Data processing method, server and storage medium

Abstract

The embodiments of the present application provide a data processing method, a server, and a storage medium. For any network object, the network operation behavior data generated by the user performing network operations on the network object can be obtained, and the association information between the network object and other network objects that have a substitution relationship with the network object can be extracted from the network operation behavior data, and the network traffic generated by the network object within a certain period of time is predicted based on these association information. In the prediction process, the substitutability between network objects is taken into account, which is conducive to improving the accuracy of the prediction results, and the prediction results can be used as a basis for the provider to perform related operations such as network object scheduling and resource allocation, providing a data basis for the provider's subsequent operations, which is conducive to improving the user's stickiness to the network object and the provider, and reducing user churn.

Description

Technical Field

[0001] The present application relates to the field of Internet technology, and in particular to a data processing method, a server and a storage medium. Background Art

[0002] With the development of the e-commerce industry and the fast-paced lifestyle, the audience of online shopping is becoming more and more extensive. Users place orders through e-commerce platforms. After receiving the user's order, the e-commerce platform selects a nearby warehouse with goods and notifies the warehouse to arrange delivery for the user. In this way, users can buy the goods they need without leaving home, which greatly facilitates users.

[0003] From the perspective of merchants or e-commerce platforms, it is necessary to predict product sales, because accurate sales forecasts can help merchants replenish stocks in a timely manner, thereby reducing user churn due to out-of-stock or untimely delivery, improving user shopping experience, or helping merchants reduce inventory management costs caused by unsalable products due to inaccurate sales.

[0004] In the prior art, the future sales volume of a product is usually predicted based on the historical sales volume of the product. This method is relatively simple, but the accuracy of the prediction result is not high. Summary of the invention

[0005] Multiple aspects of the present application provide a data processing method, a server, and a storage medium to improve the accuracy of future sales forecasts for goods, thereby improving user stickiness of goods, reducing user churn, and reducing management costs for merchants.

[0006] An embodiment of the present application provides a data processing method, including: obtaining network operation behavior data generated by a user and associated with a first network object; extracting, from the network operation behavior data, an association relationship between a second network object that has a replaceable relationship with the first network object and the first network object; and predicting the network traffic generated by the first network object within a certain period of time based on the association relationship between the second network object and the first network object and attribute information of the first network object in at least one dimension.

[0007] An embodiment of the present application also provides a method for predicting commodity sales, including: obtaining network operation behavior data generated by a user and associated with a first commodity; extracting, from the network operation behavior data, an association relationship between a second commodity that has a substitutable relationship with the first commodity and the first commodity; predicting future sales of the first commodity based on the association relationship between the second commodity and the first commodity and attribute information of the first commodity in at least one dimension; and outputting the future sales to a provider of the target commodity so that the provider can perform operations related to the target commodity based on the future sales.

[0008] An embodiment of the present application also provides a warehouse scheduling method, including: obtaining network operation behavior data generated by a user and associated with a first commodity; extracting, from the network operation behavior data, an association relationship between a second commodity that has a substitutable relationship with the first commodity and the first commodity; predicting future sales of the first commodity based on the association relationship between the second commodity and the first commodity and attribute information of the first commodity in at least one dimension; and scheduling the first commodity among multiple warehouses in advance based on the future sales of the first commodity.

[0009] The embodiment of the present application also provides a warehouse resource configuration method, including: obtaining network operation behavior data generated by a user and associated with a first commodity; extracting the association relationship between multiple commodities that have a substitutable relationship with the first commodity and the first commodity from the network operation behavior data; predicting the future sales volume of the first commodity based on the association relationship between the multiple commodities and the first commodity and the attribute information of the first commodity in at least one dimension; and dynamically configuring the picking resources in the warehouse between the local commodity and other commodities based on the future sales volume of the first commodity.

[0010] An embodiment of the present application also provides a data processing server, comprising: a processor and a memory storing a computer program; the processor is used to execute the computer program to: obtain network operation behavior data generated by a user and associated with a first network object; extract, from the network operation behavior data, an association relationship between a second network object that has a replaceable relationship with the first network object and the first network object; and predict the network traffic generated by the first network object within a certain period of time based on the association relationship between the second network object and the first network object and attribute information of the first network object in at least one dimension.

[0011] The embodiment of the present application also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor implements the steps in the method of the embodiment of the present application.

[0012] In an embodiment of the present application, for any network object, the network operation behavior data generated by the user performing network operation behavior on the network object can be obtained, and the association information between the network object and other network objects that have a substitution relationship with the network object can be extracted from the network operation behavior data, and the network traffic generated by the network object within a certain period of time is predicted based on these association information. In the prediction process, the substitutability between network objects is taken into account, which is conducive to improving the accuracy of the prediction results, and the prediction results can be used as the basis for the provider to perform related operations such as network object scheduling and resource allocation, providing a data basis for the provider's subsequent operations, which is conducive to improving the user's stickiness to the network object and the provider, reducing user churn, and reducing the management costs of merchants. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0014] Figure 1 A schematic diagram of the structure of a data processing system provided in an embodiment of the present application;

[0015] Figure 2 A flowchart of a data processing method provided in an embodiment of the present application;

[0016] Figure 3a A schematic diagram of a data processing result provided in an embodiment of the present application;

[0017] Figure 3b A flowchart of a commodity sales forecasting method provided in an embodiment of the present application;

[0018] Figure 4 A schematic diagram of the structure of a data processing server provided in an embodiment of the present application. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0020] The embodiment of the present application provides a data processing method, which can combine the association relationship between network objects with substitutable relationships, predict the network traffic generated by network objects within a certain period of time, and provide the predicted network traffic generated by network objects within a certain period of time to the provider, which can be used as the basis for the provider to perform related operations such as network object scheduling and resource allocation, and provide a data basis for the subsequent operations of the provider, which is conducive to improving the user's stickiness to the network object and the provider, reducing user churn, and reducing the management cost of the merchant. Among them, the provider refers to a party that can provide network objects to users, for example, it can be a direct provider of network objects, such as a merchant, or it can be an intermediate provider, such as various e-commerce platforms. Depending on the application scenario, the network objects, the association relationship between network objects, the provider, and the subsequent operations performed by the provider will all be different, and please refer to the exemplary description in the following embodiment. The data processing method can be applied to the following data processing system, but is not limited to this.

[0021] Figure 1 The structure diagram of a data processing system using the data processing method of the embodiment of the present application is shown in FIG. Figure 1 As shown, the data processing system 100 includes: a data processing server 10 , a provider server 20 and a plurality of terminal devices 30 .

[0022] like Figure 1 As shown in ①, the provider server 20 can provide network objects to the terminal device 30. Among them, the provider server 20 can publish the network object to the corresponding client, APP or web page. The terminal device 30 provides the network object to the user by running the client, APP or web page. Optionally, the provider server 20 can be an e-commerce server in the e-commerce field, and accordingly, the network object can be a commodity or service in the e-commerce field. Alternatively, the provider server 20 can be an audio and video server in the audio and video field, and accordingly, the network object can be audio and video content such as movies, TV series or songs in the audio and video field. Alternatively, the provider server 20 can be a software development platform in the software field, and accordingly, the network object can be various application software provided by the software development platform.

