An information processing method and device, equipment, and storage medium

Abstract

The application discloses an information processing method, which comprises the following steps: inputting an object feature of a target object into a first network model to obtain a first output vector output by the first network model, wherein the first output vector represents a supply degree of the target object; inputting a user input feature of the target object into a second network model to obtain a second output vector output by the second network model, wherein the second output vector represents a search degree of the target object by the user; and determining supply and demand state positioning information of the target object according to the first output vector and the second output vector, wherein the supply and demand state positioning information represents a predicted supply and demand relationship of the target object. In addition, the application further discloses an information processing device, an electronic equipment and a storage medium.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of information processing, and relates to but is not limited to an information processing method and device, equipment, and a storage medium. BACKGROUND

[0002] When determining the supply and demand of a certain commodity (such as: commodity A), a data analyst needs to analyze the search data of users searching for the commodity A and the supply data of a supply platform for the commodity A to determine the demand of the users for the commodity A and the supply of the commodity A. When the analysis is performed by the data analyst, for other commodities (such as: commodity B), the data analyst needs to perform analysis again, which will cause the problem of low efficiency of determining the supply and demand of commodities. SUMMARY

[0003] Embodiments of the present application provide an information processing method and device, and a storage medium to solve at least one problem in the related art, and can improve the analysis efficiency of analyzing the demand and supply of commodities.

[0004] The technical solution of the present application is implemented as follows:

[0005] In a first aspect, an embodiment of the present application provides an information processing method, which includes:

[0006] inputting an object feature of a target object into a first network model to obtain a first output vector output by the first network model, the first output vector representing a supply degree of the target object;

[0007] inputting a user input feature for the target object into a second network model to obtain a second output vector output by the second network model, the second output vector representing a search degree of the user for the target object;

[0008] determining supply and demand state positioning information of the target object according to the first output vector and the second output vector, the supply and demand state positioning information representing a predicted supply and demand relationship of the target object.

[0009] In a second aspect, an embodiment of the present application provides an information processing device, which includes:

[0010] a first determination unit configured to input an object feature of a target object into a first network model to obtain a first output vector output by the first network model, the first output vector representing a supply degree of the target object;

[0011] a second determining unit, configured to input a user input feature for the target object into a second network model to obtain a second output vector output by the second network model, the second output vector representing a search degree of the target object by the user;

[0012] a third determining unit, configured to determine supply-demand state positioning information of the target object according to the first output vector and the second output vector, the supply-demand state positioning information representing a predicted supply-demand relationship of the target object.

[0013] In a third aspect, an embodiment of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the information processing method when executing the computer program.

[0014] In a fourth aspect, an embodiment of the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executable on a processor to implement the information processing method.

[0015] The present application provides an information processing method and device, and a storage medium. An information processing device inputs an object feature of a target object into a first network model to obtain a first output vector output by the first network model, the first output vector representing a supply degree of the target object, inputs a user input feature for the target object into a second network model to obtain a second output vector output by the second network model, the second output vector representing a search degree of the target object by the user, and determines supply-demand state positioning information of the target object according to the first output vector and the second output vector, the supply-demand state positioning information representing a predicted supply-demand relationship of the target object. In this way, when determining the supply degree and the search degree of different target objects, the object features of different target objects can be input into the first network model to obtain the first output vectors representing the supply degrees of different target objects, and the user input features of different target objects can be input into the second network model to obtain the second output vectors representing the search degrees of different target objects by the user, so that the supply-demand relationships of different target objects can be determined according to the first output vectors of different target objects and the second output vectors of different target objects. In this way, when determining the supply degree and the search degree of different target objects, the data analyst can be avoided from analyzing again, so that the analysis efficiency of the supply degree and the search degree can be improved, and the efficiency of determining the supply-demand relationships of different target objects can be improved. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 An optional structure schematic diagram of an information processing system provided by an embodiment of the present application is shown in the figure;

[0017] Figure 2 An optional flowchart of an information processing method provided by an embodiment of the present application is shown in FIG. 1.

[0018] Figure 3 An optional flowchart of determining a data source provided by an embodiment of the present application is shown in FIG. 2.

[0019] Figure 4 An optional flowchart of determining a data source provided by an embodiment of the present application is shown in FIG. 3.

[0020] Figure 5 An optional flowchart of determining a data source provided by an embodiment of the present application is shown in FIG. 4.

[0021] Figure 6 An optional flowchart of an information processing method provided by an embodiment of the present application is shown in FIG. 5.

[0022] Figure 7 An optional flowchart of an information processing method provided by an embodiment of the present application is shown in FIG. 6.

[0023] Figure 8 An optional structure diagram of a network model provided by an embodiment of the present application is shown in FIG. 7.

[0024] Figure 9 An optional structure diagram of an information processing apparatus provided by an embodiment of the present application is shown in FIG. 8.

