present application. The method can be Figure 1 The computer system 100 shown in the figure executes the method, which includes the following steps:

[0129] Step 601: The client displays the game interface.

[0130] For example, Figure 7 As shown, a game item recommendation control 701 is displayed on the game interface.

[0131] Optionally, the game interface also displays at least one of virtual resources, virtual objects, battle controls, task controls, purchase controls, and activity controls. The battle controls are used to start a game battle, for example, Figure 7In the game interface shown, "Weekly Tournament" and "Classic Mode" are the battle controls. The task control is used to view game tasks, for example, Figure 7 In the game interface shown, "Task" is the task control. The purchase control is used to display the game store or purchase game props. The activity control is used to display game activities.

[0132] Step 602: In response to the game item recommendation operation on the client, the client sends an item recommendation request to the server.

[0133] Optionally, in response to a trigger operation on the game prop recommendation control, the client sends a recommended prop request to the server. Figure 7 As shown, the user clicks the game prop recommendation control 701 to complete the game prop recommendation operation.

[0134] For example, Figure 8 As shown, the client sends a prop recommendation request to the online recommendation service of the server.

[0135] Step 603: The server counts similar props of each historically purchased game prop in the historical purchase records of the user account to obtain a candidate recommendation set.

[0136] The historical purchase record includes at least one of user account information, purchase time, and game item type.

[0137] For example, Figure 8 As shown, the server stores historical purchase records of different user accounts, and the server determines the historical purchase records of the current user account.

[0138] Step 604: The server calculates the recommendation weight of each historically purchased game item based on the item purchase time and item similarity.

[0139] Optionally, the recommendation weight includes a time weight and a similarity weight. The time weight is a weight related to the time of purchasing the game props, and the similarity weight is a weight related to the similarity of the props.

[0140] Exemplarily, the time weight of each historically purchased game prop is calculated based on the purchase time of the prop, and the time weight is negatively correlated with the difference between the purchase time of the historically purchased game prop and the current time, that is, the closer the purchase time of the historically purchased game prop is to the current time, the greater the time weight.

[0141] Optionally, based on the item similarity, similarity weights of the items purchased in each historical game are calculated.

[0142] Step 605: The server determines the weighted recommendation degree of each candidate recommended game prop according to the recommendation weight of each historically purchased game prop and each candidate recommended game prop in the candidate recommendation set.

[0143] In one implementation of the present application, the similarity between the pth historically purchased game item and the qth candidate recommended game item is calculated, where p and q are positive integers and their initial values ​​are 1; the product of the recommendation weight and the similarity of the pth historically purchased game item is calculated to obtain the weighted similarity of the pth historically purchased game item; p is updated to p+1, and the above two steps are repeated k times until k weighted similarities of the qth candidate recommended game item are obtained, where k represents the number of historically purchased game items; the sum of the k weighted similarities of the qth candidate recommended game item is calculated to obtain the weighted recommendation degree of the qth candidate recommended game item; q is updated to q+1, p is initialized, and the above four steps are repeated until the weighted recommendation degrees of each candidate recommended game item are obtained.

[0144] For example, Figure 8 As shown, the server obtains the weighted recommendation degree of each candidate recommended game item through the combined time weight and similarity weight, as well as the historical purchase record.

[0145] Step 606: The server determines the recommended game props to recommend to the user account according to the ranking of the weighted recommendation degrees of each candidate recommended game prop.

[0146] Optionally, the server sorts each candidate recommended game prop in descending order according to the weighted recommendation degree, and determines the top N candidate recommended game props as recommended game props, where N is a positive integer and can be set by the technician.

[0147] Step 607: The server returns the recommended game props to the client.

[0148] Alternatively, if Figure 8 As shown, the server stores the recommended game props of the user account in a remote dictionary server (Remote Dictionary Server, Redis), and provides the recommended game props to the client through the remote dictionary service.

[0149] Step 608: The client displays recommended game props on the game interface.

[0150] Alternatively, if Figure 8 As shown, the client obtains recommended game props through the remote dictionary service.

[0151] For example, Fig. 9 As shown, recommended game props 901 are displayed on the game interface, wherein the recommended game props 901 may include multiple different types of game props.

[0152] Optionally, the recommended game props need to be purchased using different virtual resources, for example, the recommended game prop 1 needs to cost 200 gold coins, and the recommended game prop 2 needs to cost 20 diamonds.

[0153] In summary, this embodiment can accurately recommend game props to users, and the accuracy of the recommendation is high. For users, they can buy their favorite game props more conveniently and quickly, saving the time cost of searching and then buying props, and improving the user's gaming experience. For game merchants, more targeted and personalized prop recommendations will stimulate users' desire to buy, thereby increasing the commercial value of the game.