[0023] Further, if Figure 1As shown in ②, the user can initiate various network operations on the network object through the terminal device 30. Depending on the network object, the network operations initiated by the user on the network object will also be different. Taking the network object as a commodity in the e-commerce field as an example, the user can browse the commodities, add the favorite commodities to the shopping cart, collect the favorite commodities, and place an order to purchase the commodities through the shopping APP or shopping website running on the terminal device 30. The user can also comment on the purchased commodities, etc., that is, the user's network operations can be operations such as browsing, adding to the shopping cart, collecting, placing orders, and commenting. Taking the network object as an audio and video content in the audio and video field as an example, the user can browse, play, cache or download the audio and video content through the audio and video playback APP or audio and video website running on the terminal device 30. It can also post bullet screens and comments on the audio and video content, that is, the user's network operations can be operations such as browsing, playing, caching or downloading, and posting bullet screens and comments. Taking the network object as an application software as an example, the user can browse, search and download various application software through the application store or download website running on the terminal device 30, that is, the user's network operations can be operations such as browsing, searching and downloading.

[0024] When the user performs a network operation, the terminal device 30 can respond to the user's network operation, cooperate with the provider server 20 to perform operations related to the user's network operation, and record the user's network operation behavior data in a log format. The user's network operation behavior data includes the time when the user initiates the network operation, identification information that identifies the network operation initiated by the user, and identification information of the network object involved in the network operation. Further, if Figure 1 As shown in ③, the terminal device 30 can also send the network operation behavior data generated by the user to the provider server 20. Figure 1 As shown in ④, the provider server 20 can send the network operation behavior data to the data processing server 10. Optionally, the terminal device 30 can also directly send the network operation behavior data to the data processing server 10. The network operation behavior data can be the network operation behavior data generated by the user in a certain historical time period, for example, the network operation behavior data generated by the user in the last month or three months, or the network operation behavior data generated by the user in the last week or two weeks, or the network operation behavior data generated by the user during one or several historical activities, or the network operation behavior data generated by the user in a specified time period, such as March 10-June 10. The selection of the historical time period can be flexibly determined according to the application requirements and is not limited to this.

[0025] The data processing server 10 may receive the network operation behavior data provided by the terminal device 30 or the provider server 20. Figure 1As shown in ⑤, the data processing server 10 can predict the network traffic generated by the relevant network objects within a certain time based on the association relationship between the network objects with substitution relationships contained in the network operation behavior data. The "certain time" here can be the current moment or a future time period, such as one hour from the current moment, one day from the current moment, 30 minutes from the current moment, or the next hour, the next day, the next week, the next month, or a time period between a certain start time in the future and a certain end time in the future, etc. If the network traffic generated by the network object at the current moment is predicted, it can be referred to as the current network traffic; if the network traffic generated by the network object in the future time period is predicted, it can be referred to as the future network traffic. Figure 1 In the example of predicting the future network traffic of network objects, we illustrate this. Figure 1 As shown in ⑥, the data processing server 10 can provide the predicted network traffic (such as future network traffic) generated by the network object within a certain period of time to the provider server 20, so that the provider server 20 can perform other operations based on the network traffic (such as future network traffic) generated by the network object within a certain period of time. Optionally, in the case of predicting the future network traffic generated by the network object, the selection of the above historical time period can also be related to the future time period. For example, if you want to predict the network traffic generated by the network object in the next month, you can select the network operation behavior data in the same month in the historical period. Among them, the future network traffic of the network object refers to the network traffic generated by the network object in a period of time in the future.

[0026] In the embodiment of the present application, the meaning of the future network traffic of the network object will be different depending on the network object. Correspondingly, the subsequent operations performed by the provider server 20 based on the network traffic (such as future network traffic) generated by the network object within a certain period of time will also be different. For example, in the case of predicting the future network traffic generated by the network object, if the network object is a commodity in the e-commerce field, the future network traffic of the network object can be the future sales volume of the commodity. Optionally, the provider server 20 can reasonably allocate the picking resources or inventory of the commodity according to the future sales volume of the commodity; if the network object is a film or TV series on an audio and video platform, the future network traffic of the network object can be the future click volume of the film or TV series. Optionally, the provider server 20 can adjust the listing or removal of different types of film and TV series according to the future click volume of the film or TV series. Figure 1 For example, if the network object is a commodity, Figure 1 As shown in ⑦, the provider server 20 can reasonably schedule the picking personnel of the goods and the quantity of goods in the warehouse according to the future sales volume of the goods to meet the supply and demand requirements of the goods.

[0027] Alternatively, if Figure 1As shown in ⑧, the system of this embodiment may also include other servers 40. The data processing server 10 may also provide the network traffic (such as future network traffic) generated by the network object within a certain period of time to the other servers 40, so that the other servers 40 can perform related operations according to the network traffic (such as future network traffic) generated by the network object within a certain period of time. The other servers 40 may be downstream servers that perform other data processing based on the network traffic (such as future network traffic) generated by the network object within a certain period of time, and the specific implementation form and function are not limited.

[0028] Combine the following Figure 2 The illustrated method embodiment provides a detailed description of the process by which the data processing server 10 predicts the network traffic (such as future network traffic) generated by a network object within a certain period of time.

[0029] Figure 2 Flow chart of the data processing method of the embodiment of the present application, as shown in Figure 2 As shown, the method includes:

[0030] S1. Acquire network operation behavior data generated by a user and associated with a first network object.

[0031] S2. Extracting, from the network operation behavior data, an association relationship between a second network object having a replaceable relationship with the first network object and the first network object.

[0032] S3. Predict the network traffic generated by the first network object within a certain period of time according to the association relationship between the second network object and the first network object and the attribute information of the first network object in at least one dimension.

[0033] Alternatively, if Figure 2 As shown, after step S3, the method further includes:

[0034] S4. Output the future network traffic of the first network object to the provider of the first network object, so that the provider can perform operations related to the first network object according to the future network traffic.

[0035] In the embodiment of the present application, the network object for which network traffic needs to be predicted is referred to as the first network object. When the user performs a network operation on the first network object on the terminal device, the terminal device can respond to the user's action of performing a network operation on the first network object, and record the network operation behavior data generated by the user in a log format according to the type of network operation, and provide the network operation behavior data directly or through the provider server to the data processing device. Among them, the network operation behavior data may include information such as the identification of the network object, the type of network operation, and the operation time. Depending on the first network object and the type of network operation, the network operation behavior data may also include other data. For example, taking the first network object as a commodity, if the network operation performed by the user is to browse the commodity, the corresponding network operation behavior data includes the name of the commodity, the address of the details page, the browsing time, and the browsing duration; if the network operation performed by the user is to purchase the commodity, the corresponding network operation behavior data includes the purchase quantity, order time, consignee, contact information, and collection address, delivery method, and other information of the commodity.

[0036] In this embodiment, considering that there are many categories of network objects and that network objects of the same category have similarities, when a user selects a certain network object, he may compare and screen different network objects. That is to say, among the network objects corresponding to the acquired network operation behavior data, there may be other network objects that can replace the first network object, which we call the second network object. Optionally, it is possible to combine the category and attribute information of the network object to determine whether there is a substitutable relationship between network objects. If the categories of two network objects are the same and the attributes are the same or similar, it can be determined that there is a substitutable relationship between the two network objects. For example, there is a substitutable relationship between brand A mobile phones and brand B mobile phones, or there is a substitutable relationship between the water heater of manufacturer C and the water heater of manufacturer D. Alternatively, it is possible to combine the functions of the network objects to determine whether there is a substitutable relationship between network objects. For example, down jackets and cotton jackets both have cold protection functions, and there is a substitutable relationship between the two. Whether there is a substitutable relationship between two network objects can be flexibly defined according to application requirements.

