Live broadcast scene item sales data prediction method, model training method and medium

Abstract

The present application provides a method for predicting sales data of items in a live broadcast scenario, a training method for a model, and a medium. In the method for predicting sales data of items in a live broadcast scenario, the live broadcast resources of the live broadcast are obtained from the live broadcast server, and the item data is obtained from the platform to which the items to be recommended for live broadcast belong. Then, based on the live broadcast resources and the item data, the item data matrix corresponding to the items to be recommended for live broadcast is obtained, and the item data matrix is ​​input into the sales data prediction model for processing. After obtaining the predicted sales data of each item, the predicted sales data is output. The sales data prediction model is trained based on the sales characteristics of multiple items in different live broadcast resources and the actual sales data. In this solution, the sales data prediction model is used to perform prediction processing based on the sales characteristics of the items. Compared with manual prediction, it effectively shortens the time required for the prediction process and improves the accuracy of the predicted sales data.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method for predicting sales data of items in a live broadcast scene, a model training method and a medium. Background Art

[0002] With the development of technology, the way users get the items they need in shopping malls or supermarkets has been gradually replaced by online shopping. With the increasing popularity of the live broadcast industry, anchors can introduce information about different items to users by recommending items to users in the live broadcast room. Users can obtain items as needed, but since it is impossible to accurately predict the number of items needed, it often causes problems such as insufficient items or too many items to be prepared.

[0003] In the prior art, before live broadcast, it is necessary to predict the sales data of items during the live broadcast in order to prepare items, which is currently mainly judged by manual experience. Specifically, before live broadcast, the staff needs to conduct surveys to understand the popularity of the items, the matching degree between the applicable time period and the current time period, the market saturation, etc., and predict the sales data of the items based on multiple factors.

[0004] In the process of implementing the present invention, the inventors found that there are at least the following problems in the prior art: in the above-mentioned method of predicting item sales data, the staff cannot understand all the required information and cannot effectively process the obtained data, the prediction process takes a long time, and the accuracy of the predicted data is low. Summary of the invention

[0005] The present application provides a method for predicting sales data of items in a live broadcast scenario, a model training method and a medium to solve the problems of existing methods for predicting sales data of items, that the staff cannot understand all the required information, cannot effectively process the obtained data, the prediction process takes a long time, and the accuracy of the predicted data is low.

[0006] In a first aspect, an embodiment of the present application provides a method for predicting sales data of items in a live broadcasting scene, comprising:

[0007] Acquire live broadcast resources for this live broadcast from the live broadcast server, and acquire item data from the platform to which the items to be recommended for live broadcast belong, wherein the live broadcast resources include live broadcast time and / or live broadcast entry resource information;

[0008] According to the live broadcast resource and the item data, an item data matrix corresponding to the items to be recommended for live broadcast is obtained; the item data matrix is ​​used to represent a matrix of sales characteristics of each item to be recommended for live broadcast;

[0009] The item data matrix is ​​input into the pre-acquired sales data prediction model for processing to obtain predicted sales data for each item; wherein the sales data prediction model is a prediction model trained based on the sales characteristics of multiple items in different live broadcast resources and actual sales data;

[0010] The predicted sales data is output.

[0011] In a possible design of the first aspect, the items to be recommended for live broadcast include multiple items, and the item data of each item includes: the name, category, purpose, sales price, sales data, traffic, number of purchases, number of add-to-carts, and price on a competing product platform of the item in at least one historical time period;

[0012] Accordingly, according to the live broadcast resource and the item data, an item data matrix corresponding to the item to be recommended for live broadcast is obtained, including:

[0013] According to the live broadcast resource and the item data of each item, the sales characteristics of each item are obtained, and the item data matrix is ​​obtained according to the sales characteristics of each item;

[0014] Among them, the sales characteristics of each item include: basic information of the item, cross-correlation coefficient between price and sales data, statistical characteristics of sales data, statistical characteristics of price, ratio information of sales data and median of sales data, and live broadcast resources.

[0015] In another possible design of the first aspect, the method further includes:

[0016] Output a stocking instruction based on the existing quantity of each item and the predicted sales data; the stocking instruction is used to indicate the replenishment quantity of each item.

[0017] In a second aspect, an embodiment of the present application provides a method for training a sales data prediction model of an item, comprising:

[0018] According to the personal information of the current anchor and the item data of the historical live broadcast items, multiple target anchors whose similarity with the current anchor reaches a preset value are obtained;

[0019] Obtaining a training set for item sales data prediction, wherein the training set for item sales data prediction includes: live broadcast resources and sku lists of each target anchor, and the sku list corresponding to each target anchor includes item data of each item historically broadcast by the target anchor;

[0020] Model training is performed according to the sales data prediction training set to obtain the sales data prediction model.

[0021] In a possible design of the second aspect, obtaining multiple target anchors whose similarity with the current anchor reaches a preset value according to the personal information of the current anchor and the item data of the historical live broadcast items includes:

[0022] According to the personal information of the current anchor and the item data of the historical live broadcast items, an information model of the current anchor is constructed;

[0023] Get information models of multiple anchors;

[0024] According to the information model of the current anchor and the information models of the multiple anchors, the multiple target anchors are determined from the other multiple anchors; wherein a cosine distance between the information model of each target anchor and the information model of the current anchor is less than a preset value.

[0025] In another possible design of the second aspect, performing model training according to the sales data prediction training set to obtain the sales data prediction model includes:

[0026] According to the sales data, the live broadcast resources in the training set and the item data of each item are predicted to obtain an item feature matrix; the item data matrix is ​​a matrix used to represent the sales characteristics of each item;

[0027] Model training is performed based on the item feature matrix and the actual sales data of each item to obtain the sales data prediction model.

[0028] Optionally, in the sales data prediction training set, the live broadcast resources include live broadcast time and / or live broadcast portal resource information, and the item data of each item includes: the name, category, purpose, sales price in at least one historical time period, sales data, traffic, number of purchases, number of add-to-carts, and price on competing platforms of the item;

[0029] Accordingly, the live broadcast resources and item data of each item in the training set are predicted based on the sales data to obtain an item feature matrix, including:

[0030] According to the live broadcast resource and the item data of each item, the sales characteristics of each item are obtained, and the item data matrix is ​​obtained according to the sales characteristics of each item;

[0031] Among them, the sales characteristics of each item include: basic information of the item, cross-correlation coefficient between price and sales data, statistical characteristics of sales data, statistical characteristics of price, ratio information of sales data and median of sales data, and live broadcast resources.

[0032] In a third aspect, an embodiment of the present application provides a device for predicting sales data of items in a live broadcasting scene, including: an acquisition module, a processing module, and an output module;

[0033] The acquisition module is used to acquire live broadcast resources for this live broadcast from the live broadcast server, and acquire item data from the platform to which the items to be recommended for live broadcast belong, wherein the live broadcast resources include live broadcast time and / or live broadcast entry resource information;

[0034] The acquisition module is further used to acquire an item data matrix corresponding to the items to be recommended for live broadcast according to the live broadcast resources and the item data; the item data matrix is ​​a matrix used to represent sales characteristics of each item to be recommended for live broadcast;

[0035] The processing module is used to input the item data matrix into the pre-acquired sales data prediction model for processing to obtain the predicted sales data of each item; wherein the sales data prediction model is a prediction model trained based on the sales characteristics of multiple items in different live broadcast resources and actual sales data;

[0036] The output module is used to output the predicted sales data.

