Intelligent product recommendation method and device based on graph thought and electronic equipment

Abstract

The invention discloses an intelligent product recommendation method and device based on a graph thought and electronic equipment, and the method comprises the steps: obtaining a directed weighted graph corresponding to a to-be-recommended target user, the directed weighted graph representing a graph constructed with the user as a starting point, an interactive product as an end point and user features as nodes; based on the weight parameter of each node in the directed weighted graph, the importance parameter of each node is determined, and the weight parameter representation of each node is determined based on the relationship between the starting point and the node and the relationship between the node and the end point in the directed weighted graph; determining target user features based on the importance parameter of each node; and determining a to-be-recommended target interactive product corresponding to the target user based on the target user feature and the product feature of each interactive product. According to the method, the user features are screened based on the directed weighted graph, and the screened user features are used as supplementary features of the interactive product, so that the problem of inaccurate recommendation due to few product features can be solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a smart product recommendation method and device based on graph idea and electronic equipment. BACKGROUND

[0002] With the development of network technology, the amount of network information is also increasing. Excessive information makes people unable to efficiently obtain the part they need, and the efficiency of information use is actually reduced.

[0003] Taking e-commerce as an example, e-commerce is a new business operation mode based on browser / server application mode, which realizes online shopping of consumers, online transaction between merchants and online electronic payment. With the development of supply chain and logistics, the types and quantities of goods provided by merchants on the network are very large, which greatly increases the time cost of consumer shopping screening and reduces the purchase efficiency of the e-commerce platform.

[0004] There are technical solutions for recommending related products to users in the prior art, but most of the solutions have problems such as inaccurate feature processing and low feature processing efficiency in the process of processing user features or product features, and cannot solve the problem of recommending accurate products to users. SUMMARY

[0005] In view of the above problems, the present application provides the following technical solutions.

[0006] A smart product recommendation method based on graph idea, comprising: obtaining a directed weighted graph corresponding to a target user to be recommended, the directed weighted graph representing a graph constructed with a user as a starting point, an interactive product as an end point, and a user feature as a node; determining an importance parameter of each node based on a weight parameter of each node in the directed weighted graph; wherein the weight parameter includes a first path weight from the starting point to the node and a second path weight from the node to the end point; the first weight represents the number of times the user uses the user feature within a target time period; the second path weight is determined based on the relationship between the user feature and the interactive product; determining a target user feature based on the importance parameter of each node; determining a target interactive product to be recommended corresponding to the target user based on the target user feature and product features of each interactive product.

[0007] Optionally, the obtaining of the directed weighted graph corresponding to the target user to be recommended comprises: determining the directed weighted graph corresponding to the target user in a directed weighted graph database; wherein the creation process of the directed weighted graph database comprises: obtaining behavior data corresponding to each user; processing each field data type in the behavior data to obtain user features of different dimensions; constructing a directed weighted graph corresponding to each user based on the relationship between the user features of each dimension and the corresponding user, and the relationship between the user features of each dimension and the corresponding interactive product, wherein the user is the starting point, the interactive product is the terminal point, and the user feature is the node in the directed weighted graph; and combining the directed weighted graphs corresponding to each user to obtain the directed weighted graph database.

[0008] Optionally, the determining of the importance parameter of each node based on the weight parameter of each node in the directed weighted graph comprises: calculating the in-degree weight sum of the node according to the first path weight of each node; and determining the ratio between the second path weight of each node and the in-degree weight sum as the importance parameter of each node.

[0009] Optionally, the determining of the target user feature based on the importance parameter of each node comprises: determining the user feature corresponding to the node whose importance parameter meets the screening condition as the target user feature; wherein the screening condition at least comprises the condition of screening based on the sorting size of the importance parameter, and the condition of screening based on the second path weight parameter if the importance parameters are the same.

[0010] Optionally, the determining of the target interactive product to be recommended corresponding to the target user based on the target user feature and the product feature of each interactive product comprises: determining the target feature of each interactive product based on the target user feature and the product feature corresponding to each interactive product; creating a target model based on the target feature of each interactive product and the user feature of the user; and processing the user feature corresponding to the target user by using the target model to obtain the target interactive product to be recommended.