[0037] If the user selects the first network object, the first network object will generate a certain amount of network traffic. However, since the second network object is replaceable for the first network object, the existence of the second network object has a certain impact on the network traffic generated by the first network object, and it is possible that part of the traffic will be diverted by the second network object. In view of this, after obtaining the network operation behavior data, the second network object that has a replaceable relationship with the first network object can be identified from the obtained network operation behavior data, and the association relationship between the second network object and the first network object can be extracted; then, based on the association relationship between the second network object and the first network object and the attribute information of the first network object in at least one dimension, the future network traffic of the first network object is predicted. Among them, the attribute information of the first network object in at least one dimension is associated with the future network traffic of the first network object.

[0038] The attribute information of the first network object in at least one dimension includes attribute information in several dimensions, such as basic attribute information of the first network object, historical sequence information of the first network object, and future information of the first network object. The historical sequence information of the first network object may be serialized information related to the network traffic generated by the first network object in the historical period. The future information of the first network object refers to the relevant information that may affect the network traffic generated by the first network object in the future period. Taking the first network object as a commodity as an example, the attribute information in at least one dimension includes: basic attribute information of the commodity, historical sequence information of the commodity, and future information of the commodity. The basic attribute information of the commodity includes but is not limited to: the category, specification, price, origin, etc. of the commodity; the historical sequence information of the commodity includes at least one of the historical sales volume, historical purchase quantity, historical collection quantity, historical browsing quantity, etc. of the commodity; the future information of the commodity includes: commodity activity information related to the commodity in the future period, such as activity time, commodity pre-order quantity, user browsing number, and number of shopping carts, etc.

[0039] Among them, the association relationship between the second network object and the first network object indicates the influence of the second network object on the first network object, which can be reflected by whether the network operations performed by the user on the second network object and the first network object are the same or similar. For example, the first network object and the second network object are mobile phone P1 and mobile phone P2 respectively. If the user adds mobile phone P1 and mobile phone P2 to the shopping cart at the same time, or the user collects mobile phone P1 and mobile phone P2 at the same time, it means that the user likes both mobile phones and may choose to buy one of them. According to the adding to the shopping cart or collecting operations performed by the user on mobile phone P1 and mobile phone P2, it can be judged that mobile phone P1 and mobile phone P2 have a substitutable relationship, and there is a mutual influence between the sales of mobile phone P1 and the sales of mobile phone P2.

[0040] In an optional embodiment, when extracting the association relationship between the second network object and the first network object that is in a replaceable relationship with the first network object from the network operation behavior data, the number of times the same user performs the same network operation on the first network object and the second network object within a set time range can be extracted from the network operation behavior data, and based on the number, an association relationship is established between the first network object and the second network object. Among them, the set time range can be different according to the network object and the network operation, and the specific time length is not limited. For example, if the second network object and the first network object are commodities of the same category and the same or similar attributes, and the corresponding network operation is to add to a shopping cart, then the set time range can be 1 hour, 3 hours, one day, three days or one week. When extracting the association relationship between the two commodities, the number of times the user adds both commodities to the shopping cart within 1 hour, 3 hours, one day, three days or one week can be extracted. The number can reflect the possibility of mutual substitution between the two commodities to a certain extent. The more times, the higher the possibility of mutual substitution between the two commodities, that is, the stronger the association relationship between the two commodities. For another example, if the second network object and the first network object are songs, and the corresponding operation is to play each song, the set time range can be half an hour or one hour. When extracting the association between two songs, the number of times the user plays the two songs within half an hour or one hour can be extracted. The number can reflect the possibility of mutual substitution between the two songs to a certain extent. The more the number, the higher the possibility of mutual substitution between the two products, that is, the stronger the association between the two products.

[0041] Optionally, when extracting the number of times the same network operation is performed on the first network object and the second network object by the same user within a set time range from the network operation behavior data, the network operation behavior data under various network operation types generated by the same user can be identified from the network operation behavior data, that is, the network operation behavior data under various network operation types generated by the same user are first classified according to the user and the type of network operation; further, the network operation behavior data under various network operation types generated by the same user are arranged according to the order of occurrence time for the network objects involved to obtain the object sequence under various network operation types, that is, the network operation behavior data of each type is sorted according to the occurrence time for the network objects involved. Furthermore, based on the object sequence under various network operation types, the number of times the same network operation is performed on the first network object and the second network object by the same user within a set time range is extracted. Taking commodities as an example, users can perform different types of network operations such as browsing, collecting, adding to shopping carts, placing orders, sharing and commenting; correspondingly, network operation behavior data under different network operation types such as browsing, collecting, adding to shopping carts, placing orders, sharing and commenting will be generated.

[0042] In the embodiment of the present application, the occurrence time refers to the generation time of the network operation behavior data, which is also the time when the user performs the network operation on the network object. Therefore, the sequence of the time can reflect the sequence in which the user performs the network operation on the network object. The network operation behavior data generated by the same user is classified according to the network operation type and sorted according to the sequence of occurrence time, which network operations the same user has performed and the order of the network operations performed can be determined. Further, if it is defined that there is a substitutable relationship between two network objects of the same category and the same or similar attributes, it can also be determined whether the user has performed the same network operation on the two network objects with a substitution relationship within the set time range according to the type, attribute and corresponding occurrence time of each network object in the object sequence obtained after sorting; and then for the two network objects with a substitution relationship, the number of times the same network operation is performed on the two network objects by all users within the set time is counted. Optionally, the network object with a substitutable relationship can be identified first according to the type and attribute of the network object; and further, according to the corresponding occurrence time of the two network objects, it can be determined whether the user has initiated the same network operation on the two network objects within the set time range. Alternatively, the network objects for which the user initiates the same network operation within a set time range may be identified first according to the occurrence time corresponding to the network objects; and then, whether there is a substitutable relationship between these network objects may be determined according to the types and properties of the network objects.

[0043] For example, taking commodities as an example, after receiving the network operation behavior data of many users, the network operation behavior data of each user is classified according to the type of network operation, for example, the behavior data of adding to the shopping cart is classified into one category, the behavior data of placing an order is classified into one category, the behavior data of collecting commodities is classified into one category, and so on. Further, for each type of network operation behavior data, taking the network operation behavior data of adding to the shopping cart as an example, the commodities added to the shopping cart each time are sorted according to the order of adding to the shopping cart, and the commodity sequence added to the shopping cart is obtained. Assume that the commodities added to the shopping cart by a certain user are mobile phone P1, mobile phone P2, boots, down jackets, and children's picture books, etc., and these commodities are added to the shopping cart on the same day. Among them, mobile phone P1 and mobile phone P2 belong to the same category and have the same or similar attributes, and are two commodities with a substitution relationship. And since the time when mobile phone P1 and mobile phone P2 are added to the shopping cart is also within the set time range (such as within one day), the number of times mobile phone P1 and mobile phone P2 are added to the shopping cart can be increased by 1. Similarly, the network operation behavior data of other users can be processed, and the number of times other users add mobile phones P1 and P2 to the shopping cart within a set time range (such as one day) can be counted, and finally the final number of times mobile phones P1 and P2 are added to the shopping cart within the set time range can be obtained.