[0025] Figure 10 An optional structure diagram of an electronic device provided by an embodiment of the present application is shown in FIG. 9. DETAILED DESCRIPTION

[0026] In order to make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the specific technical solutions of the present application will be further described below with reference to the accompanying drawings of the embodiments of the present application. The following embodiments are used to explain the present application, but are not used to limit the scope of the present application.

[0027] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the description herein is for describing the specific embodiments of the present application only and is not intended to be limiting of the present application. As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0028] In the following description, the terms "first", "second", "third" are only used to distinguish different objects, do not represent the specific order of the objects, and do not have the limitation of the order. It can be understood that "first", "second", "third" can be interchanged in a specific order or sequence as allowed, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein.

[0029] In the following description, "some embodiments" are involved, which describe a subset of all possible embodiments, but it can be understood that "some embodiments" can be the same subset or different subset of all possible embodiments, and can be combined with each other without conflict.

[0030] It can be understood that in the embodiments of the present application, related data such as user age, user gender and user occupation are involved. When the embodiments of the present application are applied to specific products or technologies, the user's permission or consent is required, the collection, use and processing of related data need to comply with relevant laws, regulations and standards of relevant countries and regions, and the option of closing the collection, use and processing of related data is provided.

[0031] The information processing method of the embodiments of the present application can be applied to Figure 1 The information processing device 100 shown.

[0032] The information processing device inputs the object features of the target object into the first network model to obtain the first output vector output by the first network model, the first output vector representing the supply degree of the target object, inputs the user input features for the target object into the second network model to obtain the second output vector output by the second network model, the second output vector representing the search degree of the user for the target object, and determines the supply-demand state positioning information of the target object according to the first output vector and the second output vector, the supply-demand state positioning information representing the predicted supply-demand relationship of the target object.

[0033] Next, in conjunction with the schematic diagram of the information processing device 100 shown in Figure 1 The embodiments of the information processing method and device provided by the embodiments of the present application are described.

[0034] Figure 2 The implementation flowchart of the information processing method provided by the embodiments of the present application, which is applied to an information processing device, such as Figure 2 The method can include the following steps:

[0035] S201, the information processing device inputs the object features of the target object into the first network model to obtain the first output vector output by the first network model.

[0036] Here, the first output vector represents a supply degree of the target object. The supply degree includes any one of a supply sufficient and a supply insufficient.

[0037] In an example, the information processing device inputs the object feature of the target object A into the first network model to obtain a first output vector output by the first network model, the first output vector representing that the supply of the target object A is sufficient.

[0038] The target object includes an object displayed to a user, such as a commodity, an advertisement, and news, and the embodiments of the present application do not make any limitation thereto.

[0039] When the target object is a commodity, the object feature of the target object includes a regional feature of the commodity, a brand feature of the commodity, and a price feature of the commodity, and the like.

[0040] S202, the information processing device inputs the user input feature of the target object into the second network model to obtain a second output vector output by the second network model.

[0041] Here, the second output vector represents a search degree of the target object by the user. The search degree includes a high search rate and a low search rate. The embodiments of the present application do not make any limitation to the definition of the high search rate and the low search rate.

[0042] In an example, the search degree can be represented by a search frequency. When the search frequency of the target object by the user is greater than or equal to a search threshold, the high search rate is obtained; and when the search frequency of the target object by the user is less than the search threshold, the low search rate is obtained. The embodiments of the present application do not make any limitation to the specific value of the search threshold. For example, the search threshold is 30.

[0043] The user input feature of the target object includes a description information of the target object input by the user, a behavior feature of the target object by the user, and a user feature.

[0044] The description information of the target object input by the user is used to represent the target object.

[0045] The behavior feature of the target object by the user includes a click, a browse, or an add-to-cart behavior of the target object by the user.

[0046] The user feature includes a user attribute feature, such as a user age, a user gender, and a user city.

[0047] It should be noted that in the embodiments of the present application, the information processing device can obtain the basic information of the user's age and gender only after obtaining the authorization of the user. If the information processing device does not receive the authorization of the user, the information processing device will not be able to obtain the basic information of the user's age and gender.

[0048] S203, the information processing device determines the supply-demand state positioning information of the target object according to the first output vector and the second output vector.

[0049] Here, the supply-demand state positioning information represents the predicted supply-demand relationship of the target object. The supply-demand relationship includes: the supply degree is greater than the search degree, the supply degree is equal to the search degree, or the supply degree is less than the search degree.

[0050] The embodiments of the present application provide an information processing method. An information processing device inputs object features of a target object into a first network model to obtain a first output vector output by the first network model, the first output vector representing a supply degree of the target object. The information processing device inputs user input features of the target object into a second network model to obtain a second output vector output by the second network model, the second output vector representing a search degree of the target object. The information processing device determines supply-demand state positioning information of the target object according to the first output vector and the second output vector, the supply-demand state positioning information representing a predicted supply-demand relationship of the target object. In this way, when determining the supply degree and the search degree of different target objects, the object features of different target objects can be input into the first network model to obtain the first output vector representing the supply degree of different target objects, and the user input features of different target objects can be input into the second network model to obtain the second output vector representing the search degree of different target objects. Therefore, the supply-demand relationship of different target objects can be determined according to the first output vector of different target objects and the second output vector of different target objects. In this way, when determining the supply degree and the search degree of different target objects, the data analyst can be avoided from analyzing again, so that the analysis efficiency of the supply degree and the search degree can be improved, and the efficiency of determining the supply-demand relationship of different target objects can be improved.

