An information processing method, device, and storage medium

Abstract

Embodiments of the present invention disclose an information processing method, apparatus, and storage medium, including: when obtaining multiple parameter information of a target object and historical quantity information of the target object, determining multiple initial weight values corresponding to the multiple parameter information according to the multiple parameter information and the historical quantity information; inputting the multiple initial weight values into an analytic hierarchy process model to obtain multiple weight values; inputting the multiple weight values into a preset convolutional neural network model to obtain a category label corresponding to the target object, and processing the target object based on the processing method corresponding to the category label, where the weight value of each processing layer in the convolutional neural network model is obtained based on the parameters of the current processing layer and the next layer.

Description

Technical Field

[0001] The present invention relates to the technical field of classification, and in particular, to an information processing method, an apparatus, and a storage medium. Background Art

[0002] With the continuous development of Internet technology, users can obtain target objects in other regions through Internet technology to improve the convenience of obtaining target objects.

[0003] In the prior art, during the process of a user obtaining a target object using the Internet, the Internet will record the acquisition record of the target object, such as the acquisition quantity information, brand, model, size, and other information of the target object. A person can classify the target object according to the acquisition quantity information, brand, model, size, and other information of the target object to determine the category label of the target object. Since a person cannot specifically quantify the critical point division of the category label and directly determines the classification label of the target object based on personal experience, the accuracy of classifying the target object is reduced. Summary of the Invention

[0004] To solve the above technical problems, embodiments of the present invention are expected to provide an information processing method, an apparatus, and a storage medium, which can improve the accuracy of an information processing apparatus in classifying a target object.

[0005] The technical solution of the present invention is implemented as follows:

[0006] An embodiment of the present application provides an information processing method, including:

[0007] In the case of obtaining multiple parameter information of a target object and historical quantity information of the target object, determining multiple initial weight values corresponding to the multiple parameter information according to the multiple parameter information and the historical quantity information;

[0008] Inputting the multiple initial weight values into an analytic hierarchy process model to obtain multiple weight values;

[0009] Inputting the multiple weight values into a preset convolutional neural network model to obtain a category label corresponding to the target object, and processing the target object based on a processing method corresponding to the category label, where the weight value of each processing layer in the convolutional neural network model is obtained based on the parameters of the current processing layer and the next layer.

[0010] An embodiment of the present application provides an information processing apparatus, where the apparatus includes:

[0011] A determination unit, configured to determine, when multiple parameter information of a target object and historical quantity information of the target object are obtained, multiple initial weight values corresponding to the multiple parameter information according to the multiple parameter information and the historical quantity information;

[0012] An input unit, configured to input the multiple initial weight values into an analytic hierarchy process model to obtain multiple weight values; input the multiple weight values into a preset convolutional neural network model to obtain a category label corresponding to the target object, and process the target object based on a processing method corresponding to the category label, where the weight value of each processing layer in the convolutional neural network model is obtained based on parameters of the current processing layer and the next layer.

[0013] An embodiment of the present application provides an information processing device, where the device includes:

[0014] A memory, a processor, and a communication bus, where the memory communicates with the processor through the communication bus, and the memory stores a program for information processing executable by the processor. When the program for information processing is executed, the above-mentioned information processing method is executed through the processor.

[0015] An embodiment of the present application provides a storage medium, on which a computer program is stored and applied to an information processing device. The computer program, when executed by a processor, implements the above-mentioned information processing method.

[0016] An embodiment of the present invention provides an information processing method, device, and storage medium. The information processing method includes: when multiple parameter information of a target object and historical quantity information of the target object are obtained, determining multiple initial weight values corresponding to the multiple parameter information according to the multiple parameter information and the historical quantity information; inputting the multiple initial weight values into an analytic hierarchy process model to obtain multiple weight values; inputting the multiple weight values into a preset convolutional neural network model to obtain a category label corresponding to the target object, and processing the target object based on a processing method corresponding to the category label, where the weight value of each processing layer in the convolutional neural network model is obtained based on parameters of the current processing layer and the next layer. By adopting the above method implementation solution, when the information processing device obtains multiple initial weight values corresponding to multiple parameter information of the target object, the information processing device can use the analytic hierarchy process model to process the multiple initial weight values to obtain multiple weight values, and use the preset convolutional neural network model to process the multiple weight values, so as to obtain the category label of the target object, improving the accuracy of the information processing device when classifying the target object. Description of the Drawings