[0044] In some embodiments of the present application, when predicting the network traffic (such as future network traffic) generated by the first network object within a certain period of time based on the association relationship between the second network object and the first network object and the attribute information of the first network object in at least one dimension, the first-order fusion feature and the second-order fusion feature of the first network object can be calculated based on the association relationship between the second network object and the first network object and the attribute information of the first network object in at least one dimension. Furthermore, the network traffic (such as future network traffic) generated by the first network object within a certain period of time is predicted based on the first-order fusion feature and the second-order fusion feature of the first network object and the attribute information of the first network object in at least one dimension. Among them, the first-order fusion feature refers to the feature information of the first network object obtained after considering the influence of the second network object that has a direct association relationship with the first network object on the first network object, and the second-order fusion feature refers to the feature information of the first network object obtained after considering the influence of the second network object that has an indirect association relationship with the first network object on the first network object. For example, user A adds product P and product Q to the shopping cart within a set time. If product P and product Q are substitutable, then product P and product Q have a direct correlation, and product Q will affect the sales of product P. In this case, the influence of product Q can be considered when calculating the features of product P, thereby obtaining the first-order fusion features of product P. For another example, if user A adds product P and product Q to the shopping cart within a set time, and user B adds product Q and product U to the shopping cart within the same set time as user A; if product P and product Q are substitutable and product Q and product U are substitutable, it indicates that product P and product U have an indirect correlation to a certain extent. Therefore, product U will also affect the sales of product P. In this case, the influence of product U can also be considered when calculating the features of product P, thereby obtaining the second-order fusion features of product P.

[0045] In some optional embodiments of the present application, before calculating the first-order fusion features and the second-order fusion features of the first network object, the association relationship between the second network object and the first network object can be feature extracted to obtain the association relationship features between the second network object and the first network object; feature extraction can be performed on the attribute information of the first network object in at least one dimension to obtain the original features of the first network object; and the first-order fusion features and the second-order fusion features of the first network object are calculated based on the association relationship features and the original features of the first network object. Accordingly, when predicting the future network traffic of the first network object, it includes: splicing the original features of the first network object and the first-order fusion features and the second-order fusion features of the first network object to obtain the overall features of the first network object; using the traffic prediction algorithm based on deep learning to predict the overall features of the first network object to obtain the network traffic (such as future network traffic) generated by the first network object within a certain period of time.

[0046] In terms of implementation, the representation of the above-mentioned association features, original features, first-order fusion features and second-order fusion features is not limited. Optionally, the association features, original features, first-order fusion features and second-order fusion features can be represented by a matrix. Based on this, in an optional embodiment of the present application, when establishing an association relationship between a first network object and a second network object, a weighted graph representing the association relationship between the first network object and the second network object can be established according to the number of times the same user performs the same or similar network operations on the first network object and the second network object within a set time, wherein the number can be the weight of the edge between the first network object and the second network object. Optionally, the network traffic (such as future network traffic) generated by each network object with a substitutable relationship within a certain period of time can be predicted. Therefore, when constructing a weighted graph, all network objects with a substitutable relationship are used as nodes in the weighted graph, and edges are added between the nodes. The corresponding "number of times" between each other is used as the weight of the edge to obtain a weighted graph that can represent the substitutable relationship between network objects and its degree. Among them, the weighted graph can be expressed as G = (V, E, W). Among them, V represents the set of nodes representing network objects in the weighted graph, E represents the set of edges between network objects in the weighted graph, and W∈R n*n Represents the weight matrix of the edge in the weighted graph, and the matrix is ​​a symmetric matrix, where n is the number of network objects with each substitutable relationship. When calculating the weight matrix W, if there is an edge connection between the nodes of two network objects, the value of the position corresponding to the edge in W is the weight of the edge. If there is no edge connection between the nodes of the two network objects, the value of the corresponding position in W is 0; the larger the weight value of the edge, the stronger the association between the two network objects connected by the edge.

[0047] For example, taking the network object as a commodity, if user A adds commodity A and commodity B to the shopping cart within the set time, and commodity A and commodity B belong to the same commodity leaf category (indicating the same category) and have the same or similar attributes, then when constructing the weighted graph, there is an edge between commodity A and commodity B, and the number of times commodity A and commodity B are added to the shopping cart within the set time range is 1, then the weight corresponding to the edge is 1. Furthermore, it is counted that user B, user C, and user D add commodity B and commodity C to the shopping cart within the set time, and commodity B and commodity C belong to the same commodity leaf category (indicating the same category) and have the same or similar attributes, then in the weighted graph, there is an edge between commodity B and commodity C, because the number of times commodity B and commodity C are added to the shopping cart within the set time range is 3, then the weight corresponding to the edge is 3. Assume that the constructed weighted graph is as follows: Figure 3a As shown. Figure 3a Taking the weighted graph shown in the figure as an example, the weight matrix W is a 3*3 matrix, and each matrix element represents the number of times the same network operation is performed by the user between two products within a set time range. Figure 3aFor the weighted graph shown in , we can also define the degree matrix D corresponding to the weighted graph and the adjacency matrix T of the weighted graph. Figure 3a As shown, the degree of commodity A is 1, the degree of commodity B is 4, and the degree of commodity C is 3. Therefore, Figure 3a The corresponding degree matrix D is Figure 3a The corresponding adjacency matrix T is

[0048] Based on the above, one way to extract the features of the association relationship between network objects and obtain the features of the association relationship is to calculate the graph Laplacian matrix L corresponding to the weighted graph according to the adjacency matrix T and degree matrix D of the weighted graph. The graph Laplacian matrix is ​​an n*n square matrix, which can be defined as L=DA, where n represents the sum of the number of the first network object and the second network object, and n is an integer greater than or equal to 2. Then Figure 3a The corresponding Laplace matrix L can be expressed as: Furthermore, after normalizing the Laplacian matrix L, the graph Laplacian matrix used to transmit feature information can be obtained: Right now

[0049] In the embodiment of the present application, the network object itself has a lot of attribute information. Taking the network object as a commodity as an example, its basic attribute information may include information such as category, brand, price, specifications and place of origin. In addition, it may also include other attribute information such as the target audience, user praise, whether it is involved in promotional activities and the place of shipment. These attribute information in different dimensions may affect the sales volume and reputation of the commodity. Therefore, when predicting the future traffic of the network object, it is necessary to consider the degree of influence on the future traffic from the attribute information of different dimensions of the network object. Optionally, in the case of obtaining the characteristics of the association relationship between the second network object and the first network object in the above embodiment, the attribute information of the first network object in at least one dimension may be extracted to obtain the original characteristics of the first network object. When calculating the original characteristics of the first network object, the attribute information of the first network object and the second network object in at least one dimension may be extracted based on the neural network algorithm to obtain an n*m multidimensional feature matrix F∈R n*m*k, where n is the number of network objects in each category, m is the dimension of the feature, and m is an integer greater than or equal to 2, and k is the length of the feature sequence. Optionally, the neural network algorithms that can be used include but are not limited to: long short-term memory network algorithm, wavenet network algorithm, etc. The multidimensional feature matrix F is a high-dimensional summary and abstraction of the information of each dimension of the network object by the neural network algorithm. The network traffic (such as future network traffic) generated by each network object within a certain period of time can be output using the summarized and abstracted multidimensional feature matrix. It should be noted that the above embodiments are only exemplary descriptions. In the case of a multidimensional feature matrix, feature extraction can also be performed from attribute information in other dimensions, which will not be elaborated here. Figure 3a Taking the weighted graph shown in the figure as an example, the multidimensional feature matrix F corresponding to products A, B and C is