[0011] A smart product recommendation device based on graph thinking includes: an acquisition unit for obtaining a directed weighted graph corresponding to a target user to be recommended, wherein the directed weighted graph represents a graph constructed with the user as the starting point, the interactive product as the end point, and the user features as nodes. A first determination unit is used to determine the importance parameter of each node based on the weight parameter of each node in the directed weighted graph; wherein the weight parameters include a first path weight from the starting point to the node and a second path weight from the node to the end point; the first weight represents the number of times the user utilizes the user feature within a target time period; the second path weight represents the relationship between the user feature and the interactive product. A second determination unit is used to determine the target user feature based on the importance parameter of each node.

[0012] The third determining unit is configured to determine a target interactive product to be recommended corresponding to the target user based on the target user characteristics and product characteristics of each of the interactive products.

[0013] Optionally, the acquisition unit is configured to: determine the directed weighted graph corresponding to the target user in the database of directed weighted graphs. The device further includes a creation unit, the creation unit is used to create the database of the directed weighted graph, and the creation unit includes: a first acquisition subunit, used to obtain the behavior data corresponding to each user. A first processing subunit, used to process the data type of each field in the behavior data to obtain user features of different dimensions. A construction subunit, based on the relationship between the user features of each dimension and the corresponding user, and the relationship between the user features of each dimension and the corresponding interactive products, constructs a directed weighted graph corresponding to each user, wherein the user is the starting point, the interactive product is the end point, and the user feature is the node in the directed weighted graph. A combination subunit, used to combine the directed weighted graphs corresponding to each user to obtain a directed weighted graph database.

[0014] Optionally, the first determination unit includes: a first calculation subunit, configured to calculate the sum of the in-degree weights of each node based on the first path weight of each node. The first determination subunit is configured to determine a ratio between the second path weight of each node and the sum of the in-degree weights as the importance parameter of each node.

[0015] Optionally, the second determining unit includes: a second determining subunit, configured to determine, as the target user feature, user features corresponding to nodes whose importance parameters satisfy a screening condition. The screening condition includes at least screening based on a ranking of importance parameters, and, if the importance parameters are the same, screening based on a second path weight parameter.

[0016] An electronic device includes a memory for storing a program, and a processor for executing the program, the program being specifically configured to implement the intelligent product recommendation method based on a graph idea as described in any one of the above.

[0017] Compared with the prior art, the present application provides an intelligent product recommendation method, device and electronic equipment based on a graph idea, a directed weighted graph corresponding to a target user to be recommended is obtained, the directed weighted graph representing a graph constructed with the user as the starting point, the interactive product as the end point and the user features as the nodes; the importance parameter of each node is determined based on the weight parameter of each node in the directed weighted graph, the weight parameter of each node representing the relationship between the starting point and the node and the relationship between the node and the end point in the directed weighted graph; the target user features are determined based on the importance parameter of each node; and the target interactive product to be recommended corresponding to the target user is determined based on the target user features and the product features of each interactive product. The present application filters the user features based on the directed weighted graph, and uses the filtered user features as the supplementary features of the interactive product, which can solve the problem of inaccurate recommendation due to few product features. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of the provided drawings.

[0019] Figure 1 A flowchart of an intelligent product recommendation method based on a graph idea provided by an embodiment of the present application;

[0020] Figure 2 A construction flowchart of a product recommendation model provided by an embodiment of the present application;

[0021] Figure 3 A schematic diagram of a directed weighted graph provided by an embodiment of the present application;

[0022] Figure 4 A structural schematic diagram of an intelligent product recommendation device based on a graph idea provided by an embodiment of the present application. DETAILED DESCRIPTION

[0023] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0024] The terms "first" and "second" in this application are used to distinguish between different objects, not to describe a specific order. Furthermore, the terms "including" and "having," and any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of steps or elements is not limited to the listed steps or elements and may include steps or elements that are not listed.

