Product cost prediction method, model training method, electronic device, and storage medium

Abstract

The embodiments of the present application relate to the technical field of communication. Disclosed are a product cost prediction method, a model training method, an electronic device, and a storage medium, which can predict a predicted price value of a target product. The product cost prediction method comprises: on the basis of the arrangement order of M production stages of a target product, determining input parameters of a product price prediction model corresponding to each of the M production stages; for each production stage, inputting the input parameters of the production stage into the product price prediction model corresponding to the production stage, and determining a predicted reference price value of the production stage, so as to obtain the reference price value of each of the M production stages; and on the basis of the reference price value of each of the M production stages, determining a predicted price value of the target product.

Description

Product cost prediction method, model training method, electronic device and storage medium Technical Field

[0001] The present disclosure relates to the field of communication technology, and in particular to a product cost prediction method, a model training method, an electronic device, and a storage medium. Background Art

[0002] Currently, display manufacturing consists of several distinct stages. For example, LCD panel manufacturing involves the array, color filter (Cf), cell, module, and packaging stages. Product development requires factoring in the costs of each manufacturing stage to estimate overall product cost, a complex calculation process.

[0003] Summary of the Invention

[0004] On the one hand, a product cost prediction method, a model training method, an electronic device and a storage medium are provided, which can accurately predict the price of a product at different stages and determine the predicted price value of the product.

[0005] The product cost prediction method includes: the electronic device can determine the input parameters of the product price prediction model corresponding to each of the M production stages based on the arrangement order of the M production stages of the target product, and for each production stage, input the input parameters of the production stage into the product price prediction model corresponding to the production stage to determine the reference price value predicted by the production stage to obtain the reference price value of each of the M production stages. Then, the electronic device determines the predicted price value of the target product based on the reference price value of each of the M production stages.

[0006] Among them, each production stage corresponds to a product price prediction model, and the input parameters of the product price prediction model corresponding to the Nth production stage include the product parameters of the target product, and the reference price values ​​predicted by the product cost prediction models corresponding to all production stages arranged in an order before the Nth production stage; among them, M is greater than or equal to N, and M and N are positive integers.

[0007] In view of this, an embodiment of the present application provides a product cost prediction method, which constructs an arrangement order of production stages based on a correlation coefficient matrix, and each time the reference price of the next production stage is predicted, the reference price values ​​of all production stages before this production stage are added as features to the input set, and the reference price value of each production stage is predicted using a price prediction model, thereby avoiding the problem of not considering the existence of a certain relationship between the prices of multiple stages and the sequential nature of the algorithm, which in turn leads to low accuracy in predicting prices of different stages of products.

[0008] In some embodiments, the method further includes: the electronic device obtaining product parameters of a target product and product parameters of multiple reference products, and determining, for each reference product, a similarity between the reference product and the target product based on the product parameters of the target product and the reference products, to obtain a similarity between each of the multiple reference products and the target product. Furthermore, the electronic device may determine the reference product with the highest similarity to the target product among the multiple reference products as the first product, wherein the price value of the first product is used to determine whether there is an anomaly in the price value of the target product.

[0009] Based on the above technical solution, the electronic device in the embodiment of the present application determines that the reference product with the highest similarity to the target product is the first product. Based on the price of the first product, the price value of the target product predicted by the price prediction model can be further verified to check the feasibility of the predicted price value of the target product.

[0010] In some embodiments, the product price prediction model corresponding to each production stage is different.

[0011] In some embodiments, the predicted price value of the target product is the sum of the reference price values ​​of each production stage in the M production stages.

[0012] In some embodiments, the product parameters include at least one of the following: product category; product size; product shape; resolution.

[0013] In another aspect, a model training method is provided, comprising: an electronic device determining an order of the M production stages of a target product based on the product cost prices of the M production stages, and determining a data set for a neural network model corresponding to each of the M production stages based on the order of the M production stages. Furthermore, for each production stage, the neural network model for that production stage is iteratively trained based on the data set for that production stage until a loss value of the neural network model falls below a preset threshold, and the neural network model is then used as a product price prediction model for that production stage.