[0017] Figure 1 It is a flowchart of an information processing method provided by an embodiment of the present application;

[0018] Figure 2 This is a flowchart of an exemplary information processing method provided by an embodiment of the present application;

[0019] Figure 3 This is a schematic diagram of the composition structure of an exemplary information processing device provided by an embodiment of the present application;

[0020] Figure 4 This is a schematic diagram of the composition structure of an information processing device provided by an embodiment of the present application Figure 1 ;

[0021] Figure 5 This is a schematic diagram of the composition structure of an information processing device provided by an embodiment of the present application Figure 2 。 Detailed implementation manners

[0022] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0023] Embodiment 1

[0024] An embodiment of the present application provides an information processing method, Figure 1 This is a process of an information processing method provided by an embodiment of the present application Figure 1 , such as Figure 1 shown, the information processing method may include:

[0025] S101. When obtaining multiple parameter information of a target object and historical quantity information of the target object, determine multiple initial weight values corresponding to the multiple parameter information according to the multiple parameter information and the historical quantity information.

[0026] The information processing method provided by an embodiment of the present application is applicable to the scenario where an information processing device classifies a target object.

[0027] In the embodiment of the present application, the information processing device may be implemented in various forms. For example, the information processing device described in the present application may include devices such as mobile phones, cameras, tablet computers, laptop computers, handheld computers, personal digital assistants (Personal Digital Assistant, PDA), portable media players (Portable Media Player, PMP), navigation devices, wearable devices, smart bracelets, pedometers, etc., and devices such as digital TVs, desktop computers, etc.

[0028] In the embodiments of the present application, the target object may be a commodity, or it may be an express delivery, or it may be other things. Specifically, it can be determined according to the actual situation, and the embodiments of the present application do not make any limitations in this regard.

[0029] In the embodiments of the present application, if the target object is a commodity, then the multiple parameter information includes information such as the brand, model, size, browsing, adding to cart, and evaluation of the commodity.

[0030] In the embodiments of the present application, if the target object is a commodity, then the historical quantity information of the target object may be the historical sales volume of the commodity.

[0031] It should be noted that the historical sales volume information may be the total historical sales volume of the commodity, or it may be the historical average sales volume of the commodity. Specifically, it can be determined according to the actual situation, and the embodiments of the present application do not make any limitations in this regard.

[0032] In the embodiments of the present application, the information processing device may obtain the multiple parameter information of the target object and the historical quantity information of the target object from the log information. The information processing device may also obtain the multiple parameter information of the target object and the historical quantity information of the target object in other ways. Specifically, it can be determined according to the actual situation, and the embodiments of the present application do not make any limitations in this regard.

[0033] In the embodiments of the present application, the multiple initial weight values are the initial weight values corresponding to the multiple parameter information, and one parameter information corresponds to one initial weight value.

[0034] It should be noted that the sum of the multiple initial weight values is 1.

[0035] In the embodiments of the present application, before the information processing device determines the multiple initial weight values corresponding to the multiple parameter information according to the multiple parameter information and the historical quantity information, the information processing device will also obtain the initial historical quantity information of the target object; the information processing device performs denoising processing on the initial historical quantity information to obtain the historical quantity information.

[0036] In the embodiments of the present application, the initial historical quantity information may be the historical sales volume information of the target object. The information processing device may obtain the historical sales volume information from the log information corresponding to the target object, or the information processing device may obtain the historical sales volume information from the database of the target object. Specifically, it can be determined according to the actual situation, and the embodiments of the present application do not make any limitations in this regard.

[0037] In the embodiments of the present application, the process of the information processing device performing denoising processing on the initial historical quantity information to obtain the historical quantity information includes the information processing device obtaining the total sales volume of the target object within a period of time, and the information processing device obtaining the average sales volume of the target object within the period of time according to the total sales volume and the period of time, that is, obtaining the historical quantity information.