[0025] In an embodiment of this application, a graph-based intelligent product recommendation method is provided. This method can be applied to business systems, enabling the business system to recommend relevant business products based on this method, where business products include products corresponding to e-commerce. This application can quickly and accurately recommend relevant products to users based on the characteristics of the current interactive product to be recommended and the user characteristics of the recommended user, thereby meeting actual business needs.

[0026] See also Figure 1 , which is a flow chart of a smart product recommendation method based on graph thinking provided in an embodiment of the present application. The method may include the following steps S101 to S103.

[0027] S101: Obtain a directed weighted graph corresponding to a target user to be recommended.

[0028] A directed weighted graph represents a graph constructed with users as the starting point, interactive products as the end point, and user features as nodes. The user is a recommended user in the current business system, such as a new user in an online shopping system, or a user who has generated a product acquisition need, such as a current user who has entered a relevant search term and needs to obtain a corresponding product. In order to efficiently and quickly determine the corresponding recommended product for the user, in an embodiment of the present application, the relationship between users, user features, and interactive products can be represented as a directed weighted graph, and a directed weighted graph database can be pre-generated, storing directed weighted graphs corresponding to different types of users, or corresponding to different interactive products, or corresponding to different application needs. In this way, when a relevant directed weighted graph is needed, the relevant directed weighted graph can be queried in the database. In another embodiment, when a directed weighted graph is needed, the directed weighted graph can be constructed based on the user features of the current user to be recommended and the products that the user has interacted with. The products that the user has interacted with can be products purchased and used by the user based on the business system, or products that the user is interested in, such as products that the user frequently visits; or products that the user has historically paid attention to, etc. This application does not limit this.

[0029] S102: Determine an importance parameter of each node based on a weight parameter of each node in the directed weighted graph.

[0030] The directed weighted graph is composed of a vertex set and an edge set, wherein each edge has a direction and a weight. Specifically, the vertex set includes a start point, a node and an end point, which correspond to the business system in the embodiments of the present application, the start point represents a user, the end point represents an interactive product, and the node is a user feature. The direction of the edge represents the direction of the above-mentioned points, for example, the edge between the start point and the node represents the user feature corresponding to the user, and the weight of the edge represents the importance of the user feature corresponding to the current node. For example, a first user corresponds to a first user feature, a second user feature and a third user feature, and the weight parameter of the first user feature is greater than the weight parameter of the second user feature, which indicates that the first user feature is more important to the first user. Specifically, the weight value corresponding to the weight parameter of the general features such as the gender and age of the user will be smaller, and the weight value corresponding to the weight parameter of the user's interest feature will be relatively larger. Thus, the importance parameter of each node can be determined according to the weight parameter of each node, so as to determine the subsequent user interest feature and provide more accurate information for the user.

[0031] S103, determining a target user feature based on the importance parameter of each node.

[0032] In the directed weighted graph, each node represents a user feature of a user. The most relevant user feature of the current user is selected as the target user feature according to the importance parameter of the user feature. These target user features can reflect the features that the user is interested in and pays attention to, and thus can be applied to the subsequent product recommendation process.

[0033] S104, determining a target interactive product to be recommended corresponding to the target user based on the target user feature and the product features of the interactive products.

[0034] The product features of each interactive product are obtained, and the target user features are used as the supplementary features of the interactive products, so as to make up for the problem of insufficient product features. Then, by matching the user features of the target user with each feature of the interactive products, the interactive product most related to the target user is obtained as the target interactive product, so as to recommend the target interactive product to the target user. The product features corresponding to each interactive product can be extracted from the product database to which the interactive product belongs. These product features can include the basic attribute features of the interactive product, such as product name, type, price, category to which the product belongs, target user group, contract period of the interactive product, and discount information. These features are usually stored in a structured form in the product database, and the corresponding interactive product features can be obtained through a database query interface or an application program interface. The application log of the interactive product can also be retrieved and analyzed to obtain the product features of the interactive product. For example, the application log of the interactive product can be used to obtain statistical features in the product features of the interactive product, such as exposure data, click data, number of subscribers, and comment data for a certain interactive product.