[0014] The product cost prices for the M production stages are the actual product cost prices for products of the same type as the target product during the corresponding production stages. Each production stage corresponds to a neural network model. The dataset corresponding to the Nth production stage includes the product parameters of the same type of product, the product cost prices of the same type of product during the Nth production stage, and the reference price values ​​predicted by the neural network model for all production stages prior to the Nth production stage. M is greater than or equal to N, and both M and N are positive integers.

[0015] In this embodiment of the present application, the dataset includes a training set and a test set. The training set includes product parameters of the same type of products and reference price values ​​predicted by the neural network model for all production stages prior to the Nth production stage. The test set includes the product cost price of the same type of products at the Nth production stage.

[0016] In an embodiment of the present application, the loss value of the product price prediction model is lower than a preset threshold, and the loss value is used to characterize the error value between the test set and the output result; the output result is the reference price value of the production stage predicted by the neural network model based on the training set; the product price prediction model is used to predict the reference price value of the target production stage.

[0017] In view of this, an embodiment of the present application provides a model training method in which an electronic device determines the order of the production stages by determining a correlation coefficient matrix for M production stages, fully considering the correlation between any two production stages in the M production stages. It is understood that the higher the correlation coefficient between any two production stages, the greater the correlation and mutual influence between the two production stages.

[0018] Since the prices between two adjacent production stages affect each other, the electronic device can determine that the training set corresponding to the Nth production stage among the M production stages is the product parameters and the reference price value of the production stage before the Nth production stage, so as to further improve the accuracy of model training.

[0019] In some embodiments, the arrangement order of the M production stages of the target product is determined based on the product cost prices of the M production stages, including: the electronic device can determine the correlation coefficient matrix of the M production stages through a covariance algorithm based on the product cost prices of the M production stages, and determine the arrangement order of the M production stages based on the maximum correlation coefficient in the correlation coefficient matrix.

[0020] In an embodiment of the present application, the correlation coefficient matrix includes multiple correlation coefficients, and any correlation coefficient is determined based on the product cost prices of any two production stages in the M production stages.

[0021] Based on the above technical solution, the electronic device in the embodiment of the present application can determine the arrangement order of the production stages by determining the correlation coefficient matrix of M production stages, and fully consider the correlation and mutual influence between any two production stages in the M production stages.

[0022] In some embodiments, based on the maximum correlation coefficient in the correlation coefficient matrix, the arrangement order of the M production stages is determined, including: the electronic device can also determine the first production stage and the second production stage corresponding to the maximum correlation coefficient, and determine the target production stage. Further, the electronic device can determine the arrangement order of the M production stages based on the first production stage, the second production stage, and the target production stage.

[0023] The first production stage is the production stage ranked first among the M production stages, and the second production stage is the production stage ranked second among the M production stages. The target production stage is the production stage after the second production stage. The row order of the target production stage in the correlation coefficient matrix is ​​the column order of the second production stage in the correlation coefficient matrix. The correlation coefficient between the second production stage and the target production stage is the largest correlation coefficient in the row order.

[0024] Based on the above technical solution, the higher the correlation coefficient between any two production stages in the embodiment of the present application, the greater the correlation and mutual influence between the two production stages. Therefore, the electronic device can determine the arrangement order of the M production stages through the correlation coefficient matrix of the M production stages, thereby improving the subsequent prediction results of the product.

[0025] In some embodiments, the column order of the target production stage is not equal to the row order of the first production stage, and the column order of the target production stage is not equal to the column order of the second production stage.

[0026] The column order is used to represent the column where the production stage is located in the correlation coefficient matrix, and the row order is used to represent the row where the production stage is located in the correlation coefficient matrix.

[0027] In some embodiments, the product parameters include at least one of the following: product category; product size; product shape; resolution.

[0028] In some embodiments, the above method also includes: the electronic device can obtain the original product parameters of the same type of products, and perform data processing on the original product parameters to determine the product parameters of the same type of products; data processing includes data preprocessing and feature engineering.

[0029] In the present embodiment, data preprocessing includes at least one of removing, correcting, and supplementing abnormal parameters from product parameters of similar products. Feature engineering is the process of converting raw product parameters of similar products after data preprocessing into product parameters of similar products.

[0030] In some embodiments, for each production stage, the neural network model of the production stage is iteratively trained based on the data set of the production stage until the loss value of the neural network model is lower than a preset threshold, and the neural network model is used as the product price prediction model for the production stage, including: the electronic device inputs the training set into the neural network model to determine the output result of the neural network model; and determines the loss value of the neural network model based on the output result and the test set.