[0051] In some embodiments, if the supply-demand state positioning information represents that the supply-demand relationship of the target object is a first supply-demand relationship, the first supply-demand relationship represents that the search degree of the target object is greater than the supply degree, and the method further includes the following steps 1 and 2:

[0052] Step 1, the information processing device determines adjustment direction information of the target object according to the supply information of the target object.

[0053] Here, in a case where the search degree of the target object is greater than the supply degree, the adjustment direction information of the target object determined by the information processing device includes: restocking the target object.

[0054] Step 2, the information processing device adjusts the supply information of the target object according to the adjustment direction information.

[0055] Here, after determining the adjustment direction information of the target object, the information processing device adjusts the supply information of the target object according to the adjustment direction information.

[0056] In some embodiments, the method further includes: the information processing device respectively trains the first initial network model and the second initial network model according to the object features of each sample object in the at least one sample object and the user input features for each sample object in the at least one sample object, to obtain the first network model and the second network model.

[0057] In some embodiments, the training of the first initial network model and the second initial network model according to the object features of each sample object in the at least one sample object and the user input features for each sample object in the at least one sample object, to obtain the first network model and the second network model, includes the following steps 3 to 5:

[0058] Step 3, the information processing device inputs the object features of each sample object in the at least one sample object into the first initial network model to obtain a first training vector output by the first initial network model, and inputs the user input features for each sample object in the at least one sample object into the second initial network model to obtain a second training vector output by the second initial network model.

[0059] Here, for the first initial network model, the information processing device inputs the object features of each sample object in the at least one sample object into the first initial network model to obtain a first training vector. For the second initial network model, the information processing device inputs the user input features for each sample object in the at least one sample object into the second initial network model to obtain a second training vector.

[0060] Step 4, the information processing device obtains first prediction information for each sample object in the at least one sample object based on the first training vector and the second training vector.

[0061] Here, the first prediction information represents a predicted supply and demand relationship for the corresponding sample object.

[0062] After obtaining the first training vector and the second training vector, the information processing device predicts the supply-demand relationship of each of the at least one sample object based on the first training vector and the second training vector.

[0063] In a case where the first initial network model and the second initial network model do not converge, the information processing device determines a loss according to the corresponding first prediction information and the corresponding label of each of the at least one sample object, updates the parameters of the first initial network model and the second initial network model according to the loss, and continues to perform the following processing until the first initial network model and the second initial network model converge to obtain the first network model and the second network model: the information processing device inputs the object features of each of the at least one sample object into the first initial network model to obtain a first training vector output by the first initial network model, and inputs the user input features of each of the at least one sample object into the second initial network model to obtain a second training vector output by the second initial network model.

[0064] Here, the label is a label in a first label set, the first label set includes at least one of the labels, and different labels in the first label set represent different supply-demand relationships corresponding to different sample objects.

[0065] In an example, the at least one sample object includes a sample object A, a sample object B, and a sample object C, and the first label set includes a label 1, a label 2, and a label 3, where the label 1 represents that the supply-demand relationship corresponding to the sample object A is that the supply degree is greater than the search degree, the label 2 represents that the supply-demand relationship corresponding to the sample object B is that the supply degree is less than the search degree, and the label 3 represents that the supply-demand relationship corresponding to the sample object C is that the supply degree is equal to the search degree.

[0066] In the embodiments of the present application, in a case where the first initial network model and the second initial network model do not converge, the information processing device continuously trains the first initial network model and the second initial network model until the first initial network model and the second initial network model converge. In a case where the first initial network model and the second initial network model do not converge, the information processing device determines a loss according to the corresponding first prediction information and the corresponding label of each of the at least one sample object, and updates the parameters of the first initial network model and the second initial network model according to the determined loss to continuously train the first initial network model and the second initial network model.

[0067] In some embodiments, before the step 3, the method further includes steps 6 to 8:

[0068] Step 6, the information processing device acquires training data.

[0069] Here, the training data includes: object features of each of at least two training objects, user input features for each of the at least two training objects, and a second label set. Wherein, the second label set includes at least two labels, and different labels in the second label set represent the supply and demand relationship corresponding to different training objects.

[0070] In an example, the at least two training objects include: training object A and training object B, and the second label set includes: label 1 and label 2, wherein label 1 represents the supply and demand relationship corresponding to training object A, and label 2 represents the supply and demand relationship corresponding to training object B.