[0038] Exemplarily, the original data in industry usually contains a lot of noisy data. For the commodity sales volume data of an e-commerce platform, it is affected by business data entry errors or other characteristic reasons such as time, season, major promotions, etc. The intensity of activities like 618 and Double 11 often lasts for two weeks, and the data during this period needs to be denoised. For example, the historical quantity information can be x i , assuming:

[0039]

[0040] Among them, is the average value of the sales volume in 6 days including the sales volume in the previous 3 days and the next 3 days, and σ is the standard deviation of the sales volume in 6 days including the sales volume in the previous 3 days and the next 3 days. If x i meets the conditions of formula (1), then x i is considered not to be an outlier; if x i does not meet the conditions of formula (1), then x i is considered to be an outlier, and the following denoising process needs to be performed on x i : i Perform the following denoising process on x:

[0041]

[0042] Using formula (2) can smooth the noise points of the historical quantity information, but at the same time can also retain the increasing or decreasing trend that the point should have at present. Therefore, formula (2) uses to replace x i , is the average value of the sales volume in the previous 3 days and the current day's sales volume. i The average value of the sales volume in the previous 3 days and the current day's sales volume.

[0043] In the embodiment of the present application, the process by which the information processing device determines multiple initial weight values corresponding to multiple parameter information according to the multiple parameter information and the historical quantity information includes that the information processing device combines each parameter information in the multiple parameter information with the historical quantity information respectively to obtain multiple groups of combined information; the information processing device performs linear regression processing on the multiple groups of combined information to obtain multiple initial weight values.

[0044] In the embodiment of the present application, the information processing device can use a linear regression model to determine multiple initial weight values corresponding to multiple parameter information according to the multiple parameter information and the historical quantity information; the information processing device can also determine multiple initial weight values corresponding to multiple parameter information in other ways, which can be specifically determined according to the actual situation, and the embodiment of the present application does not limit this.

[0045] In the embodiments of the present application, it is assumed that a given data set D = {(x1, y1), (x2, y2),..., (x i , y i )}, where x i = (x i1 ; x i2 ;...; x id ), x i is a vector obtained based on multiple parameter information of the i-th commodity, y i is the historical sales volume information of the i-th commodity, and the linear regression model is a function for prediction through a linear combination of attributes, that is

[0046] f(x) = ω T x + b (3)

[0047] where ω T = (ω1; ω2;...; ω d ), d is the number of attributes. After ω and b are learned, the linear regression model can start to predict. Usually, the root mean square error is used as the minimization loss function of the function, and the gradient descent algorithm is used to solve the loss function formula (4).

[0048]

[0049] where Then Finally, by taking the derivative, ω and b can be obtained.

[0050] It should be noted that the obtained vector ω is the multiple initial weight values.

[0051] S102. Input the multiple initial weight values into the analytic hierarchy process model to obtain multiple weight values.

[0052] In the embodiments of the present application, after the information processing device determines multiple initial weight values corresponding to multiple parameter information according to the multiple parameter information and historical quantity information, the information processing device inputs the multiple initial weight values into the analytic hierarchy process model to obtain multiple weight values.

[0053] It should be noted that the analytic hierarchy process model can be a model obtained according to the Analytic Hierarchy Process (AHP).

[0054] In the embodiments of the present application, the process by which the information processing device inputs multiple initial weight values into the analytic hierarchy process model to obtain multiple weight values includes: the information processing device determines an initial matrix according to the multiple initial weight values; the information processing device determines the logarithm of each element in the initial matrix to obtain an anti-symmetric matrix; the information processing device compares the importance between any two elements in the anti-symmetric matrix according to the three-scale rule to obtain a comparison matrix; the information processing device determines the optimal transfer matrix of the comparison matrix, and obtains multiple weight values according to the optimal transfer matrix.

[0055] In the embodiments of the present application, the information processing device initializes multiple initial weight values using a linear regression model to obtain an initial matrix A, that is, a judgment matrix A. Then, the information processing device determines the logarithm of each element in the initial matrix to obtain an anti-symmetric matrix B. The information processing device then compares the importance between any two elements in the anti-symmetric matrix B according to the three-scale rule to obtain a comparison matrix C; the information processing device determines the optimal transfer matrix O of matrix C, and obtains matrix D according to the optimal transfer matrix O. The eigenvector of the determined matrix D is the multiple weight values.