[0050] Since the multidimensional feature matrix calculated in the above embodiment is only a summary and abstraction of the feature information of the network object itself, and does not consider the mutual influence of the feature information between network objects, in order to more accurately predict the network traffic (such as future network traffic) generated by the first network object within a certain period of time, it is necessary to consider the influencing factors between network objects. For example, before buying a mobile phone, the user adds two mobile phones A and B of different brands to the shopping cart, but mobile phone A has a promotion in 3 days, then compared with mobile phone B, the user is more likely to buy mobile phone A in three days. Therefore, when predicting the future sales of mobile phone B, it is necessary to consider the factor that the promotion of mobile phone A will affect the sales of mobile phone B. In an optional embodiment of the present application, when the association relationship feature between the second network object and the first network object and the original feature of the first network object are obtained, the first-order fusion feature and the second-order fusion feature of the first network object can also be calculated according to the association relationship feature between the second network object and the first network object and the original feature of the first network object, and the influence of the second network object on the first network object is fully considered. Optionally, the first-order fusion feature of the first network object is calculated, and the normalized graph Laplacian matrix L sys Multiply it with the multidimensional feature matrix F to obtain a first-order fusion matrix F′, where each row of the first-order fusion matrix F′ represents a first-order fusion feature of a network object, including the first network object. Further, optionally, when the first-order fusion feature F′ of the first network object is obtained, the normalized graph Laplacian matrix L sys Multiplying the first-order fusion matrix F′ with the second-order fusion matrix F″, each row of the second-order fusion matrix F″ represents the second-order fusion feature of a network object, including the first network object. Figure 3a As an example of the weighted graph shown in the figure, the first-order fusion matrix Second-order fusion matrix

[0051] When the first-order fusion features and the second-order fusion features of the first network object are obtained, the future network traffic of the first network object can be predicted based on the first-order fusion features and the second-order fusion features of the first network object and the attribute information of the first network object in at least one dimension. Optionally, the original features of the first network object and the first-order fusion features and the second-order fusion features of the first network object can be spliced ​​to obtain the overall features of the first network object, and then the overall features of the first network object can be predicted using a traffic prediction algorithm based on deep learning to obtain the network traffic (such as future network traffic) generated by the first network object within a certain period of time. Since the overall features obtained after splicing not only include the summary and abstraction of the feature information of the network objects themselves under each category, but also reflect the factors affecting the feature information of each network object, therefore, the network traffic prediction based on the overall features after splicing has a higher accuracy.

[0052] Furthermore, in the case where the network traffic generated by the first network object within a certain period of time is predicted, the predicted network traffic can be output to the provider server of the first network object, so that the provider can perform operations related to the first network object according to the predicted network traffic. For example, in the above example of predicting the future sales of mobile phone P1, assuming that based on the association between mobile phone P1 and mobile phone P2 and the attribute information of mobile phone P1, it is predicted that the sales volume of mobile phone P1 in the next week will be 500 units. The provider can determine whether the inventory quantity of mobile phone P1 needs to be allocated based on the prediction result. If the current inventory of mobile phone P1 is 100 units, the provider can add another 400 units to avoid sales losses and user loss due to the inability to place orders or deliver in time due to the shortage of mobile phones while meeting the supply and demand requirements. For another example, if the first network object is a TV series, assuming that based on the attribute information and associations of different types of TV series, it is predicted that the playback volume of costume dramas is 30% higher than that of youth idol dramas; the playback volume of action movies is 40% higher than that of literary films, then the audio and video provider can adjust the number of different types of TV series on the shelves according to the prediction results. For example, for TV series, the number of costume dramas on the shelves can be appropriately increased, and the number of youth idol dramas on the shelves can be reduced; for movies, the number of action movies on the shelves can be appropriately increased, and the number of literary films on the shelves can be reduced.

[0053] The data processing method of the embodiment of the present application can be applied to e-commerce scenarios to predict the future sales of goods. In this scenario, the first network object is the first product, and the second network object is the second product. Based on this, a method for predicting the sales of goods is as follows: Figure 3b As shown, the following steps are included:

[0054] S1b. Obtain network operation behavior data generated by the user and associated with the first product.

[0055] S2b. Extracting the association relationship between the first commodity and the second commodity that is in a substitutable relationship with the first commodity from the network operation behavior data.

[0056] S3b. Predicting future sales of the first product based on the association relationship between the second product and the first product and attribute information of the first product in at least one dimension.

[0057] S4b. Output the future sales volume to the provider of the target product, so that the provider can perform operations related to the target product according to the future sales volume.

[0058] The commodity sales prediction method provided in the embodiment of the present application can obtain the network operation behavior data generated by the user and associated with the first commodity, and when the network operation behavior data is obtained, the association relationship between the second commodity and the first commodity that has a substitutable relationship with the first commodity is extracted from the network operation behavior data, and the future sales of the first commodity can be predicted based on the association relationship between the second commodity and the first commodity and the attribute information of the first commodity in at least one dimension, and the future sales can be output to the provider of the target commodity, so that the provider can perform operations related to the target commodity based on the future sales. The process of predicting the future sales of the first commodity is consistent with the implementation process of the above-mentioned data processing method. For details, please refer to the above-mentioned embodiment, and no repetition will be made here.

[0059] In e-commerce scenarios, users can perform a series of network operations on products through terminal devices, such as browsing, collecting, adding to shopping carts, commenting or sharing. When users perform these network operations, the terminal device can record the network operation behavior data generated by users when performing network operations in the form of logs, and provide these network operation behavior data to the data processing server in the cloud through the provider server or directly. The data processing server in the cloud can predict the sales volume of each product in the future based on the network operation behavior data, and return the predicted results to the provider. Based on the prediction results, the provider can reasonably allocate the inventory and picking resources corresponding to each product to avoid problems such as users being unable to place orders or unable to ship in time due to product shortages, causing user loss, or losses due to the squeezing of a large number of products, and resource waste due to uneven distribution of picking resources.

[0060] The following is a detailed description of the application of the method embodiments of the present application in different application scenarios.

[0061] Scenario Example 1:

[0062] In a scenario in the e-commerce field, providers can reasonably allocate the inventory of each commodity in the warehouse according to the future sales of the commodity, ensuring that the inventory of each commodity is reasonable and sufficient to avoid backlogs or shortages of some commodities, thereby reducing user experience decline, user churn and economic losses caused by users' inability to place orders or failure to deliver goods in a timely manner.

[0063] In this scenario embodiment, it can be used Figure 3b The method shown predicts the future sales of each commodity; then, based on the future sales of each commodity, the inventory of each commodity in the warehouse is adjusted in advance.

[0064] For example, when users choose a long sofa, they often compare it with a short sofa. If the sales volume of the long sofa in the next week is to be predicted based on the network operation behavior data, the prediction can be made based on the correlation between the long sofa and the short sofa and the attribute information of the long sofa in at least one dimension. Among them, the attribute information of the long sofa may include: historical weekly sales volume, brand, specifications, brand, current price, and price changes in the next week. Based on this information, the user's purchase possibility of the long sofa is measured, and then the sales volume of the long sofa is predicted. For example, according to the historical sales volume, the brand of the long sofa has higher sales volume than the brand of the short sofa, but in the next week, the short sofa has a promotion activity. According to the weekly sales volume of this long sofa under the promotion activity of the short sofa in the past, the sales volume of this long sofa will be reduced by 20% compared with the usual sales volume. Assuming that 300 people have added this long sofa to their shopping carts, it can be predicted that if there is a promotion for short sofas in the next week, 240 people may buy this long sofa. If the current stock of this long sofa is only 100, the provider can add about 250 long sofas based on the prediction results to avoid losses due to inventory backlogs.

[0065] Scenario Example 2:

[0066] In another scenario in the e-commerce field, providers can dispatch goods between multiple warehouses based on the future sales of goods, so that the inventory in each warehouse is reasonable and sufficient, to avoid uneven distribution of goods in different warehouses, thereby reducing user experience decline, user churn and economic costs caused by users' inability to place orders or inability to deliver goods in a timely manner.