[0037] In a possible design of the third aspect, the items to be recommended for live broadcast include multiple items, and the item data of each item includes: the name, category, purpose, sales price, sales data, traffic, number of purchases, number of add-to-carts, and price on a competing product platform of the item in at least one historical time period;

[0038] The acquisition module is specifically used for:

[0039] Obtain the live broadcast resources of this live broadcast and the item data of all items to be recommended for live broadcast; wherein the live broadcast resources include the live broadcast time and / or live broadcast portal resource information, and the item data of each item includes: the name, category, purpose, sales price in at least one historical time period, sales data, traffic, number of purchases, number of add-to-carts, and price on competing product platforms of the item;

[0040] According to the live broadcast resource and the item data of each item, the sales characteristics of each item are obtained, and the item data matrix is ​​obtained according to the sales characteristics of each item;

[0041] Among them, the sales characteristics of each item include: basic information of the item, cross-correlation coefficient between price and sales data, statistical characteristics of sales data, statistical characteristics of price, ratio information of sales data and median of sales data, and live broadcast resources.

[0042] In another possible design of the third aspect, the processing module is further used to output a stocking instruction based on the existing quantity of each item and the predicted sales data; the stocking instruction is used to indicate the replenishment quantity of each item.

[0043] In a fourth aspect, an embodiment of the present application provides a training device for a sales data prediction model of an item, comprising: an acquisition module and a training module;

[0044] The acquisition module is used to acquire multiple target anchors whose similarity with the current anchor reaches a preset value based on the personal information of the current anchor and the item data of the historical live broadcast items;

[0045] The acquisition module is also used to acquire a training set of item sales data prediction, wherein the training set of item sales data prediction includes: live broadcast resources and sku lists of each target anchor, and the sku list corresponding to each target anchor includes item data of each item that was historically live broadcasted by the target anchor;

[0046] The training module is used to perform model training based on the sales data prediction training set to obtain the sales data prediction model.

[0047] In a possible design of the fourth aspect, the acquisition module is specifically used to:

[0048] According to the personal information of the current anchor and the item data of the historical live broadcast items, an information model of the current anchor is constructed;

[0049] Get information models of multiple anchors;

[0050] According to the information model of the current anchor and the information models of the multiple anchors, the multiple target anchors are determined from the other multiple anchors; wherein a cosine distance between the information model of each target anchor and the information model of the current anchor is less than a preset value.

[0051] In another possible design of the fourth aspect, the training module is specifically used to:

[0052] According to the sales data, the live broadcast resources in the training set and the item data of each item are predicted to obtain an item feature matrix; the item data matrix is ​​a matrix used to represent the sales characteristics of each item;

[0053] Model training is performed based on the item feature matrix and the actual sales data of each item to obtain the sales data prediction model.

[0054] Optionally, in the sales data prediction training set, the live broadcast resources include live broadcast time and / or live broadcast portal resource information, and the item data of each item includes: the name, category, purpose, sales price in at least one historical time period, sales data, traffic, number of purchases, number of add-to-carts, and price on competing platforms of the item;

[0055] Accordingly, the acquisition module is specifically used for:

[0056] According to the live broadcast resource and the item data of each item, the sales characteristics of each item are obtained, and the item data matrix is ​​obtained according to the sales characteristics of each item;

[0057] Among them, the sales characteristics of each item include: basic information of the item, cross-correlation coefficient between price and sales data, statistical characteristics of sales data, statistical characteristics of price, ratio information of sales data and median of sales data, and live broadcast resources.

[0058] In a fifth aspect, an embodiment of the present application provides an electronic device, comprising: a processor, a memory, and computer program instructions stored in the memory and executable on the processor, wherein the processor implements the first aspect and the methods provided by various possible designs when executing the computer program instructions.

[0059] In a sixth aspect, an embodiment of the present application provides an electronic device, comprising: a processor, a memory, and computer program instructions stored in the memory and executable on the processor, wherein the processor implements the second aspect and the methods provided by various possible designs when executing the computer program instructions.

[0060] In a seventh aspect, the present application provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the methods provided in the first aspect and various possible designs.

[0061] In an eighth aspect, the present application provides a computer-readable storage medium, in which computer program instructions are stored. When the computer program instructions are executed by a processor, they are used to implement the methods provided in the second aspect and various possible designs.

[0062] The present application provides a method for predicting sales data of items in a live broadcast scenario, a training method for a model, and a medium. In the method for predicting sales data of items in a live broadcast scenario, the live broadcast resources of the live broadcast are obtained from the live broadcast server, and the item data is obtained from the platform to which the items to be recommended for live broadcast belong. Then, based on the live broadcast resources and the item data, the item data matrix corresponding to the items to be recommended for live broadcast is obtained, and the item data matrix is ​​input into the sales data prediction model for processing. After obtaining the predicted sales data of each item, the predicted sales data is output. The sales data prediction model is trained based on the sales characteristics of multiple items in different live broadcast resources and the actual sales data. In this solution, the sales data prediction model is used to perform prediction processing based on the sales characteristics of the items. Compared with manual prediction, it effectively shortens the time required for the prediction process and improves the accuracy of the predicted sales data. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1A schematic diagram of an application scenario of the method for predicting sales data of items in a live broadcast scenario provided in an embodiment of the present application;

[0064] Figure 2 A flowchart of a first embodiment of a method for predicting sales data of items in a live broadcasting scenario provided in an embodiment of the present application;

[0065] Figure 3 A flowchart of a first embodiment of a method for training a sales data prediction model for an item provided in an embodiment of the present application;

[0066] Figure 4 A flowchart of a second embodiment of a method for training a sales data prediction model for an item provided in an embodiment of the present application;

[0067] Figure 5 A flowchart of a third embodiment of a method for training a sales volume data prediction model for an item provided in an embodiment of the present application;

[0068] Figure 6 A schematic diagram of the flow of a single tree training process for a sales data prediction model provided in an embodiment of the present application;

[0069] Figure 7 A flowchart of an embodiment of a method for training a sales data prediction model for an item provided in an embodiment of the present application;

[0070] Figure 8 A schematic diagram of the structure of a device for predicting sales data of items in a live broadcasting scenario provided in an embodiment of the present application;

[0071] Fig. 9 A schematic diagram of the structure of a training device for a sales data prediction model of an item provided in an embodiment of the present application;

[0072] Fig.10 A schematic diagram of the structure of a first embodiment of an electronic device provided in an embodiment of the present application;

[0073] Fig.11 A schematic diagram of the structure of a second electronic device embodiment provided in the present application. DETAILED DESCRIPTION

[0074] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0075] Before introducing the embodiments of the present application, the background technology of the present application is first explained.