[0071] Step 7, the information processing device samples the object features of each of the at least two training objects to obtain sampling data.

[0072] Here, the sampling data includes: object features of each of the at least one sample object.

[0073] After the information processing device acquires the training data, it will sample the object features of each of the at least two training objects included in the training data to obtain the object features of each of the at least one sample object.

[0074] Step 8, the information processing device determines the user input features for each of the at least one sample object from the training data according to the at least one sample object corresponding to the sampling data, and determines the first label set from the second label set according to the at least one sample object corresponding to the sampling data.

[0075] Here, after the information processing device obtains the sampling data, it will determine the user input features for each of the at least one sample object from the training data according to the at least one sample object corresponding to the sampling data, and determine the first label set from the second label set according to the at least one sample object corresponding to the sampling data.

[0076] In some embodiments, determining the loss according to the first prediction information corresponding to each of the at least one sample object and the corresponding label includes: for the first prediction information corresponding to each of the at least one sample object, the information processing device adjusts the first prediction information through the sampling probability corresponding to the sample object to obtain second prediction information, normalizes the second prediction information through an activation function, and determines the loss according to the normalized second prediction information corresponding to each of the at least one sample object and the corresponding label.

[0077] Here, the activation function can be as shown in the following formula (1):

[0078]

[0079] Wherein, x represents the object feature of the target object, y represents the user input feature of the target object, θ represents the model parameter of the first network model and the second network model, and M represents the number of goods.

[0080] After obtaining the activation function, the loss function shown in the following formula (2) can be determined according to the activation function:

[0081]

[0082] Wherein, r represents the label.

[0083] In some embodiments, the method further comprises: the information processing device obtains a sampling interval between two adjacent samplings of the sample object, and determines the sampling probability corresponding to the sample object according to the sampling interval.

[0084] Here, determining the sampling probability corresponding to the sample object according to the sampling interval comprises: determining the reciprocal of the sampling interval according to the sampling interval, and determining the sampling probability according to the reciprocal of the sampling interval.

[0085] With the development of big data and artificial intelligence, more and more users like online shopping. For e-commerce platforms and users, achieving dynamic balance between supply and demand is a win-win goal. In practical applications, when users search for goods on e-commerce platforms, the searched goods often have the problems of high search and low conversion, insufficient goods sources, or excess goods, wherein high search and low conversion refer to high search rate and low conversion rate, wherein high search rate is used to represent high search rate of goods, and low conversion rate is used to represent low purchase rate of goods. How to accurately locate the goods with high search and low conversion, insufficient goods sources, or excess goods, and provide reasonable goods supply suggestions for the goods with high search and low conversion, insufficient goods sources, or excess goods is a matter worth studying.

[0086] Under the above background, the present application takes automobile products as the breakthrough point and proposes an information processing method and device, which can effectively guide business side operation and bring online income.

[0087] In related technologies, the business side operation is guided according to the conclusion corresponding to the automobile category obtained by data analysts, so for other new categories, data analysts need to analyze the data corresponding to the other categories to obtain analysis conclusions, so as to guide the business side to configure resources under the other categories. In this way, it will cause the problem that the generality is not strong, a large number of data analysts are needed for analysis, and the support for new categories needs cost.

[0088] The technical problem solved by the present application includes: for different categories, the corresponding data of different categories can be analyzed by a deep learning double-tower network model to obtain analysis conclusions corresponding to different categories, and the analysis conclusions corresponding to different categories are used to guide the business party to configure resources for different categories. In this way, for different categories, the deep learning double-tower network model can be used to analyze different categories to obtain analysis conclusions corresponding to different categories, without the need for data analysts to re-analyze, so as to better guide the business party to configure resources under different categories.

[0089] Here, the deep learning double-tower network model is the first network model and the second network model described in the above embodiments.

[0090] In the embodiments of the present application, the supply-demand matching score can be used to guide the business party to configure resources for different categories. The supply-demand matching score can be determined according to a supply-demand matching implementation process, as shown in Figure 3

[0091] For feature development, the features include: commodity image features, user features, etc.

[0092] For label construction, the goods judged to be high-search and low-conversion are considered to be goods that need to be replenished and are labeled with a specific label. In addition, according to the value of search clicks, a threshold is designed to divide the training data into different categories of labels.

[0093] After feature development and label construction, the developed features and constructed labels constitute a training sample.

[0094] For model training, the features and labels are combined into a training sample, and the double-tower network model is trained according to the training sample.

[0095] For supply-demand automatic prediction by the double-tower network model file, according to the newly obtained data, the features are constructed according to the feature design mode, and then the features are input into the trained model, and the model gives a supply-demand matching score.

[0096] ​In the present application, batch-softmax-optimization and sample-bias are added in the deep learning twin-tower network model. Through batch-softmax-optimization, the training efficiency of training the twin-tower network model can be improved, and through sample-bias, the over-learning of popular goods can be avoided.

[0097] The information processing method provided by the present application will be described in detail below.