[0154] Please refer to Fig.10 , which shows a block diagram of a prop recommendation device provided by an embodiment of the present application. The above functions can be implemented by hardware, or by hardware executing corresponding software. The device 1000 includes:

[0155] The statistical module 1001 is used to count similar items of each historically purchased item in the historical purchase records of the user account to obtain a candidate recommendation set;

[0156] The calculation module 1002 is used to calculate the recommendation weight of each of the historically purchased props according to the prop purchase time and the prop similarity, wherein the prop purchase time is used to measure the influence of the purchase time of the historically purchased prop on the recommended prop, and the prop similarity is used to measure the influence of the similar prop on the recommended prop;

[0157] The calculation module 1002 is further used to determine the weighted recommendation degree of each candidate recommended item according to the recommendation weight of each historically purchased item and each candidate recommended item in the candidate recommendation set;

[0158] The calculation module 1002 is further configured to determine the recommended item to be recommended to the user account according to the ranking of the weighted recommendation degrees of the candidate recommended items.

[0159] In an optional design of the present application, the calculation module 1002 is further used to calculate the time weight of the i-th historical purchase item in the historical purchase record according to the item purchase time, the time weight and the difference between the purchase time of the i-th historical purchase item and the current time are negatively correlated, i is a positive integer, and the initial value of i is 1; calculate the similarity weight of the i-th historical purchase item according to the item similarity, the similarity weight is used to indicate the degree of correlation between the i-th historical purchase item and the user portrait of the user account; weightedly calculate the time weight of the i-th historical purchase item and the similarity weight of the i-th historical purchase item to obtain the recommended weight of the i-th historical purchase item; update i to i+1, and repeat the above three steps until the recommended weights of the various historical purchase items are obtained.

[0160] In an optional design of the present application, the calculation module 1002 is also used to determine the time interval between the purchase time of the i-th historical purchase item and the purchase time of the first purchase item, the first purchase item refers to the earliest purchase item in the historical purchase record of the user account; determine the time span from the current time to the last time the recommended item was obtained; and obtain the time weight of the i-th historical purchase item based on the ratio of the time interval and the time span.

[0161] In an optional design of the present application, the calculation module 1002 is also used to extract a subset of historical purchase records from the historical purchase records according to a time window; calculate the overall similarity between the i-th historical purchase item and the subset of historical purchase records; and obtain the similarity weight of the i-th historical purchase item according to the ratio between the overall similarity and the number of items in the subset of historical purchase records.

[0162] In an optional design of the present application, the calculation module 1002 is further used to calculate the similarity between the i-th historical purchase item and each historical purchase item in the historical purchase record subset; and calculate the sum of the similarities to obtain the overall similarity.

[0163] In an optional design of the present application, the calculation module 1002 is also used to calculate the similarity between the p-th historical purchase item and the q-th candidate recommended item; calculate the product of the recommendation weight of the p-th historical purchase item and the similarity to obtain the weighted similarity of the p-th historical purchase item; update p to p+1, repeat the above two steps k times, until k weighted similarities of the q-th candidate recommended item are obtained, k represents the number of the historical purchase items; calculate the sum of the k weighted similarities of the q-th candidate recommended item, and obtain the weighted recommendation degree of the q-th candidate recommended item; update q to q+1, initialize p, and repeat the above four steps until the weighted recommendation degree of each candidate recommended item is obtained.

[0164] In an optional design of the present application, the calculation module 1002 is further used to calculate the cosine similarity between the p-th historically purchased item and the q-th candidate recommended item.

[0165] In an optional design of the present application, the statistical module 1001 is also used to determine similar props of each historically purchased prop in the historical purchase record; delete duplicate props in the similar props of each historically purchased prop to obtain the candidate recommendation set, and the duplicate props refer to props whose number of props is greater than 1 in the similar props of each historically purchased prop.

[0166] In an optional design of the present application, the calculation module 1002 is further used to sort the candidate recommended items in descending order according to the weighted recommendation degree; and determine the first N candidate recommended items as the recommended items, where N is a positive integer.

[0167] In an optional design of the present application, the calculation module 1002 is also used to call the prop recommendation model, perform data processing on the historical purchase records and the candidate recommendation set, and output the weighted recommendation degree of each candidate recommended prop. The prop recommendation model includes a time weight calculation network and a similarity weight calculation network. The time weight calculation network is used to calculate the time weight of each historically purchased prop, and the similarity weight calculation network is used to calculate the similarity weight of each historically purchased prop.

[0168] In an optional design of the present application, the calculation module 1002 is also used to determine similar user accounts of the user account; and perform secondary training on the item recommendation model based on the recommended item purchase status of the similar user accounts.

[0169] To sum up, in this embodiment, since the purchase time of historically purchased props can reflect the impact of the user's recent purchase of props on the recommended props, and similar props can reflect the impact of the user's interests on the recommended props, this method takes into account the impact of the purchase time of the above-mentioned historically purchased props and similar props on the recommended props, and can accurately recommend props to users with a high accuracy rate.