[0035] The embodiment of the present application provides an intelligent product recommendation method based on a graph idea. A directed weighted graph corresponding to a target user to be recommended is obtained. The directed weighted graph represents a graph constructed with the user as the starting point, the interactive product as the ending point, and the user features as the nodes. The importance parameter of each node is determined based on the weight parameter of each node in the directed weighted graph. The weight parameter of each node is determined based on the relationship between the starting point and the node in the directed weighted graph, and the relationship between the node and the ending point. The target user features are determined based on the importance parameter of each node. The target interactive product to be recommended corresponding to the target user is determined based on the target user features and the product features of each interactive product. The user features are screened based on the directed weighted graph, and the screened user features are used as the supplementary features of the interactive products, so as to solve the problem of inaccurate recommendation due to insufficient product features.

[0036] The technical features of the embodiments of the present application will be described below in combination with specific application scenarios.

[0037] In an embodiment of the embodiment of the present application, obtaining the directed weighted graph corresponding to the target user to be recommended includes: determining the directed weighted graph corresponding to the target user in the database of the directed weighted graph.

[0038] The database of the directed and weighted graph includes: obtaining behavior data corresponding to each user; processing field data types in the behavior data to obtain user features of different dimensions; constructing a directed and weighted graph corresponding to each user based on a relationship between the user features of each dimension and the corresponding user and a relationship between the user features of each dimension and the corresponding interactive product, wherein the user is the starting point, the interactive product is the terminal point, and the user feature is the node in the directed and weighted graph; and combining the directed and weighted graphs corresponding to each user to obtain the database of the directed and weighted graph.

[0039] Specifically, the user behavior data can be interactive behavior data of the user in the corresponding business system and the related product, such as access data, browsing data, search data of the user to the product, and can also include confirmation data, purchase data and the like of the user to the related product. The behavior data includes a plurality of field data, each field data can represent key information of the corresponding data, such as user behavior data analysis can determine the access time field, the access content field, the interactive feature field data and the like of the user. Different field data types can obtain user features of different dimensions, such as user basic attribute dimension, user access dimension, user purchase dimension and the like. Then, according to the relationship between the user features of each dimension and the corresponding user and the relationship with the interactive product, the directed and weighted graph corresponding to the user is determined, so as to combine the directed and weighted graphs of each user to obtain the database of the directed and weighted graph.

[0040] For example, taking the interactive product as a user network data package as an example. The user features can include features of user product demand, such as a basic package handled by the user when entering the network, a quasi-real-time updated usage amount and saturation of the user flow, voice and the like package, and application program (APP) flow data; features describing user right preference, such as the number of times of subscribing to different categories of rights by the user in the last half year, the number of times of subscribing to different brands of rights by the user in the last half year and the like; basic attributes of the user, such as age, gender, city, entering time and the like; a family portrait of the user, including family members, whether there is a child, family DOU (Dataflow of Usage, representing average monthly internet flow per household), family MOU (Minutes of Usage, representing average monthly call time per household) and the like, it should be noted that the above user data and the corresponding user features are obtained based on user-authorized public data collection. Still taking this scenario as an example, the product features of the corresponding interactive product can include basic product price, basic product containing flow package right information, activity discount amount, activity contract period and the like product attributes; statistical features such as the number of people in the last 7 days and the number of people subscribing in the last 7 days.

[0041] In an embodiment of the present application, the directed weighted graph corresponding to each user includes: a first path weight from the starting point to the node and a second path weight from the node to the end point; wherein the first path weight represents the number of times the user utilizes the user feature within the target time period; the second path weight represents the relationship between the user feature and the interactive product. Specifically, the values ​​of each field of the user feature are extracted, with the user as the starting point, the user feature field value as the node, and the interactive product as the end point. There is no path connecting the nodes. The first path weight from the starting point (user) to the node (user feature) is determined by the number of times the user reaches the feature value within a certain time period; for example, the second weight parameter from the node (user feature) to the end point (interactive product) may include: the path weight for the user who has processed the product under the feature is 2, and the path weight for the user who has been exposed to but not processed is -1.