[0098] In the embodiments of the present application, as shown in Figure 4 To help the auxiliary service party understand the data present situation 401 under the automobile category, the user portrait data 402 and the commodity portrait data 403 under the automobile category are first investigated. After investigating the user portrait data 402, the following data are obtained: user gender data 4021, user age data 4022, user city data 4023, user education data 4024 and user occupation data 4025; after investigating the commodity portrait data 403, the following data are obtained: commodity brand data 4031, commodity price data 4032 and commodity attribute data 4033.

[0099] In the embodiments of the present application, to design a supply and demand scheme under the automobile category, the data of the third-level category under the automobile category are first obtained, as shown in Figure 5 The data 501 of the third-level category under the automobile category include: search statistics data 5011, order statistics data 5012, conversion rate statistics data 5013 and single price statistics data 5014.

[0100] In the embodiments of the present application, as shown in Figure 6 The mining scheme for determining the goods with high search and low conversion under the automobile category includes: high search and low conversion judgment logic 610, statistics data 620, the statistics data 620 including: search statistics data 6201, order statistics data 6202, conversion rate statistics data 6203, etc., determining whether it is high search and low conversion according to the statistics data 630, third-level category search ratio 640, third-level category order ratio 650, determining whether it is high search and low conversion according to the third-level category search ratio 640 and the third-level category order ratio 650 660.

[0101] The present application has done a lot of analysis work, and the analysis conclusion obtained after a lot of analysis work is the basis for training the twin-tower network model. The analysis conclusion is used for the design of the following contents: features, labels, sample weighting and model construction. The features are the input of the model, and the labels are the output of the model.

[0102] The double-tower network model comprises a supply-side tower and a demand-side tower, wherein the supply-side tower is used to indicate a commodity goods side and is recommended to a user, and the demand-side tower is used to indicate a demand of a user side.

[0103] In the embodiments of the present application, the label of the double-tower network model can be designed according to the analysis conclusion of high search and low conversion of the automobile category. For example, the commodity judged as high search and low conversion will be considered as a commodity that needs to be replenished, and a specific label is marked on the commodity. In addition, the threshold value can be designed according to the value of the user search and click commodity, and the training data is divided into different categories of labels, such as label 1 representing that the value of search and click is greater than the designed threshold value, and label 2 representing that the value of search and click is less than the designed threshold value.

[0104] As shown in Figure 7 The six modules included in the present application are: an automobile product high search and low conversion analysis module 701, an automobile service category search analysis module 702, an automobile seat cushion portrait-search word analysis module 703, an automobile user-hub supply and demand analysis module 704, an automobile decoration-mat analysis module 705, and a double-tower network model abstraction module 706.

[0105] For the automobile product high search and low conversion analysis module 701, the functions implemented by the automobile product high search and low conversion analysis module include: analyzing regional distribution 7011, analyzing brand distribution 7012, analyzing attribute distribution 7013, and analyzing price distribution 7014.

[0106] When analyzing the regional distribution, the "automobile product" first-level category users are compared with the "automobile product-safety self-driving-motorcycle" users, and Guangdong, Beijing, Sichuan, Jiangsu, Shaanxi and Shandong account for the most, accounting for about 28.47% of the overall automobile product first-level category users.

[0107] When analyzing the attribute distribution, the users pay more attention to five commodity attributes, including style, tire type, whether to support foot pedal, vehicle type, and engine type, which account for about half of the overall.

[0108] When analyzing the price distribution, the exposure rate of the commodity with a price interval of 0-1000 is relatively high, which is 24.12%, and the actual conversion rate is relatively low, which is 0.65%.

[0109] For the automobile service category search analysis module 702, the functions implemented by the automobile service category search analysis module 702 include: big plate overview 7021, data judgment 7022, comparative analysis 7023, and category statistics 7024.

[0110] For the data judgment 7022, the search rate of the third-level category will be determined according to the search statistical data of the third-level category.

[0111] For comparative analysis 7023, different data under different categories will be analyzed.

[0112] For car cushion image-search word analysis module 703, the functions implemented by car cushion image-search word analysis module 703 include: analysis background 7031, user image analysis 7032, product image analysis 7033, and search potential word analysis 7034.

[0113] For analysis background 7031, the order amount of seat cushions accounts for about 20% in car decoration, and when the order amount or GMV of seat cushions fluctuates abnormally, the business party will demand positioning reasons. The business party's appeal is to understand the current situation through user image analysis and product image analysis, to mine the current user demand, and to improve the subsequent seat cushion search user outbound order amount and outbound order GMV.

[0114] For user image analysis 7032, the gender, age, province, city, and education of the user will be analyzed.

[0115] For product image analysis 7033, the brand, price, and attributes of the product will be analyzed.

[0116] For search potential word analysis 7034, search words with less than a set threshold number of returned products and words with no user behavior after searching will be determined as potential words.