[0076] With the development of technology, the way users get the items they need in shopping malls or supermarkets has gradually been replaced by online shopping. Users can browse items on online shopping platforms, place orders for items according to their needs, fill in the address of receiving the items, and receive the items sent by the merchant through logistics at home without leaving home. With the increasing popularity of the live broadcast industry, more and more users choose to gain knowledge or entertain themselves by watching live broadcasts. Among them, in some live broadcast rooms, the anchor introduces different item information to users and recommends items suitable for them according to different user groups. Users can place orders in the live broadcast room based on the advantages and disadvantages of the items and their own needs. Since the anchor can let users have a more comprehensive understanding of the items by introducing the characteristics of the items, it effectively stimulates the user's desire to buy, thereby effectively promoting the sales data of the items. However, due to the inability to accurately predict the number of items required, when the number of items is insufficient or too many items are prepared, it may cause a loss of sales data or a backlog of goods.

[0077] In the prior art, before the live broadcast, it is necessary to predict the sales data of items during the live broadcast in order to prepare the items, which is currently mainly judged by manual experience. Specifically, before the live broadcast, the staff needs to understand the popularity of the items, the matching degree between the applicable time period and the current time period, the market saturation, etc. through investigation, and predict the sales data of the items based on various factors. However, since the staff cannot understand all the required information, the collected information is not complete, and the factors analyzed are not comprehensive enough, the staff cannot effectively process the obtained data. Not only that, since the data needs to be processed manually, the labor cost is high, and the prediction process takes a long time, the accuracy of the predicted data cannot be guaranteed.

[0078] In view of the above problems, the invention of this application is as follows: when predicting the sales data of items, in the current solution, since there is no way to manually ensure the rigor of the processing process, the accuracy of the predicted sales data cannot be guaranteed. Based on this, the inventor found that if a sales data prediction model can be obtained, the sales data prediction model can replace the manual processing process, which can solve the problem of low accuracy of the predicted sales data processed manually, thereby saving computing time and improving computing efficiency.

[0079] The following are explanations of the terms used in the embodiments of the present application:

[0080] List of smallest item units (Stock Keeping Unit, sku): sku refers to the smallest indivisible item unit for measuring inventory in and out, usually in units of pieces, boxes, pallets, etc. Each item has a unique sku number. For example, in textiles, a sku usually represents: specifications, colors, and styles. A pink (there are 3 colors: pink, yellow, and black) women's windbreaker in size M (there are 4 sizes: S, M, L, and X), in which size M and pink are a combination of a set of sku. The sku list refers to a list formed by summarizing and sorting all sku.

[0081] Cosine distance: Cosine similarity, also known as cosine similarity, evaluates the similarity between two vectors by calculating the cosine value of the angle between them. The shorter the distance, the higher the similarity.

[0082] For example, the method for predicting sales data of items in a live broadcast scene provided in the embodiment of the present application can be applied to Figure 1 In a schematic diagram of an application scenario shown. Figure 1 This is a schematic diagram of an application scenario of the method for predicting item sales data in a live broadcast scenario provided by an embodiment of the present application, which is used to solve the above technical problems. Figure 1 As shown, Figure 1 As shown, the electronic device can display the sales data prediction interface on the screen by running software, etc., so that the user can obtain the results of the sales data prediction. In the conventional processing method, the electronic device can select the data required for the sales data prediction process from the data stored in itself, or the user can input the data required for the sales data prediction, and can also download and store these data from the cloud or server. In the process of downloading these data, the cloud or server can pre-process these data, extract relevant data according to needs and send them to the electronic device for prediction processing. It is also possible that the electronic device processes according to its own needs after downloading all the relevant data and extracts the required data. After obtaining the data, the electronic device processes the obtained data according to the sales data prediction model to obtain the predicted sales data of each item.

[0083] To sum up, the electronic device that serves as the executor of the technical solution provided in this application can be any device with data processing and human-computer interaction capabilities, such as a mobile phone, a tablet computer, a computer, other smart terminals, etc., or a cloud or server and other entities with processing functions. This application does not impose too many restrictions on this.

[0084] The following specific embodiments are used to describe in detail the technical solution of the present application and how the technical solution of the present application solves the above technical problems. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0085] Figure 2 This is a flow chart of the first embodiment of the method for predicting sales data of items in a live broadcast scenario provided in the present application. Figure 2 As shown, the method for predicting sales data of items in a live broadcast scene may include the following steps:

[0086] S101: Acquire live broadcast resources for this live broadcast from a live broadcast server, and acquire item data from a platform to which the items to be recommended for live broadcast belong.

[0087] In this step, the electronic device can obtain data from the live broadcast server, process the obtained data, and parse and obtain the live broadcast resource. Specifically, the electronic device can send a command to obtain the live broadcast resource to the live broadcast server. After receiving the command, the live broadcast server sends the live broadcast resource to the electronic device, so that the electronic device can process it. The live broadcast server can also actively send the live broadcast resource to the electronic device, and the electronic device receives the live broadcast resource sent by the live broadcast server and processes it.

[0088] Furthermore, the electronic device can obtain the item data from the platform to which the item to be recommended for live broadcast belongs. Specifically, the electronic device can send an instruction to obtain the item data to the platform to which the item to be recommended for live broadcast belongs. After receiving the instruction, the live broadcast server sends the item data to the electronic device to facilitate the electronic device to process it. The platform can also actively send the item data to the electronic device, and the electronic device receives the item data of the platform and processes it.

[0089] Specifically, the item data and live broadcast resources can be carried in the data received by the electronic device from the outside. The data can be a real-time message queue (MQ) message, or it can be automatically captured from the World Wide Web by a web crawler according to a certain rule. The embodiment of the present application does not limit the specific form of the data, which can be determined according to actual needs.

[0090] Among them, live broadcast resources are mainly used to represent the live broadcast information when recommending items, including live broadcast time and / or live broadcast entry resource information. Item data is mainly used to represent the information of items to be recommended for live broadcast, including: the name, category, purpose, sales price in at least one historical time period, sales data, traffic, number of purchases, number of add-to-carts, and price on competing platforms.

[0091] S102: According to the live broadcast resources and the item data, an item data matrix corresponding to the items to be recommended for live broadcast is obtained.

[0092] In this step, the electronic device obtains the live broadcast resources and item data of the multiple items to be recommended for live broadcast, analyzes and processes them according to the live broadcast resources and the item data of each item, obtains the sales characteristics of each item, and obtains the item data matrix according to the sales characteristics of each item. The item data matrix is ​​a matrix determined according to the item data of multiple items and the live broadcast resources, and is mainly used to represent the sales characteristics of each item.

[0093] The sales characteristics of each item include: basic information of the item, cross-correlation coefficient between price and sales data, statistical characteristics of sales data, statistical characteristics of price, ratio information of sales data to median of sales data, and live broadcast resources. It should be understood that the sales characteristics of each item mentioned above may also include other sales characteristics of the item, which is not limited in the embodiments of the present application.