[0170] Fig.11 1 is a schematic diagram of the structure of a computer device according to an exemplary embodiment. The computer device 1100 includes a central processing unit (CPU) 1101, a system memory 1104 including a random access memory (RAM) 1102 and a read-only memory (ROM) 1103, and a system bus 1105 connecting the system memory 1104 and the central processing unit 1101. The computer device 1100 also includes a basic input / output system (I / O system) 1106 for helping to transmit information between various devices in the computer device, and a large-capacity storage device 1107 for storing an operating system 1113, application programs 1114 and other program modules 1115.

[0171] The basic input / output system 1106 includes a display 1108 for displaying information and an input device 1109 such as a mouse and a keyboard for user inputting information. The display 1108 and the input device 1109 are connected to the central processing unit 1101 through an input / output controller 1110 connected to the system bus 1105. The basic input / output system 1106 may also include an input / output controller 1110 for receiving and processing inputs from a plurality of other devices such as a keyboard, a mouse, or an electronic stylus. Similarly, the input / output controller 1110 also provides output to a display screen, a printer, or other types of output devices.

[0172] The mass storage device 1107 is connected to the central processing unit 1101 through a mass storage controller (not shown) connected to the system bus 1105. The mass storage device 1107 and its associated computer device readable medium provide non-volatile storage for the computer device 1100. That is, the mass storage device 1107 may include a computer device readable medium (not shown) such as a hard disk or a Compact Disc Read-Only Memory (CD-ROM) drive.

[0173] Without loss of generality, the computer device readable medium may include computer device storage media and communication media. Computer device storage media include volatile and non-volatile, removable and non-removable media implemented by any method or technology for storing information such as computer device readable instructions, data structures, program modules or other data. Computer device storage media include RAM, ROM, Erasable Programmable Read Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), CD-ROM, Digital Video Disc (DVD) or other optical storage, cassette, tape, disk storage or other magnetic storage devices. Of course, those skilled in the art will know that the computer device storage medium is not limited to the above. The above-mentioned system memory 1104 and mass storage device 1107 can be collectively referred to as memory.

[0174] According to various embodiments of the present disclosure, the computer device 1100 can also be connected to a remote computer device on a network through a network such as the Internet. That is, the computer device 1100 can be connected to the network 1111 through the network interface unit 1112 connected to the system bus 1105, or the network interface unit 1112 can be used to connect to other types of networks or remote computer device systems (not shown).

[0175] The memory also includes one or more programs, which are stored in the memory. The central processing unit 1101 implements all or part of the steps of the above-mentioned prop recommendation method by executing the one or more programs.

[0176] In an exemplary embodiment, a computer-readable storage medium is also provided, in which at least one instruction, at least one program, a code set or an instruction set is stored, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by a processor to implement the prop recommendation method provided by the above-mentioned various method embodiments.

[0177] The present application also provides a computer-readable storage medium, in which at least one instruction, at least one program, a code set or an instruction set is stored. The at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the prop recommendation method provided in the above method embodiment.

[0178] The present application also provides a computer program product or a computer program, the computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the prop recommendation method provided in the above embodiment.

[0179] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0180] A person skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware or by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.

[0181] The above description is only an optional embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A prop recommendation method, characterized in that: The method comprises: Count the similar items of each historically purchased item in the historical purchase records of the user account to obtain a candidate recommendation set; Determine the time interval between the purchase time of the i-th historically purchased item and the purchase time of the first-purchased item, wherein the first-purchased item refers to the earliest purchased item in the historical purchase record of the user account; determine the time span from the current time to the time when the recommended item was last obtained; obtain the time weight of the i-th historically purchased item according to the ratio of the time interval and the time span, wherein the time weight is negatively correlated with the difference between the purchase time of the i-th historically purchased item and the current time, wherein i is a positive integer, and the initial value of i is 1; Extracting a subset of historical purchase records from the historical purchase records according to the time window; calculating the overall similarity between the i-th historical purchase item and the subset of historical purchase records; obtaining a similarity weight of the i-th historical purchase item according to a ratio between the overall similarity and the number of items in the subset of historical purchase records, wherein the similarity weight is used to indicate a degree of relevance between the i-th historical purchase item and the user profile of the user account; Weighted calculation of the time weight of the i-th historically purchased item and the similarity weight of the i-th historically purchased item to obtain a recommendation weight of the i-th historically purchased item; Update i to i+1, and repeat the above three steps until the recommendation weights of the various historically purchased items are obtained; Determining a weighted recommendation degree of each candidate recommended item according to the recommendation weight of each historically purchased item and each candidate recommended item in the candidate recommendation set; The recommended items to be recommended to the user account are determined according to the ranking of the weighted recommendation degrees of the candidate recommended items.