[0042] Correspondingly, based on the weight parameter of each node in the directed weighted graph, the importance parameter of each node is determined, including: according to the first path weight of each node, the sum of the in-degree weights of the node is calculated. For each node in the directed weighted graph, that is, the node representing the user feature, the sum of its in-degree weights refers to the sum of the weights of all directed edges pointing to the node. The directed edge represents the path from the starting point to the node. In the embodiment of the present application, the user is the starting point and the user feature is the node in the directed weighted graph, that is, the directed edge represents the path from the user to the corresponding user feature. The first path weight represents the number of times the user uses the user feature within the target time period. For example, when the first path weight is 2, it represents that the user has used the user feature it points to twice within the target time period. For example, if the user feature is the user occupation type feature, the user will screen the first interactive product and the second interactive product according to their user occupation type within a week, that is, it represents that the user has used their occupational saliency feature twice within a week. For example, the user feature node V1 is pointed to by users U1, U2, and U3 with weights (i.e., first path weights) w1=2, w2=1, and w3=3, respectively. Then the sum of the in-degree weights S of the node V1 is: S=w1+w2+w3=2+1+3=6.

[0043] The ratio between the second path weight and the sum of the in-degree weights of each node is determined as the importance parameter of each node.

[0044] Furthermore, the target user characteristics are determined based on the importance parameters of each node, including: determining the user characteristics corresponding to the nodes whose importance parameters of each node meet the filtering conditions as the target user characteristics; wherein the filtering conditions at least include filtering based on the ranking size of the importance parameters, and if the importance parameters are the same, filtering based on the second path weight parameter.

[0045] This allows the user features most relevant to the user to be screened out as supplementary features to the product features, thereby increasing the product's ability to differentiate, improving the vector similarity between product features and user features, and solving the problem of operators having few quantifiable features and sparse features in product features.

[0046] In the embodiment of the present application, when determining target interactive products to be recommended, product recommendations can be made based on a large model. Accordingly, determining target interactive products to be recommended for the target user based on the target user characteristics and the product characteristics of each interactive product includes: determining target characteristics for each interactive product based on the target user characteristics and product characteristics corresponding to each interactive product; creating a target model based on the target characteristics of each interactive product and the user characteristics of the user; and using the target model to process the user characteristics corresponding to the target user to obtain the target interactive product to be recommended.

[0047] After using target user characteristics as supplementary features to obtain target features for interactive products, they can be used in the model training phase. This allows the model to capture relevant information between product characteristics and users, facilitating accurate application of the trained target model in interactive product recommendation scenarios. User characteristics represent basic characteristics of the user, such as age, interests, occupation, and interactions with related products, among other publicly available attributes.

[0048] The following uses the product recommendation scenario as an example to illustrate the embodiment of this application. Figure 2 , which is a schematic diagram of the construction process of a product recommendation model provided in an embodiment of the present application, and the product recommendation model is used to recommend relevant products to target users. Specifically, the process includes the following steps: S201, data processing of user features. S202, constructing a directed weighted graph with the user as the starting point, the product processing as the end point, and the user features as nodes. S203, filtering user features according to the path weight. S204, using the filtered user features as supplementary features of the product features, conducting model construction and training, and obtaining the target model.

[0049] Step S201 primarily involves processing the data types of each field in the user behavior data. For example, continuous fields are binned to ensure that all data is not concentrated in a single bin; discrete fields are processed using one-hot or multi-hot processing. The ordering relationship between users and products is then matched.

[0050] In step S202, the user feature field values are extracted, the user is taken as the starting point, the user feature field values are taken as nodes, and the interactive product is taken as the end point to construct a directed and weighted graph. There is no path connection between the nodes. The path weight from the starting point (user) to the node (user feature) is determined by the number of times the user reaches the feature value within a certain time period. The path weight from the node (user feature) to the end point (interactive product) is 2 if the user has handled the product under the feature, and -1 if the user has contacted but not handled the product.

[0051] Referring to Table 1, the related information of the user feature and the contact and handling product relationship is shown.