[0094] Specifically, the basic information of an item includes the name, category, and purpose of the item, and may also include other item information, which is not specifically limited in the present embodiment. The cross-correlation coefficient between price and sales data is mainly used to characterize the cross-correlation between item price and sales data, and can be calculated using formula (1).

[0095] corr=cov(X,Y) / √(Var(X)Var(Y)) (1)

[0096] Among them, cov(X,Y) represents covariance, Var represents covariance, X represents sales data, and Y represents price.

[0097] The statistical characteristics of sales data include the mean, maximum, minimum, variance, median, regression coefficient, maximum position, difference sum, autocorrelation coefficient and wavelet coefficient of the sales data of the historical 1 week, 1 month and 2 months, and the historical same period of 1 month and 2 months.

[0098] The price statistical features include the mean, maximum, minimum, variance, median, regression coefficient, maximum position, difference sum, autocorrelation coefficient and wavelet coefficient of the prices in the historical period of 1 week, 1 month and 2 months, and the same period of 1 month and 2 months.

[0099] The sales data and the median sales data ratio information include the ratio of single-day sales data in the past 1 to 7 days to the median sales data in the history of 1 month.

[0100] For example, the item data matrix is ​​determined based on the item data of multiple items and live broadcast resources, and is mainly used to represent the sales characteristics of each item. It can be represented by the following matrix (a).

[0101] Matrix (a):

[0102]

[0103] Among them, in matrix (a), baseinfo represents the basic information of the item, saleprice_cross represents the cross-correlation coefficient of price and sales data, saleTimeseries represents the statistical characteristics of sales data, priceTimeseries represents the statistical characteristics of price, priceTimeseries represents the ratio information of sales data and the median of sales data, and priceTimeseries represents live broadcast resources.

[0104] It should be understood that, in addition to the above-mentioned matrix format, the sales characteristics of each item can also be represented in other formats such as lists and sets, and the embodiments of the present application are not limited to this.

[0105] S103: Input the item data matrix into the pre-acquired sales data prediction model for processing to obtain predicted sales data for each item.

[0106] In this solution, the sales data prediction model is a prediction model trained based on the sales characteristics of multiple items in different live broadcast resources and actual sales data. It is a model used to predict the sales data of each item based on a matrix representing the sales characteristics of each item.

[0107] In this step, after the electronic device obtains the item data matrix, it inputs the item data matrix into the sales data prediction model, and the sales data prediction model analyzes and processes the item data matrix to obtain the predicted sales data of each item. The predicted sales data refers to the sales data of the item predicted by the sales data prediction model.

[0108] Exemplarily, the item data matrix can be input into the sales data prediction model through the input box of the sales data prediction interface for processing, or the item data matrix can be automatically obtained by the system of the electronic device and input into the sales data prediction model for processing. The sales data prediction model analyzes and processes the input item data matrix to obtain predicted sales data for each item.

[0109] Optionally, the item data matrix can be saved in a document, and the document can be imported into the sales data prediction interface through the selection control of the sales data prediction interface. The sales data prediction model receives and reads the document, obtains the item data matrix carried in the document, and obtains the predicted sales data of each item by processing the item data matrix.

[0110] The embodiment of the present application does not limit the specific method of obtaining the item data matrix by the sales data prediction model, which can be determined according to actual needs.

[0111] S104: Output forecast sales data.

[0112] In this step, the electronic device obtains the predicted sales data of the item through the sales data prediction model, and displays the predicted sales data on the sales data prediction interface.

[0113] Furthermore, after the electronic device predicts the sales volume forecast of each item, it can also output a stocking instruction according to the existing quantity of each item and the sales volume forecast. The stocking instruction is used to indicate the replenishment quantity of each item, and remind the staff to prepare the stock according to the stocking instruction.

[0114] For example, the replenishment quantity can be calculated by using formula (2) through conventional target turnover inventory, predicted sales data and existing inventory.

[0115] H=A+sales live -B (2)

[0116] Among them, H represents the replenishment quantity, A represents the regular target turnover inventory, and sales live A represents predicted sales data, and B represents existing inventory.

[0117] The method for predicting sales data of items in a live broadcast scenario provided in an embodiment of the present application is achieved by obtaining the live broadcast resources of the live broadcast from the live broadcast server, and obtaining the item data from the platform to which the items to be recommended for live broadcast belong. Then, based on the live broadcast resources and the item data, the item data matrix corresponding to the items to be recommended for live broadcast is obtained, and the item data matrix is ​​input into the sales data prediction model for processing. After obtaining the predicted sales data of each item, the predicted sales data is output. The sales data prediction model is trained based on the sales characteristics of multiple items in different live broadcast resources and the actual sales data. According to the sales characteristics of the items, the sales data prediction model is used for prediction processing, which effectively shortens the time required for the prediction process and improves the accuracy of the predicted sales data compared to manual prediction.

[0118] On the basis of the above embodiment, in order to ensure the accuracy of the predicted sales data of each item obtained using the sales data prediction model, the sales data prediction model needs to be trained to reduce the error between the predicted sales data and the actual sales data, which is beneficial for the staff to pre-stock the items.

[0119] Figure 3 This is a flow chart of the first embodiment of the training method of the sales data prediction model of the article provided in the embodiment of the present application. Figure 3 As shown, the training method of the sales data prediction model of the item may include the following steps:

[0120] S201: Acquire multiple target hosts whose similarity with the current host reaches a preset value according to the personal information of the current host and the item data of historical live broadcast items.

[0121] In this embodiment, since the number of items broadcast live by the current anchor is limited, if only the items broadcast live by the current anchor are analyzed, the sales data prediction model trained based on the item data of the items broadcast live by the current anchor may have low accuracy. Therefore, it is necessary to obtain a target anchor similar to the current anchor, and use the item data of the target anchor's historical live items to train the sales data prediction model of the item, so as to achieve the purpose of improving the accuracy of the sales data prediction model.

[0122] Specifically, the electronic device can obtain the required data from the data stored in itself, and then analyze it to obtain the personal information of the current anchor and the item data of the historical live broadcast items. The electronic device can also obtain data from the cloud or server, process the obtained data, and analyze it to obtain the personal information of the current anchor and the item data of the historical live broadcast items.

[0123] In this step, the electronic device obtains the personal information of the current anchor and the item data of the historical live broadcast items, processes and analyzes the obtained data, and obtains multiple target anchors whose similarity with the current anchor reaches a preset value.

[0124] Among them, the personal information of the current anchor mainly includes the anchor's occupation (for example, actor, businessman, professional live streamer, manufacturer live streamer, etc.), the anchor's basic information (for example, age, gender, region, popularity, etc.). It should be understood that the personal information of the current anchor may also include other personal information, and this plan does not restrict this.

[0125] S202: Obtain item sales data prediction training set.