[0031] Furthermore, the electronic device needs to determine whether the loss value of the neural network model is below a preset threshold. If the loss value is above the preset threshold, the electronic device can adjust the model parameters of the neural network model. If the loss value is below the preset threshold, the electronic device can use the neural network model as a product price prediction model.

[0032] Based on the above technical solution, in the embodiment of the present application, when the loss value is higher than the preset threshold, the electronic device iteratively adjusts the model parameters of the neural network model until the loss value is lower than the preset threshold and the conditions are met, and then the current neural network model is used as a prediction model to improve the reliability of subsequent product price predictions.

[0033] In yet another aspect, an electronic device is provided, comprising a processor and a communication interface. The communication interface is coupled to the processor. The processor is configured to execute a computer program or instructions to implement the product cost prediction method or model training method of the first aspect or any embodiment of the first aspect.

[0034] In another aspect, a computer-readable storage medium is provided. The computer-readable storage medium stores computer program instructions, which, when executed on a computer (e.g., a receiving node), cause the computer to execute the product cost prediction method or model training method according to any of the above embodiments.

[0035] In another aspect, a computer program product is provided, comprising computer program instructions, which, when executed on a computer (eg, a receiving node), cause the computer to perform the product cost prediction method or model training method according to any of the above embodiments.

[0036] In another aspect, a computer program is provided. When the computer program is executed on a computer (eg, a receiving node), the computer program causes the computer to execute the product cost prediction method or model training method according to any one of the above embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] To more clearly illustrate the technical solutions of the present disclosure, the following briefly introduces the drawings required for use in some embodiments of the present disclosure. Obviously, the drawings described below are only drawings of some embodiments of the present disclosure, and those skilled in the art can also derive other drawings based on these drawings. Furthermore, the drawings described below are schematic diagrams and are not intended to limit the actual dimensions of the products, actual processes of the methods, actual timing of signals, and the like involved in the embodiments of the present disclosure.

[0038] FIG1 is a block diagram of an electronic device according to some embodiments;

[0039] FIG2 is a flow chart of a model training method according to some embodiments;

[0040] FIG3 is a flowchart of a method for predicting product costs according to some embodiments;

[0041] FIG4 is a flowchart of a product cost prediction method according to some other embodiments;

[0042] FIG5 is a schematic diagram of contribution analysis of product parameters according to some embodiments;

[0043] FIG6 is a schematic diagram of a scenario of a model training method according to some embodiments;

[0044] FIG7 is a schematic diagram of a scenario of a product cost prediction method according to some embodiments;

[0045] FIG8 is a structural diagram of a product cost prediction device according to some embodiments;

[0046] FIG9 is a block diagram of a model training apparatus according to some embodiments;

[0047] FIG10 is a block diagram of an electronic device according to some embodiments. DETAILED DESCRIPTION

[0048] The following will be combined with the accompanying drawings to clearly and completely describe the technical solutions in some embodiments of the present disclosure. Obviously, the embodiments described are only some embodiments of the present disclosure, not all embodiments. Based on the embodiments provided by the present disclosure, all other embodiments obtained by ordinary technicians in this field are within the scope of protection of the present disclosure.

[0049] Unless the context requires otherwise, throughout the specification and claims, the term "comprise" and its other forms, such as the third person singular form "comprises" and the present participle form "comprising", are to be interpreted as open and inclusive, that is, "including, but not limited to". In the description of the specification, the terms "one embodiment", "some embodiments", "exemplary embodiments", "example", "specific example" or "some examples" are intended to indicate that the specific features, structures, materials or characteristics associated with the embodiment or example are included in at least one embodiment or example of the present disclosure. The schematic representation of the above terms does not necessarily refer to the same embodiment or example. In addition, the specific features, structures, materials or characteristics may be included in any one or more embodiments or examples in any appropriate manner.

[0050] In the following, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the embodiments of the present disclosure, unless otherwise specified, "plurality" means two or more.

[0051] “At least one of A, B and C” has the same meaning as “at least one of A, B or C” and both include the following combinations of A, B and C: A only, B only, C only, the combination of A and B, the combination of A and C, the combination of B and C, and the combination of A, B and C.