[0056] Exemplarily, the judgment matrix A = [a ij n×n

[0057] Find the anti-symmetric matrix of the judgment matrix A

[0058] Among them,

[0059] According to matrix B, construct a comparison matrix using the 3-scale method

[0060] Among them,

[0061] Find the optimal transfer matrix of C

[0062] Among them,

[0063] Determine matrix according to matrix O Among them,

[0064] The eigenvector W of the obtained matrix D i , that is, [W1, W2,..., W n T is the multiple weight values.

[0065] ​​It can be understood that the optimal transfer matrix is essentially used to obtain reasonable index weight coefficients by converting the original matrix through different mathematical means. Considering that the traditional analytic hierarchy process constructs the judgment matrix in the form of a positive reciprocal matrix, which largely ignores the influence degree of other indexes on the current index, the optimal transfer matrix method is adopted in this application to optimize the judgment matrix, without ignoring the influence degree of other indexes on the current index, and improving the accuracy when the information processing device determines multiple weight values.

[0066] S103. Input multiple weight values into a preset convolutional neural network model to obtain a category label corresponding to the target object, and process the target object based on the processing method corresponding to the category label. The weight value of each processing layer in the preset convolutional neural network model is obtained based on the parameters of the current processing layer and the next layer.

[0067] In the embodiment of this application, after the information processing device inputs multiple initial weight values into the analytic hierarchy model to obtain multiple weight values, the information processing device inputs the multiple weight values into the preset convolutional neural network model to obtain a category label corresponding to the target object.

[0068] In the embodiment of this application, if the target object can be a commodity, the category label corresponding to the target object can be a popular commodity label, a best-selling commodity label, a regular sales volume commodity label, and a low sales volume commodity label.

[0069] It should be noted that in the case where the information device determines that the category label is a popular commodity label, the information processing device can sell the target object based on the sales method of the popular commodity; in the case where the information device determines that the category label is a best-selling commodity label, the information processing device can sell the target object based on the sales method of the best-selling commodity; in the case where the information device determines that the category label is a regular sales volume commodity label, the information processing device can sell the target object based on the sales method of the regular sales volume commodity; in the case where the information device determines that the category label is a low sales volume commodity label, the information processing device can sell the target object based on the sales method of the low sales volume commodity.

[0070] In the embodiment of this application, the processing layer can be a convolutional layer in the preset convolutional neural network model, or a downsampling layer (pooling layer) of the preset convolutional neural network model, or a fully connected layer of the preset convolutional neural network model, which can be specifically determined according to the actual situation, and this application embodiment does not limit this.

[0071] It should be noted that the number of convolutional layers can be one, two, or multiple, which can be specifically determined according to the actual situation, and this application embodiment does not limit this.

[0072] It should be noted that the number of pooling layers can be one, two, or multiple, and can be specifically determined according to the actual situation. The embodiments of the present application do not limit this.

[0073] It should be noted that the number of fully connected layers can be one, two, or multiple, and can be specifically determined according to the actual situation. The embodiments of the present application do not limit this.

[0074] In the embodiments of the present application, the process in which the information processing device inputs multiple weight values into the preset convolutional neural network model to obtain the category label corresponding to the target object includes: the information processing device inputs multiple weight values into the preset convolutional neural network model and uses the preset convolutional neural network model to predict the processing quantity of the target object; the information processing device determines the category label according to the processing quantity.

[0075] It should be noted that if the target object can be a commodity, the processing quantity of the target object can be the predicted sales information of the commodity in a future period of time.

[0076] In the embodiments of the present application, there is a corresponding relationship between the preset weight value and the preset sales volume in the preset convolutional neural network model. The preset convolutional neural network model can predict the processing quantity of the target object according to the corresponding relationship between the preset weight value and the preset sales volume and multiple weight values.

[0077] In the embodiments of the present application, before the information processing device inputs multiple weight values into the preset convolutional neural network model to obtain the category label corresponding to the target object, the information processing device also acquires the initial weight of the current processing layer of the initial convolutional neural network model, the learning rate of the current processing layer, the learning rate of the next layer, the activation function value of the current processing layer, and the activation function value of the next layer; the information processing device updates the initial weight using the learning rate of the current processing layer, the learning rate of the next layer, the activation function value of the current processing layer, and the activation function value of the next layer to obtain the updated weight; the information processing device obtains the preset convolutional neural network model according to the updated weight.