[0126] In this step, after obtaining multiple target anchors, the electronic device obtains the item sales data prediction training set according to the live broadcast resources and sku lists corresponding to the target anchors. In other words, the item sales data prediction training set includes the live broadcast resources and sku lists of each target anchor.

[0127] The live broadcast resources include the live broadcast time and / or live broadcast portal resource information. The item data of each item includes: the name, category, purpose, sales price in at least one historical time period, sales data, traffic, number of purchases, number of add-on purchases, and price on competing platforms.

[0128] Exemplarily, the sku list corresponding to each target anchor includes the item data of each item in the target anchor's historical live broadcast. The item data can be obtained from data stored in an electronic device, or from data stored in a cloud or server. The embodiment of the present application does not limit the method of obtaining the item data of each item in the target anchor's historical live broadcast.

[0129] Exemplarily, the electronic device obtains the item data of each item in the historical live broadcasts of multiple target anchors, and processes the data to form a corresponding sku list. When the electronic device cannot obtain all the item data constituting the sku list, it is necessary to supplement the missing item data in the sku list. The missing item data can be supplemented to the sku list manually by the staff, or the missing item data in the required sku list in the World Wide Web can be automatically captured by a web crawler according to a certain rule and filled into the sku list to form a complete sku list.

[0130] S203: Perform model training according to the sales data prediction training set to obtain a sales data prediction model.

[0131] In this step, since the sales data prediction training set has been obtained through multiple target anchors in the above steps, the electronic device needs to train the sales data prediction training set multiple times so that the sales data prediction model obtained has higher accuracy.

[0132] After obtaining the sales data prediction training set, the electronic device conducts multiple training on it. The training target is the sales data of each live broadcast, and the sales data is standardized during training. An effective standardization method is to divide the sales data of the day by the median of the historical 60 days, and the prediction training target becomes the multiple of the live broadcast relative to the daily sales data. After the prediction, the sales data is restored by multiplying it by the median of the historical 60 days. The sales data prediction model is obtained based on the training results, so that the user can use the sales data prediction model to process the predicted sales data of each item according to the item data matrix.

[0133] The training method of the sales data prediction model of the item provided in this embodiment obtains multiple target anchors whose similarity with the current anchor reaches a preset value based on the personal information of the current anchor and the item data of the historical live broadcast items, and obtains the item sales data prediction training set based on the multiple target anchors. Model training is performed based on the sales data prediction training set to obtain a sales data prediction model. The item sales data prediction training set is obtained through multiple target anchors, and the sales data prediction model is obtained by training based on the item sales data prediction training set, so that the sales data prediction model is more accurate, which lays a foundation for the subsequent application of the sales data prediction model to obtain the predicted sales data of each item, and improves the accuracy of subsequent prediction work.

[0134] Exemplarily, based on the above embodiments, Figure 4 This is a flow chart of the second embodiment of the training method of the sales data prediction model of the article provided in the embodiment of the present application. Figure 4 As shown, the above S201 can be implemented by the following steps:

[0135] S301: construct an information model of the current anchor according to the personal information of the current anchor and the item data of historical live broadcast items.

[0136] In an embodiment of the present application, in order to identify multiple target hosts that are similar to the current host, the electronic device first needs to construct an information model of the current host.

[0137] In this step, the electronic device needs to obtain the personal information of the current live broadcast and the item data of the historical live broadcast items, and process the obtained data to obtain the information model of the current anchor.

[0138] Exemplarily, the information model of the current anchor can be represented by the following matrix (b).

[0139] [B] = [name cate sales price ...] Matrix (b)

[0140] Among them, B represents the name of the current anchor, name represents the name of the item, cate represents the category of the item, sales represents the sales data of the item, and price represents the price of the item.

[0141] It should be understood that in addition to the above-mentioned matrix format, the information model of the current anchor may also be represented by other matrix formats, which is not limited in the embodiments of the present application.

[0142] S302: Obtain information models of multiple anchors.

[0143] In this step, the electronic device obtains the personal information of multiple anchors and the item data of historical live broadcast items, processes and analyzes the obtained data, and obtains the information models of the multiple anchors.

[0144] Among them, the personal information of multiple anchors mainly includes the anchor's occupation (for example, actor, businessman, professional live streamer, manufacturer live streamer, etc.), the anchor's basic information (for example, age, gender, region, popularity, etc.). It should be understood that the personal information of multiple anchors may also include other personal information, and the embodiments of the present application do not limit this.

[0145] Exemplarily, the information model of multiple anchors can be represented by the following matrix (c).

[0146]

[0147] Among them, A represents the names of multiple anchors, name represents the name of the item, cate represents the category of the item, sales represents the sales data of the item, and price represents the price of the item.

[0148] It should be understood that in addition to the above-mentioned matrix format, the information model of multiple anchors can also be represented by other matrix formats, which is not limited in the embodiments of the present application.

[0149] S303: Determine multiple target anchors from other multiple anchors according to the information model of the current anchor and the information models of multiple anchors.

[0150] In this step, the electronic device has obtained the information model of the current anchor and the information models of multiple anchors according to the above steps, and can determine multiple target anchors by comparing the two. Among them, the target anchor refers to an anchor whose similarity with the current anchor reaches a preset value among the multiple anchors.

[0151] Exemplarily, multiple target anchors may be determined by calculating the cosine distance between the information model of the current anchor and the information models of multiple anchors. Specifically, the cosine distance may be calculated using the following formula (3).

[0152]

[0153] Among them, A represents multiple anchors, and B represents the current anchor.

[0154] In a specific implementation, a preset cosine distance may be preset, and the preset cosine distance is a specific value, such as 1, 2, 3, etc., and the present application does not impose specific restrictions on this. The calculated cosine distance between the information model of the current anchor and the information models of multiple anchors is compared with the preset cosine distance. When the cosine distance is less than the preset cosine distance, the anchor is considered to be a target anchor similar to the current anchor.

[0155] The number of target anchors is less than or equal to 100. If the number of anchors selected according to the above method is greater than 100, 100 anchors with the smallest cosine distances are selected as target anchors according to the cosine distances of these anchors.

[0156] In another specific implementation, a target number of anchors may be preset, and the target number of anchors is a specific value less than or equal to 100, such as 1, 2, 3, etc., and the present application does not impose any specific restrictions on this. The cosine distance between the information model of the current anchor and the information models of multiple anchors is calculated, and the anchor with the smallest cosine distance equal to the target number of anchors is used as the target anchor.

[0157] The training method of the sales data prediction model of an item provided in this embodiment constructs an information model of the current anchor based on the personal information of the current anchor and the item data of the historical live broadcast items. The information models of multiple anchors are obtained, and multiple target anchors are determined from multiple other anchors based on the information model of the current anchor and the information models of multiple anchors. By comparing the information model of the current anchor and the information models of multiple anchors, multiple target anchors can be confirmed, which lays the foundation for confirming the item sales data prediction training set based on the target anchors.