[0052] “A and / or B” includes the following three combinations: A only, B only, and a combination of A and B.

[0053] As used herein, the term "if" is optionally interpreted to mean "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrases "if it is determined that" or "if [stated condition or event] is detected" are optionally interpreted to mean "upon determining" or "in response to determining" or "upon detecting [stated condition or event]" or "in response to detecting [stated condition or event]," depending on the context.

[0054] The use of "adapted to" or "configured to" herein is intended to be open and inclusive language that does not exclude devices adapted or configured to perform additional tasks or steps.

[0055] Additionally, the use of “based on” is intended to be open and inclusive, as a process, step, calculation, or other action “based on” one or more conditions or values ​​may, in practice, be based on additional conditions or beyond values.

[0056] As used herein, “about,” “substantially,” or “approximately” includes the stated value and an average value that is within an acceptable range of deviation from the particular value, where the acceptable range of deviation is determined by one of ordinary skill in the art taking into account the measurements in question and the errors associated with the measurement of the particular quantity (i.e., the limitations of the measurement system).

[0057] As used herein, "equal" includes the stated conditions and conditions similar to the stated conditions, where the range of the similar conditions is within an acceptable range of deviation, where the acceptable range of deviation is determined by one of ordinary skill in the art taking into account the measurement in question and the errors associated with the measurement of the particular quantity (i.e., the limitations of the measurement system). "Equal" includes absolute equality and approximate equality, where the acceptable range of deviation for approximate equality can be, for example, that the difference between the two is less than or equal to 5% of either.

[0058] The following explains the terms involved in the embodiments of the present application to facilitate readers' understanding.

[0059] 1. Liquid Crystal Display (LCD) is a type of flat-panel display. Its structure consists of a liquid crystal cell placed between two parallel glass substrates. TFTs (thin-film transistors) are located on the lower glass substrate, and color filters are located on the upper glass substrate. The rotation direction of the liquid crystal molecules is controlled by changing the signal and voltage on the TFTs, thereby controlling the polarization of light emitted from each pixel and achieving the display purpose.

[0060] 2. The array process, also known as alignment substrate design, is the fundamental material for PCB manufacturing. Typically, the substrate is based on a thin, cored copper-clad laminate. Conductive pattern layers and prepregs are alternately laminated and bonded together in a single process, forming at least three interconnected conductive pattern layers. Alignment substrate design involves rationally arranging the elements on the substrate according to appropriate design principles to achieve the desired circuit functionality.

[0061] 3. The color filter substrate (CF) is a key component of the display panel. Backlight passing through the liquid crystal layer or white light emitted by the organic electroluminescent device (OLED) passes through the array of red, green, and blue color resists on the CF substrate, emitting red, green, and blue light, achieving color display on the display panel. To prevent crosstalk between light transmitted through adjacent color resists, the CF substrate is typically provided with a black matrix (BM). The CF substrate is formed with a plurality of spaced-apart hollow areas arranged in an array. Adjacent color resists are coated within the hollow areas of the black matrix to prevent crosstalk.

[0062] 4. Color filter on array (COA) technology refers to a technique used in LCD manufacturing that integrates color filters with array substrates. This involves applying color photoresist onto the completed array to form a color filter layer, which improves the low aperture ratio of traditional color filters.

[0063] 5. In the cell forming process, the first step is to coat the alignment layer film, then apply liquid crystal on the CF substrate, and finally apply edge-sealing UV glue on the TFT substrate, which means cutting the outer edge size of the LCD screen.

[0064] 6. An LCD display module (LCM) is an assembly that combines a liquid crystal display device, connectors, peripheral circuits such as control and drive circuits, a printed circuit board (PCB), a backlight source, and structural components. During the module assembly process, anisotropic conductive adhesive is first applied and the FPC is bonded to the LCD panel. The driver board (PCB) is then bonded to the plastic frame using conductive adhesive. Finally, the module is bonded to the plastic frame using adhesive material.

[0065] 7. Each element of the covariance matrix is ​​the covariance between the elements of a vector. It is a natural generalization from scalar random variables to high-dimensional random vectors. The following is a brief overview of the origin of the formula for the covariance matrix.