2. The method according to claim 1, characterized in that The calculating the overall similarity between the i-th historically purchased item and the historical purchase record subset includes: Calculating the similarity between the i-th historical purchase item and each historical purchase item in the historical purchase record subset; The sum of the similarities is calculated to obtain the overall similarity.

3. The method according to claim 1 or 2, characterized in that: The step of determining the weighted recommendation degree of each candidate recommended item according to the recommendation weight of each historically purchased item and each candidate recommended item in the candidate recommendation set includes: Calculate the similarity between the pth historically purchased item and the qth candidate recommended item, where p and q are positive integers and their initial values ​​are 1; Calculate the product of the recommendation weight of the p-th historically purchased item and the similarity to obtain the weighted similarity of the p-th historically purchased item; Update p to p+1, and repeat the above two steps k times until k weighted similarities of the qth candidate recommended item are obtained, where k represents the number of historically purchased items; Calculate the sum of k weighted similarities of the qth candidate recommended item to obtain a weighted recommendation degree of the qth candidate recommended item; Update q to q+1, initialize p, and repeat the above four steps until the weighted recommendation degree of each candidate recommended item is obtained.

4. The method according to claim 3, characterized in that The calculating of the similarity between the pth historically purchased item and the qth candidate recommended item includes: The cosine similarity between the p-th historically purchased item and the q-th candidate recommended item is calculated.

5. The method according to claim 1 or 2, characterized in that: The counting of similar items of each historically purchased item in the historical purchase record to obtain a candidate recommendation set includes: Determine similar items to each of the historically purchased items in the historical purchase records; The candidate recommendation set is obtained by deleting duplicate props from similar props of each of the historically purchased props, wherein the duplicate props refer to props whose number of props is greater than 1 from similar props of each of the historically purchased props.

6. The method according to claim 1 or 2, characterized in that: The step of determining the recommended props according to the ranking of the weighted recommendation degrees of the candidate recommended props includes: Sorting the candidate recommended items in descending order according to the weighted recommendation degree; The top N candidate recommended items are determined as the recommended items, where N is a positive integer.

7. The method according to claim 1 or 2, characterized in that: The method further comprises: Call the item recommendation model, perform data processing on the historical purchase records and the candidate recommendation set, and output the weighted recommendation degree of each candidate recommended item. The item recommendation model includes a time weight calculation network and a similarity weight calculation network. The time weight calculation network is used to calculate the time weight of each historically purchased item, and the similarity weight calculation network is used to calculate the similarity weight of each historically purchased item.

8. The method according to claim 7, characterized in that The item recommendation model is trained based on the historical purchase records of the user account and the candidate recommendation set; The method further comprises: determining similar user accounts to the user account; The item recommendation model is trained again according to the recommended item purchase status of the similar user accounts.

9. A prop recommendation device, characterized in that: The device comprises: A statistical module is used to count similar items of each historically purchased item in the historical purchase records of the user account to obtain a candidate recommendation set; A calculation module, used to determine the time interval between the purchase time of the i-th historical purchase item and the purchase time of the first purchase item, wherein the first purchase item refers to the earliest purchase item in the historical purchase record of the user account; determine the time span from the current time to the time when the recommended item was last obtained; obtain the time weight of the i-th historical purchase item according to the ratio of the time interval and the time span, wherein the time weight is negatively correlated with the difference between the purchase time of the i-th historical purchase item and the current time, wherein i is a positive integer, and the initial value of i is 1; Extracting a subset of historical purchase records from the historical purchase records according to the time window; calculating the overall similarity between the i-th historical purchase item and the subset of historical purchase records; obtaining a similarity weight of the i-th historical purchase item according to a ratio between the overall similarity and the number of items in the subset of historical purchase records, wherein the similarity weight is used to indicate a degree of relevance between the i-th historical purchase item and the user profile of the user account; Weighted calculation of the time weight of the i-th historically purchased item and the similarity weight of the i-th historically purchased item to obtain a recommendation weight of the i-th historically purchased item; Update i to i+1, and repeat the above three steps until the recommendation weights of the various historically purchased items are obtained; The calculation module is further used to determine the weighted recommendation degree of each candidate recommended item according to the recommendation weight of each historically purchased item and each candidate recommended item in the candidate recommendation set; The calculation module is further used to determine the recommended props to be recommended to the user account according to the ranking of the weighted recommendation degrees of the candidate recommended props.

10. A computer device, characterized in that: The computer device includes: a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the prop recommendation method as described in any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that: At least one program code is stored in the computer-readable storage medium, and the program code is loaded and executed by the processor to implement the prop recommendation method according to any one of claims 1 to 8.

12. A computer program product comprising a computer program or instructions, characterized in that When the computer program or instruction is executed by a processor, the prop recommendation method according to any one of claims 1 to 8 is implemented.