[0158] Exemplarily, based on the above embodiments, Figure 5 This is a flow chart of the third embodiment of the training method of the sales data prediction model of the article provided in the embodiment of the present application. Figure 5 As shown, the above S202 can be implemented by the following steps:

[0159] S401: Predict the live broadcast resources in the training set and the item data of each item according to the sales data to obtain the item feature matrix.

[0160] In this step, the electronic device obtains the sales data prediction training set, and analyzes and processes the live broadcast resources and the item data of each item therein to obtain an item feature matrix. The live broadcast resources include the live broadcast time and / or live broadcast entry resource information, and the item data of each item includes: the name, category, purpose, sales price in at least one historical time period, sales data, traffic, number of purchases, number of add-on purchases, and price on competing product platforms.

[0161] Specifically, the electronic device obtains the sales characteristics of each item based on the live broadcast resources and the item data of each item, and obtains the item data matrix based on the sales characteristics of each item. The item data matrix is ​​a matrix determined based on the item data of multiple items and the live broadcast resources, and is mainly used to represent the sales characteristics of each item.

[0162] Among them, the sales characteristics of each item include: basic information of the item, cross-correlation coefficient between price and sales data, statistical characteristics of sales data, statistical characteristics of price, ratio information of sales data and median of sales data, and live broadcast resources.

[0163] S402: Model training is performed based on the item feature matrix and the actual sales data of each item to obtain a sales data prediction model.

[0164] In this step, the electronic device is trained based on the acquired item feature matrix and the actual sales data of each item. The actual sales data of each item is mainly used to compare with the training results to verify the accuracy of the model. After multiple trainings, a sales data prediction model is obtained.

[0165] Specifically, training can be performed using a gradient descent tree regression method, or using a least squares method, Newton's method, or the like, which is not limited in this embodiment of the present application.

[0166] For example, this solution uses the gradient descent tree regression method to train the model. M features are randomly selected in the matrix (a) to find the best cut point. The best cut point refers to the continuous variable such as [1,100]. During the training process, the cut point in this interval must be found to minimize the loss function, that is, to find a value between [1,100] that minimizes the loss function. Among them, the mean square error (MSE) is used as an example. At the tth tree, the form can be expressed by formula (4).

[0167]

[0168] in, Represents the actual sales data of items in t trees, Represents the predicted sales data of items in t trees, represents the predicted sales data of the i-th item in the t-1 tree, f t (x i ) represents the fitted sales data of the items in the tth tree.

[0169] Through Taylor's second-order expansion, that is, formula (5): By transforming formula (4), formula (4) can be converted into formula (5).

[0170]

[0171] Among them, g i Represents the first-order derivative of the loss function, h i Represents the second-order derivative of the loss function.

[0172] Repeat the above process, and the electronic device will calculate and obtain the predicted sales data of the items in the prediction training set, which will gradually approach the true value as the training progresses. At the same time, the sales data prediction training set data is used to supervise the fitting of the sales data prediction model. If the model keeps approaching the true value during the continuous iteration process and the prediction effect on the sales data prediction training set does not deteriorate, the training can continue. When the prediction effect of the sales data prediction training set data no longer improves, it means that the training can be stopped and the model training is completed.

[0173] Specifically, the trained sales data prediction training set can be expressed by formula (6).

[0174]

[0175] in, Represents the trained sales data prediction model, sales live Represents the predicted sales data, inputData predic Represents the item data matrix.

[0176] Figure 6 A flowchart of a single tree training process for a sales data prediction model provided in an embodiment of the present application. Figure 6 As shown, first, the item data matrix is ​​input into the sales data prediction model. The sales data prediction model analyzes and processes the item data matrix, and enters step 1 to determine whether the item is on sale. If there is a promotion, step 2 is executed to determine whether there is a full discount; if not, step 3 is executed to determine whether there is a gift. In step 2, if the item is on sale, step 4 is entered to determine whether it meets the full 199-100; if not, step 5 is entered to analyze the full discount strength of the product. For step 4, if the item meets the full 199-100, the predicted sales data of the items in the prediction training set are output as 13; if not, the predicted sales data of the items in the prediction training set are output as 5. The predicted sales data of the items in the output prediction training set are compared with the data in the prediction training set to determine the training status of the model. If the predicted sales data is consistent with the data in the prediction training set, and the model is constantly approaching the true value during the continuous iteration process, the training can continue. When the prediction effect of the sales data prediction training set data no longer improves, it means that the training can be stopped and the model training is completed.

[0177] Figure 7 The flowchart of an example of a method for training a sales data prediction model for an item provided in an embodiment of the present application is shown in FIG. Figure 7 As shown:

[0178] Step 1: Build the information model of the current anchor and multiple anchors based on the anchor's personal information and the item data of historical live broadcast items.

[0179] Step 2: Obtain a data matrix that can represent the information model of the current anchor by obtaining the information model of the current anchor, and proceed to step 3.

[0180] Step 3: Get multiple target anchors whose similarity with the current anchor reaches a preset value. Then proceed to step 4.

[0181] Step 4: Get the live broadcast resources and sku lists of multiple target anchors, and proceed to step 5 based on the acquired data.

[0182] Step 5: Create a training set for item sales data prediction.

[0183] Step 6: Perform training based on the item sales data prediction training set to obtain the sales data prediction model.

[0184] Step 7: Store the sales data prediction model. In actual use, proceed to step 8.

[0185] Step 8: Read the sales data prediction model.

[0186] In step 9, the electronic device obtains the item data matrix corresponding to the items to be recommended in the live broadcast through the information model of the current anchor, inputs the item data matrix into the sales data prediction model for processing, predicts the sales data of the items to be recommended in the live broadcast, and obtains the predicted sales data of each item.

[0187] Finally, in step 10, stock items based on predicted sales data.

[0188] The training method of the sales data prediction model of an item provided in this embodiment obtains an item feature matrix based on the live broadcast resources and item data of each item in the sales data prediction training set. This enables training to be performed based on the item feature matrix and the actual sales data of each item to obtain a sales data prediction model, thereby ensuring the accuracy of the model and laying the foundation for confirming the item sales data prediction training set based on the target anchor.

[0189] In combination with the above embodiments, in the implementation of the technical solution provided in the present application, it should be understood that the electronic device for sales data prediction and the electronic device for sales data prediction model training of items can be the same electronic device or different electronic devices. That is to say, the model training process can be trained on other devices and prediction can be performed on the electronic device that needs to be used. This application does not impose any restrictions on this.

[0190] The following is an embodiment of the device of the present application, which can be used to execute the embodiment of the method of the present application. For details not disclosed in the embodiment of the device of the present application, please refer to the embodiment of the method of the present application.