[0066] a. Variance

[0067] Variance is used to measure the degree of dispersion of a single random variable. Its calculation formula 1 is:

[0068] Where n represents the sample size, Represents the mean of the sample.

[0069] b. Covariance

[0070] Covariance is used to characterize the similarity between two random variables, and its calculation formula 2 is defined as:

[0071] in, Respectively represent the mean of the observed samples corresponding to the two random variables. Variance It can be viewed as the covariance σ(x,x) of the random variable x with respect to itself.

[0072] c. Covariance matrix

[0073] Given d random variables x k ,k=1,2,…,d, the variance of the random variable is given by the following formula 3:

[0074] Among them, x ki represents the random variable x k The i-th observation sample of n represents the sample size, and the number of observation samples of each random variable is n. The covariance between two random variables is formula 4:

[0075] Therefore, the covariance matrix is ​​Equation 5:

[0076] Among them, the elements on the diagonal are the variances of each random variable, the elements on the off-diagonal are the covariances between any two random variables, and the matrix ∑ is a symmetric matrix with a size of d×d.

[0077] The above is a brief introduction to the related technologies of this application.

[0078] Currently, display panel production consists of several distinct stages, including the array process, color filter (Cf) process, cell process, module process, and packaging process. Each model building method can be tailored to the specific stage of the product.

[0079] In traditional technologies, there are several types of model training methods and price prediction methods.

[0080] Type 1: Spectral clustering price index dynamic optimization integrated prediction method. This method can establish a dynamic selective integration strategy through sliding window technology and multi-objective programming theory, and use the spectral clustering method to analyze the data characteristics between observation points. It selectively selects the observation point with the most similar data structure to the current observation point as the basis for generating the optimal integration strategy, and then determines the optimal integration weight to complete the spectral clustering model.

[0081] Type II: A multi-objective model predictive control method for hierarchical optimization of regional integrated energy systems. This method sets the maximum renewable energy absorption capacity and comprehensive energy efficiency within the regional integrated energy system's scheduling cycle as the upper-level objective function, and sets the minimum operating cost and energy cost as the lower-level objective function. Within each scheduling cycle, the distributed wind and solar output power and load in the regional integrated energy system are predicted, and the prediction error is corrected based on the system's real-time status to obtain the predicted value. Furthermore, the scheduling constraints, upper-level objective function, lower-level objective function, and predicted value are solved using the model predictive control method for online rolling hierarchical optimization scheduling. The rolling optimization solution is obtained, and the scheduling plan for the regional integrated energy system within a scheduling period is output.

[0082] Type 3: Prediction model training method, product cost forecasting method. This method obtains multiple factors influencing the price of a target product and the price of the target product at multiple times. It also obtains the factor data corresponding to each factor at each time from a preset website. Furthermore, a preset algorithm is used to train the product price at each time and the factor data corresponding to each factor at each time to obtain a price prediction model.

[0083] Type 4: Product cost prediction method. This method trains multiple models h1, h2, h3, h4, and h5 for different stages. Each model uses the same dataset to determine prices at different stages. Specifically, this method performs K-fold cross-validation on model training and testing. This involves splitting the dataset into k parts, taking k-1 parts as the training set and 1 part as the test set. This is used to train neural network models corresponding to different stages. After the neural network model is trained, it is used to predict product prices at different stages.

[0084] The above-mentioned technical solutions do not take into account the existence of a certain relationship between prices at multiple stages and the nature of the order of the algorithm, which leads to the problem of low accuracy in predicting prices of products at different stages.

[0085] In view of this, an embodiment of the present application provides a product cost prediction method, which constructs an arrangement order of production stages based on a correlation coefficient matrix, and each time the reference price of the next production stage is predicted, the reference price values ​​of all production stages before this production stage are added as features to the input set, and the reference price value of each production stage is predicted using a price prediction model, thereby avoiding the problem of not considering the existence of a certain relationship between the prices of multiple stages and the sequential nature of the algorithm, which in turn leads to low accuracy in predicting prices of different stages of products.

[0086] The following will describe in detail the implementation of the embodiment of the present application in conjunction with the accompanying drawings.

[0087] As shown in Figure 1, Figure 1 is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device can be a model training device or a product cost prediction device. As shown in Figure 1, the electronic device 100 includes a display 201, at least one processor 202, a transceiver 203, and may also include a memory 204. The processor 202, the memory 204, and the transceiver 203 can be connected via a communication line.