[0078] In the embodiments of the present application, the preset convolutional neural network model can specifically be a multi-class convolutional neural network (MCNN). Among them, MCNN contains two convolutional layers, two downsampling layers (pooling layers), and two fully connected layers. Let I be the input layer, where the input matrix is x, O1 and O2 represent adding classifiers in the network, FC represents the fully connected layer, and the classification output functions of FC, O1, and O2 are y, y1, and y2 respectively, and FC contains classifiers, C1 and C2 represent convolutional layers, and P1 and P2 represent two downsampling layers.

[0079] There are n sample commodities (x i , y i ). The network has L layers, and the output of the last layer is f(x i ). The expected output is y i ′. The loss function in the form of cross-entropy of the three classifiers can be determined according to formula (7);

[0080]

[0081] where: E = (E1, E2, E FC ) T ; i = 1, 2, 3.

[0082] In the weight update method in the training of the convolutional neural network in the prior art, only the output of the classifier of the current network layer is considered, but it has nothing to do with the output results of other classifiers. This will lead to the situation that once the current classifier cannot effectively obtain the classification result, its convolution calculation result cannot be well retrained by the deeper network, and thus the optimal solution of classification cannot be obtained.

[0083] In the embodiment of the present application, it is based on the improvement of the original propagation algorithm. Let the weight of the convolutional layer be ω = (ω1, ω2, ω FC ) T , the learning rate α = (α1, α1, α1) T , and there is an activation function after each convolutional layer. The activation value is used as the input signal θ = (θ1, θ2, θ FC ) T of the convolutional layer, and a new weight update method is obtained, as shown in formulas (8)-(10):

[0084]

[0085]

[0086]

[0087] In the embodiment of the present application, the intermediate result can be obtained by the chain rule of differentiation. For example, the weight update of the first convolutional layer can be obtained by performing the gradient descent algorithm operation according to the output values of the first and second classifiers and the FC.

[0088] It can be understood that, similar to the general forward propagation algorithm, the training of the initial convolutional neural network model is divided into two processes: forward propagation and backward propagation. In forward propagation, each convolutional layer obtains an activation value, and each classifier calculates the error using cross-entropy. In the backward propagation process, the weight update of each convolutional layer requires the error signals of other layer classifiers. In this way, the parameter training of each layer of the network is affected by the outputs of both this layer and other classifiers, making the weight update of each convolutional layer related not only to the output of the classifier of this layer but also to the outputs of other deeper network classifiers. As a result, deeper features can be learned, improving the accuracy of classification using the preset convolutional neural network model.

[0089] In an embodiment of the present application, after the information processing device inputs multiple weight values into the preset convolutional neural network model and obtains the class label corresponding to the target object, the information processing device will also determine the target parameter information corresponding to the class label; the information processing device outputs the class label and the target parameter information.

[0090] It should be noted that the target parameter information is one of the multiple parameter information.

[0091] In an embodiment of the present application, the target parameter information further includes the weight value corresponding to the target parameter information.

[0092] In an embodiment of the present application, the manner in which the information processing device outputs the class label and the target parameter information can be that the information processing device displays the class label and the target parameter information; or the information processing device transmits the class label and the target parameter information to other devices, which can be specifically determined according to the actual situation, and the embodiments of the present application do not limit this.

[0093] Exemplarily, as Figure 2 shown, the information processing device inputs the training data (multiple parameter information of the target object and the historical quantity information of the target object) into the linear regression model (regression model), performs linear regression processing on the training data using the linear regression model to obtain multiple initial weight values (initializing the weights of important factors affecting the label) corresponding to the multiple parameter information, and determines multiple weight values (updating the weight coefficients) corresponding to the multiple initial weight values using the analytic hierarchy process model (analytic hierarchy process); inputs the multiple weight values into the preset convolutional neural network model (improved convolutional neural network) to obtain the class label (multi-class label) corresponding to the target object.