[0052] Table 1

[0053]

[0054] Correspondingly, the directed and weighted graph is shown in Figure 3 In Figure 3 , u1, u2, u3, and u4 represent user 1, user 2, user 3, and user 4, respectively, v11, v12, v21, v22, and v23 represent the corresponding user features, and p1 and p2 represent the corresponding handling products. The path weight (such as w1-11) from the starting point (user) to the node (user feature) is determined by the number of times the user reaches the feature value within a certain time period. For example, user u1 has feature value v11 at time t1 and t2, and the path weight w1-11 = 2 (because it appears twice).

[0055] The path weight (such as w11-1) from the node (user feature) to the end point (interactive product) is 1 if the user has handled the product under the feature, and -1 if the user has contacted but not handled the product. It should be noted that Figure 3 the related examples and weight calculation process in Figure 3 are only illustrative. The specific calculation can be performed according to the actual application scenario. For example, when calculating w11-1, the weight path of all user features v11 and product p1 interactions needs to be counted. That is, wXX-Y represents the sum of all path weights from user feature node XX to product Y. For example, w22-1 needs to count all user features v2=v22 and product p1 interaction records. For example, user u2 has handled product p1 twice based on user feature v22, and w22-1=2, Figure 3 The other calculation formulas in

[0056] The importance score of each field value of the user feature is calculated, the importance calculation method is to take the node (user feature) in-degree weight sum as the denominator, and the node (user feature) and the path weight of the terminal point (interactive product) as the numerator. The field value with the maximum importance score is taken, if the importance scores are the same, the node (user feature) with the maximum weight to the terminal point (interactive product) is taken as the product supplementary feature.

[0057] The product supplementary feature is spliced to the tail of the existing product feature. When the directed and weighted graph is as shown in Figure 3 , the p1 supplementary feature is v11 and v21, and the p2 supplementary feature is v12 and v22. Finally, the product feature is subjected to PAC dimension reduction processing. Specifically, the importance score calculation method of v11 in Figure 3 is (w1-11+w2-11+w3-11+w4-11) / (w11-1+w11-2), the path weight calculation method of w1-11 can be obtained according to the user feature and the contact handling relationship table. At t1 time, the user u1 handles the product p1, and at that time, the feature value of the v1 feature is v11. At t2 time, the user u1 handles the product p1 again, and at that time, the feature value of the v1 feature is v11, so the path w1-11 weight is 2, and the path features such as w2-11 and w11-1 are calculated in the same way. The field value with the maximum importance score is taken, if the importance scores are the same, the node (user feature) with the maximum weight to the terminal point (interactive product) is taken as the product supplementary feature.

[0058] Subsequently, a training set, feature processing, model training, and model saving can be generated, and the inference stage process is to load the model, obtain the user recall list generated in the recall stage, read the user feature, read the product feature, perform feature processing, and perform forward operation of the model, so as to obtain the target interactive product to be recommended corresponding to the target user.

[0059] In the embodiment of the present application, based on the idea of directed and weighted graph, a directed and weighted graph is constructed with the user as the starting point, the user feature as the node, and the interactive product as the terminal point. The degree operation of the graph is used to calculate the weight of the feature node, which is efficient and has low complexity. The weight of the path in the directed and weighted graph is used to screen the user feature and supplement it as the product feature, effectively alleviating the problems of few product features and sparse product features, thereby improving the accuracy of product recommendation.

[0060] In another embodiment of the present application, an intelligent product recommendation device based on the graph idea is also provided, as shown in Figure 4 , comprising:

[0061] The acquisition unit 401 is configured to obtain a directed and weighted graph corresponding to a target user to be recommended, wherein the directed and weighted graph represents a graph constructed with the user as the starting point, the interactive product as the terminal point, and the user feature as the node.

[0062] The first determination unit 402 is used to determine the importance parameter of each node based on the weight parameter of each node in the directed weighted graph, where the weight parameter representation of each node is determined based on the relationship between the starting point and the node, and the relationship between the node and the end point in the directed weighted graph.

[0063] The second determining unit 403 is configured to determine target user characteristics based on the importance parameter of each node.

[0064] The third determining unit 404 is configured to determine a target interactive product to be recommended corresponding to the target user based on the target user characteristics and product characteristics of each of the interactive products.

[0065] Optionally, the acquisition unit is configured to: determine a directed weighted graph corresponding to the target user in a database of directed weighted graphs.