[0088] In the embodiment of the present application, the processor 202 can be a chip. The chips can include five categories: logic chips, memory chips, sensor chips, power chips and communication chips. Among them, the processor class mainly undertakes chips for specific calculation and control tasks in the system, such as microcontroller unit (MCU), central processing unit (CPU), graphics processing unit (GPU), neural processing unit (NPU), etc. The storage class mainly undertakes chips for data storage in the system, as well as some storage controller chips, such as dynamic random access memory (DRAM), static random access memory (SRAM), flash eeprom memory (Flash), etc. The sensor class mainly undertakes chips for information collection, presentation and interaction in the system, such as input and output devices, some signal processing chips, etc. Communications chips (wired and wireless) mainly undertake communication functions in the system, such as some Ethernet chips, switching chips, wide area and local area network, point-to-point and ad hoc network chips, as well as filtering, amplification, power and other devices that assist in communication. Commonly known products include wireless fidelity (WiFi), Bluetooth, fifth-generation mobile communication technology (5G) baseband, global positioning system (GPS), narrowband internet of things (NB-IoT), network cards, switches, etc.

[0089] The communication line may include a channel for transmitting information between the above components.

[0090] The memory 204 may be a read-only memory (ROM) or other type of static storage device that can store static information and instructions, a random access memory (RAM) or other type of dynamic storage device that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, an optical disc storage (including a compact disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to contain or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited to these.

[0091] In one possible design, the memory 204 can exist independently of the processor 202, that is, the memory 204 can be a memory external to the processor 202. In this case, the memory 204 can be connected to the processor 202 via the communication line 102, and is used to store execution instructions or application code, and the execution is controlled by the processor 202 to implement the network quality determination method provided in the following embodiment of this application. In another possible design, the memory 204 can also be integrated with the processor 202, that is, the memory 204 can be the internal memory of the processor 202. For example, the memory 204 is a cache that can be used to temporarily store some data and instruction information.

[0092] As an implementation manner, the processor 202 may include one or more CPUs.

[0093] It should be noted that the electronic device described in the embodiment of the present application is intended to more clearly illustrate the technical solution of the embodiment of the present application, and does not constitute a limitation on the technical solution provided in the embodiment of the present application. Ordinary technicians in this field can know that with the evolution of electronic devices and the emergence of other electronic devices, the technical solution provided in the embodiment of the present application is also applicable to similar technical problems.

[0094] The methods in the following embodiments can all be implemented in the electronic device 100 having the above hardware structure. The methods in the embodiments of the present application are described below.

[0095] The model training method and product cost prediction method provided in the embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0096] The embodiment of the present application can train the neural network model to obtain a price prediction model of the product cost prediction method. The following is a detailed introduction to a model training method provided by the embodiment of the present application in conjunction with the accompanying drawings. As shown in Figure 2, the model training method may include S201-S203. Among them, S201 can also be called the "determining the arrangement order of M production stages" process, S202 can also be called the "determining the data sets corresponding to different production stages" process, and S203 can be called the "determining the product price prediction model for different production stages" process. The following is a detailed description of S201-S203.

[0097] S201. Determine the arrangement order of the M production stages of the target product based on the product cost prices of the M production stages.

[0098] In the embodiment of the present application, the product cost prices of the M production stages are the actual product cost prices of the same type of products of the target product in the corresponding production stages.

[0099] In the embodiments of the present application, the target product can be understood as the product for which price prediction is required. The M production stages of the target product include, but are not limited to, the array process stage, the color film (Cf) process stage, the cell process stage, the module process stage, and the packaging process stage.

[0100] In one possible implementation, the electronic device can obtain the original product parameters of the same type of products and the product cost prices of M production stages (test set), and perform data preprocessing on the product parameters of the same type of products to determine the product parameters of the same type of products.

[0101] Optionally, data processing includes data preprocessing and feature engineering. Data preprocessing includes at least one of eliminating, correcting, and supplementing abnormal parameters from product parameters of similar products. Feature engineering is the process of converting raw product parameters of similar products after data preprocessing into product parameters of similar products.

[0102] In the embodiment of the present application, product parameters include but are not limited to product category, product size, product form, and resolution.