[0067] In this scenario embodiment, taking the first product as an example, Figure 3bThe method shown predicts the future sales volume of the first commodity. Considering that warehouses in different regions may have different inventory requirements for the same commodity, the first commodity may be scheduled among multiple warehouses in advance according to the future sales volume of the first commodity. In an optional embodiment, when scheduling the first commodity among multiple warehouses, the reserve inventory of the first commodity in the multiple warehouses may be determined according to the historical shipment volume of the first commodity in the multiple warehouses and the future sales volume of the first commodity; and the first commodity may be scheduled among the multiple warehouses according to the current inventory and the reserve inventory of the first commodity in the multiple warehouses.

[0068] For example, a brand provider wants to put a down jacket on the shelves before winter. Before putting it on the shelves, the provider uses Figure 3b The method shown predicts that the future sales volume of this down jacket will be 8,000 pieces, taking into account the impact of other brands and other styles of down jackets on this down jacket. Assume that the provider has warehouses in areas A and B respectively, and area A is close to the south and area B is close to the north. In historical sales, the warehouse in area A shipped 1,000 pieces of this down jacket, and the warehouse in area B shipped 3,000 pieces of this down jacket. Then, based on the proportional relationship between the historical shipments of this down jacket of the two warehouses, the reserve inventory of this down jacket that each warehouse needs to have can be determined. For example, the warehouse in area A needs 2,000 pieces of inventory, and the warehouse in area B needs 6,000 pieces of inventory. If, when initially replenishing the warehouse, 5,000 pieces of this down jacket are stored in the warehouse in area A, and 5,000 pieces of this down jacket are stored in the warehouse in area B, then 1,000-2,000 pieces of this down jacket can be transferred from the warehouse in area A to the warehouse in area B to ensure the inventory in the warehouse in area B, and to ensure that the goods can be delivered to the user in time after the user places an order, thereby improving the user's shopping experience.

[0069] Scenario Example 3:

[0070] In another scenario in the e-commerce field, the provider can dynamically allocate the picking resources in the warehouse among multiple commodities according to the future sales volume of the commodities, so as to avoid uneven resource allocation and waste of resources. In this embodiment, the form of picking resources is not limited, and can be human resources or material resources, such as storage and handling robots or picking robots.

[0071] In this scenario embodiment, it can be used Figure 3b The method shown predicts the future sales volume of each commodity; further, according to the future sales volume of each commodity, the picking resources in the warehouse are reasonably allocated.

[0072] For example, in e-commerce sales, there will be promotions during Double 11, Double 12 and other holidays, and the promotion is relatively strong. During the event, the sales of goods may be several times or even dozens of times higher than usual. Before the event starts, the provider can configure the picking resources in the warehouse according to the predicted sales of the goods to avoid affecting the user's consumption experience during the event. Assuming that a store conducts a buy one get one free promotion for some of its products during Double 11, the predicted sales of these products will be 10 times higher than usual. However, the current staffing is difficult to bear the pressure of packing and shipping goods during the event, so the provider needs to make appropriate adjustments to the staffing. For example, if two people are usually needed to arrange packaging and shipping, then according to the forecast results, the provider can temporarily recruit some temporary workers to pack and ship the goods during the event. For example, when recruiting 18 temporary workers to pack and ship the goods during the event, or, considering the cost issue, the provider can also appropriately reduce the number of recruits and adjust the working hours and remuneration to obtain the maximum benefit.

[0073] Scenario Example 4:

[0074] In another scenario in the e-commerce field, when the data processing server in the cloud predicts the future sales volume of a product, in addition to providing the predicted future sales volume to the provider, the predicted future sales volume can also be provided to other servers in the cloud or downstream servers of the provider, so that other servers can perform corresponding processing according to the predicted results. For example, in the process of providing data services and support to the provider, in addition to processing the network operation behavior data generated by users, the cloud can also perform other operations such as statistics and analysis on the predicted results, as well as test the accuracy of the predicted results, and then optimize and upgrade the processing algorithm to provide better services to the provider.

[0075] For example, in large-scale promotional activities such as Double Eleven, in addition to the transaction sales that are several times or even dozens of times higher than usual, the number of returns also increases. Many users buy a lot of unnecessary or inappropriate goods due to impulse consumption during the event, and may return a large number of goods after receiving them. Therefore, other servers in the cloud can also perform statistical analysis on the prediction results of the data processing server based on the sales predicted by the data processing server, the actual transaction sales, and the number of returns and refunds, and continuously optimize the prediction algorithm to improve the accuracy of the prediction algorithm. For another example, the cloud can also provide the predicted future sales of goods according to the type of goods, the place of shipment, and the place of receipt to different after-sales service platforms, such as logistics manufacturers. The after-sales service platform can reasonably arrange the number of staff at each logistics station based on the predicted future sales of the goods to avoid the situation where a large number of goods are piled up at the logistics station and no one delivers them.

[0076] In an embodiment of the present application, the network operation behavior data generated by the user performing the network operation is processed to obtain the attribute information of the network object corresponding to the user performing the network operation in each dimension, and based on this information, the future network traffic of the network object in each dimension can be predicted. Moreover, when predicting the future network traffic of the network object, it is not just predicted according to the attribute information of the individual network object, but the attribute information of the network object with a substitution relationship is extracted and fused to obtain a matrix containing the attribute information of the influence relationship between the network objects, and the matrix is ​​predicted through a neural network algorithm. In this way, when predicting the future network traffic of a certain network object, not only the factors that affect the prediction results of the future network traffic of the network object itself are taken into account, but also the attribute information of other network objects that have a substitution relationship with the network object can be taken into account, which helps to improve the accuracy of the prediction results.

[0077] It should be noted that the execution subject of each step of the method provided in the above embodiment can be the same device, or the method can be executed by different devices. For example, the execution subject of steps S1 to S4 can be device A; for another example, the execution subject of steps S1 and S2 can be device A, and the execution subject of steps S3 and S4 can be device B; and so on.

[0078] In addition, in some of the processes described in the above embodiments and the accompanying drawings, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel, and the sequence numbers of the operations, such as S1, S2, etc., are only used to distinguish between different operations, and the sequence numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0079] The present application also provides a data processing server, such as Figure 4 As shown, the data processing server includes: a processor 41 and a memory 42 storing a computer program; wherein the processor 41 and the memory 42 can be one or more.

[0080] The memory 42 is mainly used to store computer programs, which can be executed by the processor, so that the processor controls the data processing server to realize corresponding functions, complete corresponding actions or tasks. In addition to storing computer programs, the memory can also be configured to store various other data to support operations on the data processing server, examples of which include instructions for any application or method operating on the data processing server.

[0081] The memory 42 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0082] In the embodiment of the present application, the implementation form of the processor 41 is not limited, for example, it can be but not limited to a CPU, a GPU or an MCU. The processor 31 can be regarded as a control system of the data processing server, which can be used to execute the computer program stored in the memory 42 to control the data processing server to implement corresponding functions and complete corresponding actions or tasks. It is worth noting that, depending on the implementation form of the data processing server and the different scenarios in which it is located, the functions to be implemented, the actions or tasks to be completed will be different; accordingly, the computer programs stored in the memory 42 will also be different, and the processor 41 can control the data processing server to implement different functions and complete different actions or tasks by executing different computer programs.

[0083] In an alternative embodiment, if Figure 4 As shown, the data processing server also includes other components such as a communication component 43 and a power supply component 44. Figure 4 The following only shows some components, which does not mean that the database engine only includes Figure 4 Components shown.