[0066] Wherein, the device also includes a creation unit, and the creation unit is used to create a database of the directed weighted graph, and the creation unit includes: a first acquisition subunit, used to obtain the behavior data corresponding to each user. A first processing subunit, used to process the data type of each field in the behavior data to obtain user features of different dimensions. A construction subunit, based on the relationship between the user features of each dimension and the corresponding user, and the relationship between the user features of each dimension and the corresponding interactive products, constructs a directed weighted graph corresponding to each user, wherein the user is the starting point, the interactive product is the end point, and the user feature is the node in the directed weighted graph. A combination subunit, used to combine the directed weighted graphs corresponding to each user to obtain a directed weighted graph database.

[0067] Optionally, the directed weighted graph corresponding to each user includes a first path weight from a starting point to a node and a second path weight from a node to an end point. The first path weight represents the number of times the user utilizes the user feature within a target time period, and the second path weight represents a relationship between the user feature and the interactive product.

[0068] Optionally, the first determination unit includes: a first calculation subunit, configured to calculate the sum of the in-degree weights of each node based on the first path weight of each node. The first determination subunit is configured to determine a ratio between the second path weight of each node and the sum of the in-degree weights as the importance parameter of each node.

[0069] Optionally, the second determining unit comprises a second determining sub-unit, configured to determine, as the target user feature, the user feature corresponding to the node whose importance parameter meets a screening condition. The screening condition comprises at least screening based on the ordering size of the importance parameter, and screening based on the second path weight parameter if the importance parameters are the same.

[0070] Optionally, the third determining unit comprises a third determining sub-unit, configured to determine, as the target feature of each interactive product, the target user feature and the product feature corresponding to each interactive product. A creating sub-unit is configured to create a target model based on the target feature of each interactive product and the user feature of the user. A second processing sub-unit is configured to process the user feature corresponding to the target user by using the target model, to obtain the target interactive product to be recommended.

[0071] It should be noted that the processing procedures of the above-mentioned units and sub-units can refer to the related procedures of the above-mentioned intelligent product recommendation method based on the graph idea, which will not be described in detail here.

[0072] Based on the above-mentioned embodiments, the embodiments of the present application provide a computer readable storage medium, which stores one or more programs, and the one or more programs can be executed by one or more processors to implement the above-mentioned intelligent product recommendation method based on the graph idea.

[0073] The embodiments of the present application also provide an electronic device, comprising a memory configured to store a program, and a processor configured to execute the program, and the program is specifically configured to implement the above-mentioned intelligent product recommendation method based on the graph idea.

[0074] It should be noted that the above-mentioned processor or CPU can be at least one of an application specific integrated circuit (ASIC), a digital signal processor (DSP), a digital signal processing device (DSPD), a programmable logic device (PLD), a field programmable gate array (FPGA), a central processing unit (CPU), a controller, a microcontroller, and a microprocessor. It can be understood that the electronic device for implementing the above-mentioned processor function can also be other electronic devices, and the embodiments of the present application are not limited specifically.

[0075] 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); it can also be various terminals including one or any combination of the above-mentioned memories, such as mobile phones, computers, tablet devices, personal digital assistants, etc.

[0076] 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.

[0077] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0078] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present application. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A graph-based intelligent product recommendation method, characterized in that: include: Obtaining a directed weighted graph corresponding to the target user to be recommended, wherein the directed weighted graph represents a graph constructed with the user as the starting point, the interactive product as the end point, and the user characteristics as the nodes; Determining an importance parameter of each node based on a weight parameter of each node in the directed weighted graph; wherein the weight parameters include a first path weight from the starting point to the node and a second path weight from the node to the end point; the first weight represents the number of times the user utilizes the user feature within a target time period; and the second path weight represents a relationship between the user feature and the interactive product. Determine target user characteristics based on the importance parameters of each node; Based on the target user characteristics and the product characteristics of each of the interactive products, a target interactive product to be recommended corresponding to the target user is determined.