[0094] Exemplarily, as Figure 3As shown in the figure, the information processing device first obtains multiple parameter information of the target object and the initial historical quantity information of the target object from the data source, then performs denoising processing on the initial historical quantity information to obtain the historical quantity information. The information processing device performs linear regression processing on the multiple parameter information and the historical quantity information by using a linear regression model to obtain multiple initial weight values corresponding to the multiple parameter information. The information processing device inputs the multiple initial weight values into the layer analysis model to obtain multiple weight values; the information processing device inputs the multiple weight values into a preset convolutional neural network model to obtain category label information and target parameter information corresponding to the category label.

[0095] In the embodiment of the present application, in order to verify the practicability, reliability and superiority of the classification method proposed in the present application, after applying the test data to this method, this experimental verification is mainly evaluated from the importance degree of the explosion product related index and the accuracy of the explosion product classification prediction. Through the experience of business and procurement and sales personnel, the products with the monthly sales volume accounting for the top 5% are defined as explosion products as the classification labels.

[0096] In the embodiment of the present application, the sales volume is divided into four categories, that is, mapped to 4 classification labels. The explosion products are still the products ranked in the top 5 in terms of sales volume. Followed by hot products, ordinary products, and cold products. The preliminary screening of the above linear regression model and the refined screening of the improved analytic hierarchy process (Interval Analytic Hierarchy Process, IAHP) decision model are used to obtain the explosion product indicators. Using the factor weights learned by the model and putting them into the deep learning model can improve the interpretability of the influencing factors of the multi-classification labels. Model training of multiple methods has been carried out on the data set, and the comparison results of the model performance are shown in Table 2.

[0097] Table 2 Performance comparison of models

[0098] Accuracy Classification Time LR + CNN 60.31％ 93.32 seconds AHP + CNN 64.57％ 157.86 seconds IAHP + CNN 74.18％ 203.86 seconds HFIAC 83.57％ 197.26 seconds

[0099] Through comparative analysis, the overall classification accuracy of further extracting more accurate explosion product factors through AHP and IAHP is significantly improved. Among them, the classification speed of the classification method (HFIAC) proposed in the present application is only 0.014% lower than that of IAHP+CNN, but the classification accuracy is as high as 83.57%, which is significantly higher than other classifiers that do not consider the weight update of each convolutional layer and are also related to the output of deeper network classifiers. Thus, it further positively proves the effectiveness and feasibility of this method.

[0100] It should be noted that Convolutional Neural Network (CNN).

[0101] It can be understood that after applying test data to this method, the weights of the important factors affecting potential hit products can be obtained from the predicted potential hit products, and the quantiles of each indicator can be obtained according to the category dimension. Business or procurement and sales personnel can intuitively analyze the indicator recommendation data, improve the understanding of the hit product critical point, and thus better support the business to create a hit business development.

[0102] It can be understood that when the information processing device obtains multiple initial weight values corresponding to multiple parameter information of the target object, the information processing device can use the analytic hierarchy process model to process the multiple initial weight values to obtain multiple weight values, and use the preset convolutional neural network model to process the multiple weight values, so as to obtain the category label of the target object, and improve the accuracy of the information processing device when classifying the target object.

[0103] Embodiment 2

[0104] Based on the same inventive concept as Embodiment 1, the embodiment of the present application provides an information processing device 1, corresponding to an information processing method; Figure 4 Schematic diagram of the composition structure of an information processing device provided by an embodiment of the present application Figure 1 The information processing device 1 may include:

[0105] A determination unit 11, configured to determine multiple initial weight values corresponding to the multiple parameter information according to the multiple parameter information and the historical quantity information when obtaining the multiple parameter information of the target object and the historical quantity information of the target object;

[0106] An input unit 12, configured to input the multiple initial weight values into an analytic hierarchy process model to obtain multiple weight values; input the multiple weight values into a preset convolutional neural network model to obtain the category label corresponding to the target object, and process the target object based on the processing method corresponding to the category label, and the weight values of each processing layer in the preset convolutional neural network model are obtained based on the parameters of the current processing layer and the next layer.

[0107] In some embodiments of the present application, the input unit 12 is configured to input the multiple weight values into a preset convolutional neural network model and use the preset convolutional neural network model to predict the processing quantity of the target object;

[0108] The determination unit 11 is configured to determine the category label according to the processing quantity.