[0117] For car user-hub supply and demand analysis module 704, the functions implemented by car user-hub supply and demand analysis module 704 include: analysis demand 7041, data definition 7042, search click probability 7043, and main conclusion 7044.

[0118] For analysis demand 7041, by analyzing search words containing "hub" keywords with low click rate and zero clicks, potential products with no stock, few product categories, and product attributes that cannot meet the buyer's demand are found.

[0119] For data definition 7042, UV represents the number of search exposures, UV_CLK represents the number of exposure clicks, PV represents the number of exposures, UV_CLK / UV represents the search click rate, and PV_CLK / UV represents the average number of browsed products.

[0120] For search click probability 7043, the search click rate of hub search words is concentrated between 50% and 70%, the low search click rate is 0%, and the proportion of users with search but no clicks is high.

[0121] For main conclusion 7044, the proportion of search click rates <10% or >90% is greater than 80%.

[0122] For the car decoration-foot mat analysis module 705, the functions implemented by the car decoration-foot mat analysis module 705 include: analyzing color materials 7051, analyzing pattern styles 7052, analyzing categories and prices 7053, and adapting to popular car models 7054.

[0123] For the analysis of color materials 7051, black is the most selected color, followed by red, gray black, khaki, brown and gray. The popular surface layer main material of the foot mat is leather + silk ring, leather + velvet, and the click order rate of these two materials is higher.

[0124] For the analysis of pattern styles 7052, the popular patterns of the foot mat are plaid, solid color, stripe and business casual, and the click order of plaid and business is higher. The popular styles of the foot mat are large enclosure, full enclosure and flat, and the click order rate of these three is higher.

[0125] For the analysis of category and price 7053, the popular category of the foot mat is five-seat full car, and the user's willingness to purchase this kit category is stronger. The popular price range of the foot mat is concentrated in 100-400 yuan, and the user's click order willingness is the highest in 300-399 yuan.

[0126] For the double-tower network model abstraction module 706, the functions implemented by the double-tower network model abstraction module 706 include: determining supply-side features 7061, determining demand-side features 7062, determining supply-demand matching networks 8063, and predicting supply-demand matching 7064.

[0127] For the supply-side features 7061, the supply-side features include: product features.

[0128] For the demand-side features 7062, the demand-side features include: user features, context background information, and the context background information includes: user's mobile phone model, etc.

[0129] For the determination of the supply-demand matching network 7063, the supply-demand matching network is as shown in Figure 8 , the supply-demand matching network includes: a supply side 801 and a demand side 802, the supply-side features 803 are input to the supply side 801 to obtain a first output vector, the demand-side features 804 are input to the demand side 802 to obtain a second output vector, and the supply-demand state positioning information 805 of the target object is determined according to the first output vector and the second output vector.

[0130] Figure 9 An information processing device provided by an embodiment of the present application is as shown in Figure 9 , the information processing device 900 includes:

[0131] The first determining unit 901 is configured to input an object feature of a target object into a first network model to obtain a first output vector output by the first network model, the first output vector representing a supply degree of the target object.

[0132] The second determining unit 902 is configured to input a user input feature for the target object into a second network model to obtain a second output vector output by the second network model, the second output vector representing a search degree of the target object by the user.

[0133] The third determining unit 903 is configured to determine supply-demand state positioning information of the target object according to the first output vector and the second output vector, the supply-demand state positioning information representing a predicted supply-demand relationship of the target object.

[0134] In some embodiments, if the supply-demand state positioning information represents that the supply-demand relationship of the target object is a first supply-demand relationship, the first supply-demand relationship representing that the search degree of the target object is greater than the supply degree, the first determining unit 901 is further configured to:

[0135] determine adjustment direction information of the target object according to the supply information of the target object;

[0136] adjust the supply information of the target object according to the adjustment direction information.

[0137] In some embodiments, the first determining unit 901 is further configured to:

[0138] train a first initial network model and a second initial network model according to the object feature of each sample object in the at least one sample object and the user input feature for each sample object in the at least one sample object, respectively, to obtain the first network model and the second network model.

[0139] In some embodiments, the first determining unit 901 is further configured to:

[0140] input the object feature of each sample object in the at least one sample object into the first initial network model to obtain a first training vector output by the first initial network model, and input the user input feature for each sample object in the at least one sample object into the second initial network model to obtain a second training vector output by the second initial network model;

[0141] obtain first prediction information for each sample object in the at least one sample object based on the first training vector and the second training vector, the first prediction information representing a predicted supply-demand relationship for the corresponding sample object;

[0142] In a case where the first initial network model and the second initial network model do not converge, loss is determined according to first prediction information corresponding to each sample object in the at least one sample object and a label corresponding to the each sample object, and parameters of the first initial network model and the second initial network model are updated according to the loss, and the following processing is continuously performed until the first initial network model and the second initial network model converge, to obtain the first network model and the second network model: wherein the label is a label in a first label set, the first label set includes at least one label, and different labels in the first label set represent supply-demand relationships corresponding to different sample objects.