[0191] Figure 8 This is a schematic diagram of the structure of the device for predicting item sales data in a live broadcast scene provided in an embodiment of the present application. Figure 8 As shown, the device includes: an acquisition module 81, a processing module 82 and an output module 83;

[0192] The acquisition module 81 is used to acquire the live broadcast resources of the current live broadcast from the live broadcast server, and acquire the item data from the platform to which the items to be recommended for live broadcast belong. The live broadcast resources include the live broadcast time and / or live broadcast entry resource information;

[0193] The acquisition module 81 is further used to acquire the item data matrix corresponding to the items to be recommended for live broadcast according to the live broadcast resources and the item data; the item data matrix is ​​used to represent the sales characteristics of each item to be recommended for live broadcast;

[0194] The processing module 82 is used to input the item data matrix into the pre-acquired sales data prediction model for processing to obtain the predicted sales data of each item; wherein the sales data prediction model is a prediction model trained based on the sales characteristics of multiple items in different live broadcast resources and actual sales data;

[0195] The output module 83 is used to output the predicted sales data.

[0196] In a possible design of an embodiment of the present application, the items to be recommended for live broadcast include multiple items, and the item data of each item includes: the name of the item, category, purpose, sales price in at least one historical time period, sales data, traffic, number of purchases, number of add-to-carts, and price on a competing product platform;

[0197] The acquisition module 81 is specifically used for:

[0198] Obtain the live broadcast resources of this live broadcast and the item data of all items to be recommended for live broadcast; the live broadcast resources include the live broadcast time and / or live broadcast portal resource information, and the item data of each item includes: the name of the item, category, purpose, sales price in at least one historical time period, sales data, traffic, number of purchases, number of add-to-carts, and price on competing platforms;

[0199] According to the live broadcast resources and the item data of each item, the sales characteristics of each item are obtained, and the item data matrix is ​​obtained according to the sales characteristics of each item;

[0200] Among them, the sales characteristics of each item include: basic information of the item, cross-correlation coefficient between price and sales data, statistical characteristics of sales data, statistical characteristics of price, ratio information of sales data and median of sales data, and live broadcast resources.

[0201] In another possible design of the embodiment of the present application, the processing module 82 is also used to output a stocking indication based on the existing quantity of each item and the predicted sales data; the stocking indication is used to indicate the replenishment quantity of each item.

[0202] The device provided in the embodiment of the present application can be used to execute the method for predicting sales data of items in the live broadcast scene in the above-mentioned embodiment. Its implementation principle and technical effects are similar and will not be repeated here.

[0203] Fig. 9 This is a schematic diagram of the structure of a training device for a sales data prediction model of an item provided in an embodiment of the present application. Fig. 9 As shown, the device includes: an acquisition module 91 and a training module 92;

[0204] The acquisition module 91 is used to acquire multiple target anchors whose similarity with the current anchor reaches a preset value according to the personal information of the current anchor and the item data of the historical live broadcast items;

[0205] The acquisition module 91 is also used to acquire a training set for predicting item sales data, wherein the training set for predicting item sales data includes: live broadcast resources and sku lists of each target anchor, and the sku list corresponding to each target anchor includes item data of each item in the historical live broadcast of the target anchor;

[0206] The training module 92 is used to perform model training based on the sales data prediction training set to obtain a sales data prediction model.

[0207] In a possible design of the embodiment of the present application, the acquisition module 91 is specifically used to:

[0208] According to the personal information of the current anchor and the item data of the historical live broadcast items, an information model of the current anchor is constructed;

[0209] Get information models of multiple anchors;

[0210] According to the information model of the current anchor and the information models of multiple anchors, multiple target anchors are determined from the other multiple anchors; wherein the cosine distance between the information model of each target anchor and the information model of the current anchor is less than a preset value.

[0211] In another possible design of the embodiment of the present application, the training module 92 is specifically used for:

[0212] According to the sales data, the live broadcast resources in the training set and the item data of each item are predicted to obtain the item feature matrix; the item data matrix is ​​a matrix used to represent the sales characteristics of each item;

[0213] The model is trained based on the item feature matrix and the actual sales data of each item to obtain a sales data prediction model.

[0214] Optionally, in the sales data prediction training set, the live broadcast resources include the live broadcast time and / or live broadcast portal resource information, and the item data of each item includes: the name, category, purpose, sales price in at least one historical time period, sales data, traffic, number of purchases, number of add-to-carts, and price on competing platforms;

[0215] Accordingly, the acquisition module 91 is specifically used for:

[0216] According to the live broadcast resources and the item data of each item, the sales characteristics of each item are obtained, and the item data matrix is ​​obtained according to the sales characteristics of each item;

[0217] Among them, the sales characteristics of each item include: basic information of the item, cross-correlation coefficient between price and sales data, statistical characteristics of sales data, statistical characteristics of price, ratio information of sales data and median of sales data, and live broadcast resources.

[0218] The device provided in the embodiment of the present application can be used to execute the training method of the sales data prediction model of the items in the above-mentioned embodiment. Its implementation principle and technical effect are similar and will not be repeated here.

[0219] It should be noted that it should be understood that the division of the various modules of the above device is only a division of logical functions. In actual implementation, they can be fully or partially integrated into one physical entity, or they can be physically separated. And these modules can all be implemented in the form of software called by processing elements; they can also be all implemented in the form of hardware; some modules can also be implemented in the form of software called by processing elements, and some modules can be implemented in the form of hardware. For example, the determination module can be a separately established processing element, or it can be integrated in a chip of the above device. In addition, it can also be stored in the memory of the above device in the form of program code, and called and executed by a processing element of the above device. The implementation of other modules is similar. In addition, these modules can be fully or partially integrated together, or they can be implemented independently. The processing element here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each module above can be completed by an integrated logic circuit of hardware in the processor element or instructions in the form of software.

[0220] Fig.10 This is a schematic diagram of the structure of the electronic device embodiment 1 provided in the present application. Fig.10 As shown, the electronic device may include: a processor 101, a memory 102, and computer program instructions stored in the memory 102 and executable on the processor 101. When the processor 101 executes the computer program instructions, the method for predicting sales data of items in a live broadcast scene provided in any of the aforementioned embodiments is implemented.

[0221] Optionally, the electronic device may further include an interface for interacting with other devices.

[0222] Optionally, the above-mentioned components of the electronic device may be connected via a system bus.

[0223] The memory 102 may be a separate storage unit or a storage unit integrated in the processor 101. The number of the processor 101 is one or more.

[0224] The electronic device provided in the embodiment of the present application can be used to execute the method for predicting sales data of items in a live broadcast scene provided in any of the above-mentioned method embodiments. Its implementation principle and technical effects are similar and will not be repeated here.

[0225] Fig.11 This is a schematic diagram of the structure of the electronic device embodiment 2 provided in the present application. Fig.11 As shown, the electronic device may include: a processor 111, a memory 112, and computer program instructions stored in the memory 112 and executable on the processor 111. When the processor 111 executes the computer program instructions, the training method of the sales data prediction model of the item provided by any of the aforementioned method embodiments is implemented.

[0226] Optionally, the electronic device may further include an interface for interacting with other devices.

[0227] Optionally, the above-mentioned components of the electronic device may be connected via a system bus.

[0228] The memory 112 may be a separate storage unit or a storage unit integrated in the processor 11. The number of the processor 111 is one or more.