[0109] In some embodiments of the present application, the device further includes an acquisition unit and an update unit;

[0110] The obtaining unit is configured to obtain the initial weight of the current processing layer of the initial convolutional neural network model, the learning rate of the current processing layer, the learning rate of the next layer, the activation function value of the current processing layer, and the activation function value of the next layer;

[0111] The updating unit is configured to update the initial weight by using the learning rate of the current processing layer, the learning rate of the next layer, the activation function value of the current processing layer, and the activation function value of the next layer to obtain an updated weight; and obtain the preset convolutional neural network model according to the updated weight.

[0112] In some embodiments of the present application, the apparatus further includes a comparison unit;

[0113] The determining unit 11 is configured to determine an initial matrix according to the multiple initial weight values; determine the logarithm of each element in the initial matrix to obtain an anti-symmetric matrix; determine an optimal transfer matrix of a comparison matrix, and obtain the multiple weight values according to the optimal transfer matrix;

[0114] The comparison unit is configured to compare the importance between any two elements in the anti-symmetric matrix according to the three-scale rule to obtain the comparison matrix.

[0115] In some embodiments of the present application, the apparatus further includes a combining unit and a processing unit;

[0116] The combining unit is configured to combine each parameter information in the multiple parameter information with the historical quantity information respectively to obtain multiple groups of combined information;

[0117] The processing unit is configured to perform linear regression processing on the multiple groups of combined information to obtain the multiple initial weight values.

[0118] In some embodiments of the present application, the obtaining unit is configured to obtain the initial historical quantity information of the target object;

[0119] The processing unit is configured to perform denoising processing on the initial historical quantity information to obtain the historical quantity information.

[0120] In some embodiments of the present application, the apparatus further includes an output unit;

[0121] The determining unit 11 is configured to determine the target parameter information corresponding to the category label; the target parameter information is one of the multiple parameter information;

[0122] The output unit is configured to output the category label and the target parameter information.

[0123] It should be noted that in practical applications, the above-mentioned determination unit 11 and input unit 12 can be implemented by the processor 13 on the information processing device 1, specifically implemented by a CPU (Central Processing Unit), an MPU (Microprocessor Unit), a DSP (Digital Signal Processing), or a field programmable gate array (FPGA, Field Programmable Gate Array), etc.; the above-mentioned data storage can be implemented by the memory 14 on the information processing device 1.

[0124] An embodiment of the present invention further provides an information processing device 1, as Figure 5 shown, the information processing device 1 includes: a processor 13, a memory 14, and a communication bus 15. The memory 14 communicates with the processor 13 through the communication bus 15. The memory 14 stores programs executable by the processor 13. When the programs are executed, the information processing method as described above is executed through the processor 13.

[0125] In practical applications, the above-mentioned memory 14 can be a volatile memory, such as a random access memory (RAM); or a non-volatile memory, such as a read-only memory (ROM), a flash memory, a hard disk drive (HDD), or a solid-state drive (SSD); or a combination of the above types of memories, and provides instructions and data to the processor 13.

[0126] An embodiment of the present invention provides a computer-readable storage medium with a computer program thereon. When the program is executed by the processor 13, the information processing method as described above is implemented.

[0127] It can be understood that when the information processing device obtains multiple initial weight values corresponding to multiple parameter information of the target object, the information processing device can use the analytic hierarchy process model to process the multiple initial weight values to obtain multiple weight values, and use the preset convolutional neural network model to process the multiple weight values, thereby obtaining the category label of the target object, improving the accuracy of the information processing device when classifying the target object.

[0128] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can be in the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can be in the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories and optical memories, etc.) containing computer-usable program code.

[0129] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one or more of the flows Figure 1 or blocks or the combination of blocks.

[0130] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in Figure 1 one or more of the flows Figure 1 or blocks or the combination of blocks.

[0131] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one or more of the flows Figure 1 or blocks or the combination of blocks.

[0132] As mentioned above, it is only the preferred embodiment of the present invention, and is not used to limit the protection scope of the present invention.