[0143] The object features of each sample object in the at least one sample object are input into the first initial network model to obtain a first training vector output by the first initial network model, and the user input features of each sample object in the at least one sample object are input into the second initial network model to obtain a second training vector output by the second initial network model.

[0144] In some embodiments, the information processing apparatus 1000 further includes a processing unit configured to:

[0145] Obtain training data, the training data including object features of each training object in at least two training objects, user input features of each training object in the at least two training objects, and a second label set including at least two labels, different labels in the second label set representing supply-demand relationships corresponding to different training objects;

[0146] Sample the object features of each training object in the at least two training objects to obtain sampling data, the sampling data including object features of each sample object in the at least one sample object;

[0147] Determine, from the training data, the user input features of each sample object in the at least one sample object according to the sampling data corresponding to the at least one sample object, and determine the first label set from the second label set according to the at least one sample object corresponding to the sampling data.

[0148] In some embodiments, the processing unit is further configured to:

[0149] Adjust, for the first prediction information corresponding to each sample object, the first prediction information by a sampling probability corresponding to the sample object to obtain second prediction information;

[0150] Normalize the second prediction information by an activation function;

[0151] According to the normalized second prediction information corresponding to each of the at least one sample object and the corresponding label, the loss is determined.

[0152] In some embodiments, the processing unit is further configured to:

[0153] The sampling interval between the sample object being sampled for the second time and the sample object being sampled for the third time is obtained.

[0154] According to the sampling interval, the sampling probability corresponding to the sample object is determined.

[0155] Embodiments of the present application also provide an electronic device, including a memory and a processor, the memory stores a computer program capable of running on the processor, and the processor implements the information processing method provided in the above embodiments when executing the program.

[0156] Embodiments of the present application also provide a storage medium, that is, a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the information processing method provided in the above embodiments.

[0157] It should be noted that the description of the above storage medium and device embodiments is similar to the description of the above method embodiments, and has similar beneficial effects as the method embodiments. For technical details not disclosed in the storage medium and device embodiments of the present application, please refer to the description of the method embodiments of the present application for understanding.

[0158] It should be noted that, Figure 10 A hardware entity diagram of the electronic device of the embodiments of the present application is shown in FIG. 10. Figure 10 As shown in FIG. 10, the electronic device 1000 includes a processor 1001, at least one communication bus 1002, at least one external communication interface 1004 and a memory 1005. The communication bus 1002 is configured to realize the connection and communication between the components. In an example, the electronic device 1000 further includes a user interface 1003. The user interface 1003 can include a display screen, and the external communication interface 1004 can include a standard wired interface and a wireless interface.

[0159] The memory 1005 is configured to store instructions and applications executable by the processor 1001, and can also cache data to be processed by the processor 1001 and data to be processed or having been processed by each module in the electronic device (for example, image data, audio data, voice communication data and video communication data), which can be realized by FLASH or RAM.

[0160] It should be understood that every feature, structure, or characteristic described herein is within a preferred embodiment of the present application. Thus, it is meant that the features, structures, or characteristics can be combined with each other in any manner within a preferred embodiment of the present application. In addition, it is contemplated that each feature, structure, or characteristic can be implemented in hardware, software, or a combination thereof.

[0161] It should be noted that, as used herein, the terms "includes," "including," or "includes" are intended to be open-ended terms that specifically permit the inclusion of other elements not specifically recited. As used herein, the terms "comprises," "comprising," or the like are to be construed as open-ended terms, indicating the presence of the stated features, groups of features, or the like, but not excluding the presence of one or more other features, groups of features, or the like.

[0162] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The above-described device embodiments are only illustrative, for example, the division of the units is only a logical function division, and actual implementation can have another division manner, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between each component part shown or discussed can be through some interface, indirect coupling or communication connection between devices or units, which can be electrical, mechanical or other forms.

[0163] The units described above as separate components can or can not be physically separated, and the components shown as units can or can not be physical units; they can be located in one place or distributed on multiple network units; some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0164] In addition, each functional unit in each embodiment of the present application can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be realized in the form of hardware or hardware plus software functional unit.

[0165] Those skilled in the art can understand that all or part of the steps of the above-mentioned method embodiments can be completed by program instruction related hardware, the foregoing program can be stored in a computer readable storage medium, and the program executes the steps of the method embodiments when executed; and the foregoing storage medium includes a mobile storage device, a read only memory (ROM), a magnetic disc or an optical disc, and various storage medium capable of storing program codes.

[0166] Alternatively, the integrated units of the present application can be stored in a computer readable storage medium if the integrated units are realized in the form of software function modules and sold or used as independent products. Based on this understanding, the technical solutions of the embodiments of the present application can be embodied in the form of software products, and the computer software products are stored in a storage medium, including a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in the embodiments of the present application. The foregoing storage medium includes a mobile storage device, a ROM, a magnetic disc or an optical disc, and various storage medium capable of storing program codes.