[0229] The electronic device provided in the embodiment of the present application can be used to execute the training method of the sales data prediction model of the item provided in any of the above-mentioned method embodiments. Its implementation principle and technical effects are similar and will not be repeated here.

[0230] In the specific implementation of any of the above electronic devices, it should be understood that the processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the present application can be directly implemented as a hardware processor, or can be implemented by a combination of hardware and software modules in the processor.

[0231] The system bus may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus. The system bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus. The memory may include random access memory (RAM) and may also include non-volatile memory (non-volatile memory), such as at least one disk storage.

[0232] All or part of the steps of the above-mentioned method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a readable memory. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the aforementioned memory (storage medium) includes: read-only memory (English: read-only memory, abbreviated as: ROM), RAM, flash memory, hard disk, solid state drive, magnetic tape (English: magnetic tape), floppy disk (English: floppy disk), optical disc (English: optical disc) and any combination thereof.

[0233] An embodiment of the present application provides a computer-readable storage medium, in which computer instructions are stored. When the computer instructions are executed on a computer, the computer executes the above-mentioned method for predicting sales data of items in a live broadcast scene.

[0234] An embodiment of the present application also provides a computer-readable storage medium, in which computer instructions are stored. When the computer instructions are executed on a computer, the computer executes the training method of the sales data prediction model of the above-mentioned item.

[0235] The computer-readable storage medium mentioned above 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 memory, flash memory, magnetic disk or optical disk. The readable storage medium can be any available medium that can be accessed by a general or special-purpose computer.

[0236] Optionally, a readable storage medium is coupled to a processor so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can be located in an application specific integrated circuit (ASIC). Of course, the processor and the readable storage medium can also exist in the device as discrete components.

[0237] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A method for predicting sales data of items in a live broadcast scene, characterized in that: include: Obtain the live broadcast resources of this live broadcast from the live broadcast server, and obtain the item data of each item from the platforms to which the multiple items to be recommended for live broadcast belong, wherein the live broadcast resources include the live broadcast time and / or live broadcast portal resource information; the item data includes: the name, category, purpose, sales price, sales data, traffic, number of purchases, number of add-to-carts, and price on competing platforms of the item; According to the live broadcast resource and the item data of each item in the multiple items to be recommended for live broadcast, the sales characteristics of each item are obtained, and an item data matrix is ​​obtained according to the sales characteristics of each item; the item data matrix is ​​a matrix used to represent the sales characteristics of each item to be recommended for live broadcast; the sales characteristics include: basic information of the item, cross-correlation coefficient between price and sales data, statistical characteristics of sales data, statistical characteristics of price, ratio information of sales data to median of sales data, and live broadcast resources; The item data matrix is ​​input into the pre-acquired sales data prediction model for processing to obtain predicted sales data for each item; wherein the sales data prediction model is a prediction model trained based on the sales characteristics of multiple items in different live broadcast resources and actual sales data; Output forecast sales data; The sales data prediction model is trained in the following way: According to the personal information of the current anchor and the item data of the historical live broadcast items, multiple target anchors whose similarity with the current anchor reaches a preset value are obtained; a training set of item sales data prediction is obtained, wherein the training set of item sales data prediction includes: the live broadcast resources and sku list of each target anchor, and the sku list corresponding to each target anchor includes the item data of each item in the historical live broadcast of the target anchor; model training is performed according to the sales data prediction training set to obtain the sales data prediction model.

2. The method according to claim 1, characterized in that The method further comprises: Output a stocking instruction based on the existing quantity of each item and the predicted sales data; the stocking instruction is used to indicate the replenishment quantity of each item.

3. The method according to claim 1, characterized in that The step of acquiring, based on the personal information of the current anchor and the item data of historical live broadcast items, a plurality of target anchors whose similarity with the current anchor reaches a preset value includes: According to the personal information of the current anchor and the item data of the historical live broadcast items, an information model of the current anchor is constructed; Get information models of multiple anchors; According to the information model of the current anchor and the information models of the multiple anchors, the multiple target anchors are determined from the other multiple anchors; wherein the cosine distance between the information model of each target anchor and the information model of the current anchor is less than a preset value.

4. The method according to any one of claims 1 to 3, characterized in that: The performing model training according to the sales data prediction training set to obtain the sales data prediction model includes: According to the sales data, the live broadcast resources in the training set and the item data of each item are predicted to obtain an item feature matrix; the item data matrix is ​​a matrix used to represent the sales characteristics of each item; Model training is performed based on the item feature matrix and the actual sales data of each item to obtain the sales data prediction model.

5. The method according to claim 4, characterized in that In the sales data prediction training set, the live broadcast resources include live broadcast time and / or live broadcast portal resource information, and the item data of each item includes: the name, category, purpose, sales price in at least one historical time period, sales data, traffic, number of purchases, number of add-to-carts, and price on competing platforms of the item; Accordingly, the live broadcast resources and item data of each item in the training set are predicted based on the sales data to obtain an item feature matrix, including: According to the live broadcast resources and the item data of each item, the sales characteristics of each item are obtained, and the item data matrix is ​​obtained according to the sales characteristics of each item.

6. A device for predicting sales data of items in a live broadcast scene, comprising: Acquisition module, processing module, output module and training module; The acquisition module is used to acquire the live broadcast resources of the live broadcast from the live broadcast server, and acquire the item data of each item from the platforms to which the multiple items to be recommended for live broadcast belong, wherein the live broadcast resources include the live broadcast time and / or live broadcast portal resource information; the item data includes: the name, category, purpose, sales price, sales volume data, traffic, number of purchases, number of add-to-carts, and price on competing platforms of the item; The acquisition module is further used to acquire the sales characteristics of each item according to the live broadcast resources and the item data of each item in the multiple items to be recommended for live broadcast, and obtain an item data matrix according to the sales characteristics of each item; the item data matrix is ​​used to represent the matrix of sales characteristics of each item to be recommended for live broadcast; the sales characteristics include: basic information of the item, cross-correlation coefficient between price and sales data, statistical characteristics of sales data, statistical characteristics of price, ratio information of sales data to median of sales data, and live broadcast resources; The processing module is used to input the item data matrix into the pre-acquired sales data prediction model for processing to obtain the predicted sales data of each item; wherein the sales data prediction model is a prediction model trained based on the sales characteristics of multiple items in different live broadcast resources and actual sales data; The output module is used to output the predicted sales data; The acquisition module is further used to acquire multiple target anchors whose similarity with the current anchor reaches a preset value based on the personal information of the current anchor and the item data of the historical live broadcast items; The acquisition module is further used to acquire a training set for predicting item sales data, wherein the training set for predicting item sales data includes: live broadcast resources and sku lists of each target anchor, and the sku list corresponding to each target anchor includes item data of each item that was historically live broadcasted by the target anchor; The training module is used to perform training based on the sales data prediction training set to obtain the sales data prediction model.

7. An electronic device, characterized in that: include: A processor, a memory, and computer program instructions stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 5 when executing the computer program instructions.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 5 when executed by a processor.

Images