Claims

1. An information processing method, characterized in that, The method includes: When obtaining multiple parameter information of a target object and historical quantity information of the target object, determining multiple initial weight values corresponding to the multiple parameter information according to the multiple parameter information and the historical quantity information; the target object is a commodity; the multiple parameter information is commodity information; the historical quantity information is the historical sales volume of the commodity; Inputting the multiple initial weight values into an analytic hierarchy process model to obtain multiple weight values; Inputting the multiple weight values into a preset convolutional neural network model to obtain a category label corresponding to the target object, and processing the target object based on the processing method corresponding to the category label, where the weight value of each processing layer in the preset convolutional neural network model is obtained based on the parameters of the current processing layer and the next layer; Among them, the determining multiple initial weight values corresponding to the multiple parameter information according to the multiple parameter information and the historical quantity information includes: Combining each parameter information in the multiple parameter information with the historical quantity information respectively to obtain multiple groups of combined information; Performing linear regression processing on the multiple groups of combined information to obtain the multiple initial weight values.

2. The method according to claim 1, characterized in that, The inputting the multiple weight values into a preset convolutional neural network model to obtain a category label corresponding to the target object includes: Inputting the multiple weight values into a preset convolutional neural network model, and using the preset convolutional neural network model to predict the processing quantity of the target object; Determining the category label according to the processing quantity.

3. The method according to claim 1, characterized in that, Before the inputting the multiple weight values into a preset convolutional neural network model to obtain a category label corresponding to the target object, the method further includes: Obtaining the initial weight of the current processing layer of the initial convolutional neural network model, the learning rate of the current processing layer, the learning rate of the next layer, the activation function value of the current processing layer, and the activation function value of the next layer; Updating the initial weight using the learning rate of the current processing layer, the learning rate of the next layer, the activation function value of the current processing layer, and the activation function value of the next layer to obtain an updated weight; Obtaining the preset convolutional neural network model according to the updated weight.

4. The method according to claim 1, characterized in that, The inputting the multiple initial weight values into an analytic hierarchy process model to obtain multiple weight values includes: Determining an initial matrix according to the multiple initial weight values; Determining the logarithm of each element in the initial matrix to obtain an anti-symmetric matrix; Comparing the importance between any two elements in the anti-symmetric matrix according to the three-scale rule to obtain a comparison matrix; Determining the optimal transfer matrix of the comparison matrix, and obtaining the multiple weight values according to the optimal transfer matrix.

5. The method according to claim 1, characterized in that, Before the determining multiple initial weight values corresponding to the multiple parameter information according to the multiple parameter information and the historical quantity information, the method further includes: Obtaining the initial historical quantity information of the target object; Performing denoising processing on the initial historical quantity information to obtain the historical quantity information.

6. The method according to claim 1, characterized in that, After the inputting the multiple weight values into a preset convolutional neural network model to obtain a category label corresponding to the target object, the method further includes: Determine the target parameter information corresponding to the category label; the target parameter information is one of the multiple parameter information; Output the category label and the target parameter information.

7. An information processing device, characterized in that, The device includes: A determination unit, configured to, when obtaining multiple parameter information of a target object and historical quantity information of the target object, determine multiple initial weight values corresponding to the multiple parameter information according to the multiple parameter information and the historical quantity information; the target object is a commodity; the multiple parameter information is commodity information; the historical quantity information is the historical sales volume of the commodity; An input unit, configured to input the multiple initial weight values into an analytic hierarchy process model to obtain multiple weight values; input the multiple weight values into a preset convolutional neural network model to obtain a category label corresponding to the target object, and process the target object based on a processing method corresponding to the category label, where the weight value of each processing layer in the convolutional neural network model is obtained based on the parameters of the current processing layer and the next layer; Wherein, the device further includes: A combination unit, configured to combine each of the multiple parameter information with the historical quantity information respectively to obtain multiple groups of combined information; A processing unit, configured to perform linear regression processing on the multiple groups of combined information to obtain the multiple initial weight values.

8. An information processing device, characterized in that, The device includes: A memory, a processor, and a communication bus, where the memory communicates with the processor through the communication bus, and the memory stores a program for information processing executable by the processor. When the program for information processing is executed, the method according to any one of claims 1 to 6 is executed through the processor.

9. A storage medium, on which a computer program is stored, is applied to an information processing device, and is characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images