[0167] The above is only an embodiment of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can easily think of changes or replacements within the technical range disclosed in the present application, which should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. An information processing method, characterized in that, The method includes: The object features of the target object are input into the first network model to obtain the first output vector output by the first network model. The first output vector represents the supply level of the target object. The user input features targeting the target object are input into the second network model to obtain a second output vector, which represents the degree of the user's search for the target object. The user input features of the target object include: descriptive information of the target object input by the user, user behavior features of the target object, and user features. The first network model and the second network model are deep learning dual-tower network models. The first network model and the second network model are trained on the first initial network model and the second initial network model respectively based on the object features of each sample object in at least one sample object and the user input features targeting each sample object in at least one sample object. Based on the first output vector and the second output vector, the supply and demand status positioning information of the target object is determined, and the supply and demand status positioning information represents the predicted supply and demand relationship of the target object.

2. The method according to claim 1, characterized in that, If the supply and demand status positioning information indicates that the supply and demand relationship of the target object is a first supply and demand relationship, and the first supply and demand relationship indicates that the search degree of the target object is greater than the supply degree, the method further includes: Based on the supply information of the target object, determine the adjustment direction information of the target object; The supply information of the target object is adjusted according to the adjustment direction information.

3. The method according to claim 2, characterized in that, Based on the object features of each sample object in at least one sample object and the user input features for each sample object in the at least one sample object, a first initial network model and a second initial network model are trained respectively to obtain the first network model and the second network model, including: The object features of each sample object in the at least one sample object are input into the first initial network model to obtain the first training vector output by the first initial network model, and the user input features of each sample object in the at least one sample object are input into the second initial network model to obtain the second training vector output by the second initial network model. Based on the first training vector and the second training vector, first prediction information is obtained for each sample object in the at least one sample object, and the first prediction information represents the predicted supply and demand relationship for the corresponding sample object. If the first initial network model and the second initial network model fail to converge, a loss is determined based on the first prediction information and corresponding label of each sample object in the at least one sample object, and the parameters of the first initial network model and the second initial network model are updated based on the loss. The following process continues until the first initial network model and the second initial network model converge, resulting in the first network model and the second network model: wherein the label is a label in a first label set, the first label set includes at least one of the labels, and different labels in the first label set represent the supply and demand relationship corresponding to different sample objects; The object features of each sample object in the at least one sample object are input into the first initial network model to obtain the first training vector output by the first initial network model, and the user input features of each sample object in the at least one sample object are input into the second initial network model to obtain the second training vector output by the second initial network model.

4. The method according to claim 3, characterized in that, Before inputting the object features of each sample object in the at least one sample object into the first initial network model to obtain the first training vector output by the first initial network model, and inputting the user input features of each sample object in the at least one sample object into the second initial network model to obtain the second training vector output by the second initial network model, the method further includes: Acquire training data, which includes: object features of each training object in at least two training objects, user input features of each training object in the at least two training objects, and a second label set, which includes at least two labels, and different labels in the second label set represent the supply and demand relationship corresponding to different training objects; The object features of each training object in the at least two training objects are sampled to obtain sampled data, the sampled data including: the object features of each sample object in the at least one sample object; Based on at least one sample object corresponding to the sampling data, user input features for each sample object in the at least one sample object are determined from the training data, and based on at least one sample object corresponding to the sampling data, the first tag set is determined from the second tag set.

5. The method according to claim 4, characterized in that, The step of determining the loss based on the first prediction information and corresponding label of each sample object in the at least one sample object includes: For each sample object, the first prediction information is adjusted by the sampling probability corresponding to the sample object to obtain the second prediction information. The second prediction information is normalized using an activation function; The loss is determined based on the normalized second prediction information and the corresponding label of each sample object in the at least one sample object.

6. The method according to claim 5, characterized in that, The method further includes: Obtain the sampling interval between two consecutive samplings of the sample object; The sampling probability corresponding to the sample object is determined based on the sampling interval.

7. An information processing device, characterized in that, The device includes: The first determining unit is used to input the object features of the target object into the first network model to obtain a first output vector output by the first network model, wherein the first output vector represents the supply level of the target object; The second determining unit is used to input user input features for the target object into the second network model to obtain a second output vector output by the second network model. The second output vector represents the degree of search by the user for the target object. The user input features of the target object include: descriptive information of the target object input by the user, behavioral features of the user towards the target object, and user features. The first network model and the second network model are deep learning dual-tower network models. The first network model and the second network model are obtained by training the first initial network model and the second initial network model respectively based on the object features of each sample object in at least one sample object and the user input features for each sample object in at least one sample object. The third determining unit is used to determine the supply and demand status positioning information of the target object based on the first output vector and the second output vector, wherein the supply and demand status positioning information represents the predicted supply and demand relationship of the target object.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the information processing method according to any one of claims 1 to 6.

9. A storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the information processing method according to any one of claims 1 to 6.

Images