[0084] In the embodiment of the present application, when the processor 41 executes the computer program in the memory 42, it is used to: obtain network operation behavior data generated by a user and associated with a first network object; extract the association relationship between a second network object that has a replaceable relationship with the first network object and the first network object from the network operation behavior data; and predict the network traffic (such as future network traffic) generated by the first network object within a certain period of time based on the association relationship between the second network object and the first network object and the attribute information of the first network object in at least one dimension.

[0085] In an optional embodiment, the processor 41 is further configured to: output the predicted network traffic to a provider of the first network object, so that the provider can perform operations related to the first network object according to the predicted network traffic.

[0086] In an optional embodiment, when the processor 41 extracts the association relationship between the second network object that has a replaceable relationship with the first network object and the first network object from the network operation behavior data, it is used to: extract the number of times the same network operation is performed on the first network object and the second network object by the same user within a set time range from the network operation behavior data; and establish an association relationship between the first network object and the second network object based on the number of times.

[0087] In an optional embodiment, when the processor 41 extracts the number of times the same network operation is performed on the first network object and the second network object by the same user within a set time range from the network operation behavior data, it is used to: identify the network operation behavior data under various network operation types generated by the same user from the network operation behavior data; arrange the network objects involved in the network operation behavior data under various network operation types generated by the same user in chronological order of occurrence to obtain object sequences under various network operation types; and extract the number of times the same network operation is performed on the first network object and the second network object by the same user within a set time range based on the object sequences under various network operation types.

[0088] In an optional embodiment, the second network object is a network object that appears in the network operation behavior data and has the same category as the first network object and the same or similar attributes as the first network object.

[0089] In an optional embodiment, when predicting the network traffic generated by the first network object within a certain period of time based on the association relationship between the second network object and the first network object and the attribute information of the first network object in at least one dimension, the processor 41 is used to: calculate the first-order fusion features and the second-order fusion features of the first network object based on the association relationship between the second network object and the first network object and the attribute information of the first network object in at least one dimension; predict the network traffic generated by the first network object within a certain period of time based on the first-order fusion features and the second-order fusion features of the first network object and the attribute information of the first network object in at least one dimension.

[0090] In an optional embodiment, when calculating the first-order fusion features and the second-order fusion features of the first network object based on the association relationship between the second network object and the first network object and the attribute information of the first network object in at least one dimension, the processor 41 is used to: extract features of the association relationship between the second network object and the first network object to obtain the association relationship features between the second network object and the first network object; extract features of the attribute information of the first network object in at least one dimension to obtain the original features of the first network object; and calculate the first-order fusion features and the second-order fusion features of the first network object based on the association relationship features and the original features of the first network object.

[0091] In an optional embodiment, when predicting the network traffic generated by the first network object within a certain period of time based on the first-order fusion features and the second-order fusion features of the first network object and the attribute information of the first network object in at least one dimension, the processor 41 is used to: splice the original features of the first network object and the first-order fusion features and the second-order fusion features of the first network object to obtain the overall features of the first network object; and predict the overall features of the first network object using a traffic prediction algorithm based on deep learning to obtain the network traffic generated by the first network object within a certain period of time.

[0092] In an optional embodiment, when establishing an association relationship between a first network object and a second network object according to a number, the processor 41 is used to: establish a weighted graph representing the association relationship between the first network object and the second network object according to the number, where the number is the weight of the edge between the first network object and the second network object.

[0093] In an optional embodiment, when the processor 41 extracts features of the association relationship between the second network object and the first network object to obtain features of the association relationship between the second network object and the first network object, it is used to: calculate a graph Laplacian matrix corresponding to the weighted graph, where the graph Laplacian matrix is ​​n*n, where n represents the sum of the number of the first network object and the second network object, and n is an integer greater than or equal to 2.

[0094] In an optional embodiment, when the processor 41 performs feature extraction on the attribute information of the first network object in at least one dimension to obtain the original features of the first network object, it is used to: perform feature extraction on the attribute information of the first network object and the second network object in at least one dimension based on a neural network algorithm to obtain an n*m multidimensional feature matrix, where m is the dimension of the feature and m is an integer greater than or equal to 2.

[0095] In an optional embodiment, when the processor 41 calculates the first-order fusion features and the second-order fusion features of the first network object based on the association relationship features and the original features of the first network object, it is used to: multiply the graph Laplacian matrix with the multidimensional feature matrix to obtain a first-order fusion matrix, wherein the first-order fusion matrix includes the first-order fusion features of the first network object; and multiply the graph Laplacian matrix with the first-order fusion matrix to obtain a second-order fusion matrix, wherein the second-order fusion matrix includes the second-order fusion features of the first network object.

[0096] In an optional embodiment, the network objects are commodities in the e-commerce field, that is, the first network object is the first commodity, and the second network object is the second commodity; accordingly, when the certain time is a future period, the network traffic generated by the first network object in the future period is the future sales volume of the first commodity. Based on this, after predicting the future sales volume of the first commodity, the processor 41 is further used to: pre-schedule the first commodity among multiple warehouses according to the future sales volume of the first commodity; and / or dynamically configure the picking resources in the warehouse between the first commodity and other commodities according to the future sales volume of the first commodity.

[0097] In an optional embodiment, when the processor 41 schedules the first commodity among multiple warehouses, it is specifically used to: determine the reserve inventory of the first commodity in multiple warehouses based on the historical shipment volume of the first commodity in multiple warehouses and the future sales volume of the first commodity; and schedule the first commodity among multiple warehouses based on the current inventory and reserve inventory of the first commodity in the multiple warehouses.

[0098] Accordingly, an embodiment of the present application also provides a computer-readable storage medium storing a computer program, which, when executed, can implement the steps that can be executed by the data processing server in the above method embodiment.

[0099] Above Figure 4 The communication component in is configured to facilitate wired or wireless communication between the device where the communication component is located and other devices. The device where the communication component is located can access a wireless network based on a communication standard, such as WiFi, 2G, 3G, 4G / LTE, 5G and other mobile communication networks, or a combination thereof. In an exemplary embodiment, the communication component receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0100] Above Figure 4The power supply component in the device provides power to various components of the device where the power supply component is located. The power supply component may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device where the power supply component is located.

[0101] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0102] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0103] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0104] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0105] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0106] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0107] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0108] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0109] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the present application.

Claims

1. A data processing method, characterized in that: include: Acquire network operation behavior data generated by a user and associated with a first network object; Extracting, from the network operation behavior data, an association relationship between a second network object having a replaceable relationship with the first network object and the first network object; According to the association relationship between the second network object and the first network object and the attribute information of the first network object in at least one dimension, a first-order fusion feature and a second-order fusion feature of the first network object are calculated; the first-order fusion feature refers to feature information obtained after considering the influence of the second network object that has a direct association relationship with the first network object on the first network object, and the second-order fusion feature refers to feature information obtained after considering the influence of the second network object that has an indirect association relationship with the first network object on the first network object; The network traffic generated by the first network object within a certain period of time is predicted according to the first-order fusion features and the second-order fusion features of the first network object and the attribute information of the first network object in at least one dimension.

2. The method according to claim 1, characterized in that Also includes: The network traffic is output to a provider of the first network object, so that the provider performs operations related to the first network object according to the network traffic.

3. The method according to claim 1, characterized in that Extracting, from the network operation behavior data, an association relationship between a second network object having a replaceable relationship with the first network object and the first network object, comprising: Extracting, from the network operation behavior data, the number of times the same network operation is performed on the first network object and the second network object by the same user within a set time range; According to the number of times, an association relationship is established between the first network object and the second network object.