2. The method according to claim 1, characterized in that The step of obtaining a directed weighted graph corresponding to the target user to be recommended includes: Determining a directed weighted graph corresponding to the target user in a database of directed weighted graphs; The process of creating the database of the directed weighted graph includes: Obtain behavioral data corresponding to each user; Processing the data types of each field in the behavioral data to obtain user features of different dimensions; Based on the relationship between the user features of each dimension and the corresponding user, and the relationship between the user features of each dimension and the corresponding interactive products, a directed weighted graph corresponding to each user is constructed, wherein the user is the starting point, the interactive product is the end point, and the user features are nodes in the directed weighted graph; The directed weighted graphs corresponding to each user are combined to obtain a directed weighted graph database.

3. The method according to claim 1, characterized in that The determining of the importance parameter of each node based on the weight parameter of each node in the directed weighted graph includes: Calculate the sum of the in-degree weights of the node according to the first path weight of each node; The ratio between the second path weight and the sum of the in-degree weights of each node is determined as the importance parameter of each node.

4. The method according to claim 3, characterized in that The determining of target user characteristics based on the importance parameter of each node includes: The user features corresponding to the nodes whose importance parameters of each node meet the screening conditions are determined as target user features; The screening conditions at least include screening based on the ranking of importance parameters, and if the importance parameters are the same, screening based on the second path weight parameter.

5. The method according to claim 1, wherein The determining of the target interactive product to be recommended corresponding to the target user based on the target user characteristics and the product characteristics of each of the interactive products includes: Determine the target features of each interactive product based on the target user features and product features corresponding to each interactive product; Create a target model based on the target characteristics of each interactive product and the user characteristics of the user; The target model is used to process the user features corresponding to the target user to obtain the target interactive product to be recommended.

6. A smart product recommendation device based on graph thinking, characterized in that: include: An acquisition unit is configured to obtain a directed weighted graph corresponding to a target user to be recommended, wherein the directed weighted graph represents a graph constructed with the user as a starting point, the interactive product as an end point, and the user characteristics as nodes; a first determining unit configured to determine an importance parameter of each node based on a weight parameter of each node in the directed weighted graph; wherein the weight parameters include a first path weight from the starting point to the node and a second path weight from the node to the end point; the first weight represents the number of times the user utilizes the user feature within a target time period; and the second path weight represents an importance parameter determined based on a relationship between the user feature and the interactive product; a second determining unit, configured to determine target user characteristics based on the importance parameter of each node; The third determining unit is configured to determine a target interactive product to be recommended corresponding to the target user based on the target user characteristics and product characteristics of each of the interactive products.

7. The device according to claim 6, characterized in that The acquisition unit is configured to: Determining a directed weighted graph corresponding to the target user in a database of directed weighted graphs; The device further includes a creation unit, which is used to create a database of the directed weighted graph, and the creation unit includes: A first acquisition subunit is used to obtain behavior data corresponding to each user; A first processing sub-unit is used to process the data type of each field in the behavior data to obtain user features of different dimensions; Constructing a sub-unit, based on the relationship between the user features of each dimension and the corresponding user, and the relationship between the user features of each dimension and the corresponding interactive products, constructing a directed weighted graph corresponding to each user, wherein the user is the starting point, the interactive product is the end point, and the user features are nodes in the directed weighted graph; The combining subunit is used to combine the directed weighted graphs corresponding to each user to obtain a directed weighted graph database.

8. The device according to claim 6, characterized in that The first determining unit includes: A first calculation subunit, configured to calculate a sum of the in-degree weights of the node according to the first path weight of each node; The first determining subunit is configured to determine a ratio between the second path weight and the sum of the in-degree weights of each node as the importance parameter of each node.

9. The device according to claim 6, characterized in that The second determining unit includes: The second determining subunit is configured to determine the user features corresponding to the nodes whose importance parameters satisfy the screening conditions as target user features; The screening conditions at least include screening based on the ranking of importance parameters, and if the importance parameters are the same, screening based on the second path weight parameter.

10. An electronic device, characterized in that: include: Memory, used to store programs; A processor is used to execute the program, wherein the program is specifically used to implement the intelligent product recommendation method based on graph concept as described in any one of claims 1 to 5.

Images