4. The method according to claim 3, characterized in that Extracting the number of times the same network operation is performed on the first network object and the second network object by the same user within a set time range from the network operation behavior data includes: From the network operation behavior data, identifying network operation behavior data of various network operation types generated by the same user; The network operation behavior data of various network operation types generated by the same user are arranged in the order of occurrence time for the network objects involved to obtain the object sequence of various network operation types; Based on the object sequences under the various network operation types, the number of times the same network operation is performed on the first network object and the second network object by the same user within a set time range is extracted.

5. The method according to claim 3 or 4, characterized in that: The second network object is a network object that appears in the network operation behavior data and has the same category as the first network object and the same or similar attributes as the first network object.

6. The method according to claim 3 or 4, characterized in that: Calculating first-order fusion features and second-order fusion features of the first network object according to the association relationship between the second network object and the first network object and attribute information of the first network object in at least one dimension includes: Extracting features of the association relationship between the second network object and the first network object to obtain features of the association relationship between the second network object and the first network object; Performing feature extraction on the attribute information of the first network object in at least one dimension to obtain original features of the first network object; According to the association relationship features and the original features of the first network object, first-order fusion features and second-order fusion features of the first network object are calculated.

7. The method according to claim 6, characterized in that Predicting network traffic generated by the first network object within a certain period of time according to the first-order fusion feature and the second-order fusion feature of the first network object and the attribute information of the first network object in at least one dimension includes: splicing the original features of the first network object and the first-order fusion features and the second-order fusion features of the first network object to obtain the overall features of the first network object; The overall characteristics of the first network object are predicted using a traffic prediction algorithm based on deep learning to obtain the network traffic generated by the first network object within a certain period of time.

8. The method according to claim 7, characterized in that Establishing an association relationship between the first network object and the second network object according to the number of times includes: According to the number, a weighted graph representing the association relationship between the first network object and the second network object is established, and the number is the weight of the edge between the first network object and the second network object.

9. The method according to claim 8, characterized in that Extracting features of the association relationship between the second network object and the first network object to obtain features of the association relationship between the second network object and the first network object includes: A graph Laplacian matrix corresponding to the weighted graph is calculated, where the graph Laplacian matrix is ​​n*n, where n represents the sum of the number of the first network objects and the number of the second network objects, and n is an integer greater than or equal to 2.

10. The method according to claim 9, characterized in that Extracting features from the attribute information of the first network object in at least one dimension to obtain original features of the first network object includes: Based on the neural network algorithm, feature extraction is performed on the attribute information of the first network object and the second network object in at least one dimension to obtain an n*m multidimensional feature matrix, where m is the dimension of the feature and m is an integer greater than or equal to 2.

11. The method according to claim 10, characterized in that Calculating first-order fusion features and second-order fusion features of the first network object according to the association relationship features and the original features of the first network object includes: Multiplying the graph Laplacian matrix with the multidimensional feature matrix to obtain a first-order fusion matrix, wherein the first-order fusion matrix includes the first-order fusion features of the first network object; The graph Laplacian matrix is ​​multiplied by the first-order fusion matrix to obtain a second-order fusion matrix, wherein the second-order fusion matrix includes the second-order fusion features of the first network object.

12. A commodity sales forecasting method, characterized in that: include: Acquire network operation behavior data generated by the user and associated with the first product; Extracting, from the network operation behavior data, an association relationship between the first commodity and a second commodity that is in a substitutable relationship with the first commodity; Calculate the first-order fusion feature and the second-order fusion feature of the first product according to the association relationship between the second product and the first product and the attribute information of the first product in at least one dimension; the first-order fusion feature refers to the feature information obtained after considering the influence of the second product that has a direct association relationship with the first product on the first product, and the second-order fusion feature refers to the feature information obtained after considering the influence of the second product that has an indirect association relationship with the first product on the first product; predicting future sales volume of the first product based on the first-order fusion features and the second-order fusion features of the first product and attribute information of the first product in at least one dimension; The future sales volume is output to a provider of the first commodity, so that the provider performs operations related to the first commodity according to the future sales volume.

13. The method according to claim 12, characterized in that The attribute information of the first commodity in at least one dimension includes: basic attribute information of the first commodity, historical sequence information of the first commodity, and future information of the first commodity; Among them, the historical sequence information of the first commodity includes at least one of the historical sales volume, historical collection volume and historical added purchase volume of the first commodity; the future information of the first commodity includes: commodity activity information related to the first commodity in a future period.

14. A storage scheduling method, characterized in that: include: Acquire network operation behavior data generated by the user and associated with the first product; Extracting, from the network operation behavior data, an association relationship between the first commodity and a second commodity that is in a substitutable relationship with the first commodity; Calculate the first-order fusion feature and the second-order fusion feature of the first product according to the association relationship between the second product and the first product and the attribute information of the first product in at least one dimension; the first-order fusion feature refers to the feature information obtained after considering the influence of the second product that has a direct association relationship with the first product on the first product, and the second-order fusion feature refers to the feature information obtained after considering the influence of the second product that has an indirect association relationship with the first product on the first product; Predict future sales of the first commodity based on first-order fusion features and second-order fusion features of the first commodity and attribute information of the first commodity in at least one dimension; and pre-schedule the first commodity among multiple warehouses based on the future sales of the first commodity.

15. The method according to claim 14, characterized in that According to the future sales volume of the first commodity, the first commodity is pre-scheduled among multiple warehouses, including: Determining a reserve inventory of the first commodity in the multiple warehouses according to the historical shipment volume of the first commodity in the multiple warehouses and the future sales volume of the first commodity; The first commodity is scheduled among the multiple warehouses according to the current inventory and the reserve inventory of the first commodity in the multiple warehouses.

16. A warehouse resource configuration method, characterized in that: include: Acquire network operation behavior data generated by the user and associated with the first product; Extracting, from the network operation behavior data, an association relationship between the first commodity and a second commodity that is in a substitutable relationship with the first commodity; Calculate the first-order fusion feature and the second-order fusion feature of the first product according to the association relationship between the second product and the first product and the attribute information of the first product in at least one dimension; the first-order fusion feature refers to the feature information obtained after considering the influence of the second product that has a direct association relationship with the first product on the first product, and the second-order fusion feature refers to the feature information obtained after considering the influence of the second product that has an indirect association relationship with the first product on the first product; predicting future sales volume of the first product based on the first-order fusion features and the second-order fusion features of the first product and attribute information of the first product in at least one dimension; According to the future sales volume of the first commodity, picking resources in the warehouse are dynamically allocated between the first commodity and other commodities.

17. A data processing server, characterized in that: include: a processor and a memory storing a computer program; The processor is configured to execute the computer program to: Acquire network operation behavior data generated by a user and associated with a first network object; Extracting, from the network operation behavior data, an association relationship between a second network object having a replaceable relationship with the first network object and the first network object; According to the association relationship between the second network object and the first network object and the attribute information of the first network object in at least one dimension, a first-order fusion feature and a second-order fusion feature of the first network object are calculated; the first-order fusion feature refers to feature information obtained after considering the influence of the second network object that has a direct association relationship with the first network object on the first network object, and the second-order fusion feature refers to feature information obtained after considering the influence of the second network object that has an indirect association relationship with the first network object on the first network object; The network traffic generated by the first network object within a certain period of time is predicted according to the first-order fusion features and the second-order fusion features of the first network object and the attribute information of the first network object in at least one dimension.

18. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the processor is caused to implement the steps in the method according to any one of claims 1 to 16.

Images