[0103] For example, as shown in Table 1, the electronic device obtains product parameters of the same type of products and prices of M production stages. For example, the product parameters of the same type of products obtained by the electronic device include 30+ product attributes, that is, 30+ product features.

[0104] Table 1

[0105] In this embodiment of the present application, after obtaining product parameters for similar products and product cost prices across M production stages (a test set), the electronic device preprocesses the data. First, it filters out abnormal data values ​​whose sizes and prices fall outside a reasonable range. Then, it corrects these filtered abnormal data values ​​to ensure maximum utilization of the data information. Furthermore, the electronic device needs to supplement any missing fields in Table 1 above.

[0106] Furthermore, the electronic device needs to perform feature engineering on the pre-processed product parameters of the same type of products and the product cost prices (test set) of M production stages. Specifically, the electronic device first eliminates data with the same attribute values ​​in the data set (Table 1) (data with unique attributes), as well as data with small attribute changes in the data set, or data with little impact on the model prediction results (data with low variance attributes). Secondly, the data information is extracted from the eliminated data to create new features. That is, the original eliminated data includes 30+ product features. These 30+ product features are refined to determine 90+ product features. Then, the created product features are feature encoded to facilitate model calculation. For example, the product forms are A, B, and C, which can be encoded as 0, 1, and 2 using labelEncoder.

[0107] Optionally, the encoding method may be onehot or WOE.

[0108] The electronic device combines the encoded product features, creating new features by combining the Cartesian products of the existing attributes. For example, if the two attributes of size 10 and product form A are encoded as 1 and 2, respectively, the feature combination could be: 1-2. For example, if the two attributes of product form A are encoded as 1 and 2, respectively, the second-order polynomial features are 1, 2, and 4, which is the square of each attribute value multiplied by the attribute value.

[0109] The electronic device in the embodiment of the present application obtains the product parameters of the same type of products and the price values ​​of M production stages in the above manner, and determines the arrangement order of the M production stages of the target product through the product cost prices of the M production stages.

[0110] In an embodiment of the present application, the electronic device can determine the correlation coefficient matrix of the M production stages based on the product cost prices of the M production stages through a covariance algorithm, and determine the arrangement order of the M production stages based on the maximum correlation coefficient in the correlation coefficient matrix.

[0111] The correlation coefficient matrix includes multiple correlation coefficients, and any correlation coefficient is determined based on the product cost prices of any two production stages in the M production stages.

[0112] For example, among the M production stages, the array process stage is y1, the color film (Cf) process stage is y2, the cell process stage is y3, the module process stage is y4, and the packaging process stage is y5. The order of the current M production stages is y1, y2, y3, y4, y5, and the product cost price corresponding to each production stage is x1, x2, x3, x4, x5. All the arrangements of the above five production stages are kind.

[0113] Substituting the product cost prices of the M production stages into the above formula 5, the following formula 6 is obtained:

[0114] Among them, x1 is the price value of the first production stage y1, x2 is the price value of the second production stage y2, and x5 is the price value of the fifth production stage y5.

[0115] After obtaining Formula 6, the electronic device calculates the correlation coefficient matrix of the M production stages according to Formula 6, and obtains the following Formula 7:

[0116] in,

[0117] In this way, the electronic device can determine the maximum correlation coefficient in the correlation coefficient matrix, and determine the arrangement order of the M production stages according to the maximum correlation coefficient.

[0118] It can be understood that, by using the maximum correlation coefficient, the correlation between two adjacent production stages in the order of the M production stages is the highest, so as to improve the subsequent prediction results of the product.

[0119] In one implementation, the electronic device determines the first production stage and the second production stage corresponding to the maximum correlation coefficient, and determines the target production stage after the second production stage, and then determines the arrangement order of the M production stages based on the first production stage, the second production stage, and the target production stage.

[0120] The first production stage is the production stage ranked first among the M production stages, and the second production stage is the production stage ranked second among the M production stages. The row order of the target production stage in the correlation coefficient matrix is ​​the column order of the second production stage in the correlation coefficient matrix, and the correlation coefficient between the second production stage and the target production stage is the largest correlation coefficient in the row order.

[0121] For example, the electronic device can determine from the correlation coefficient matrix that the first production stage corresponding to the maximum correlation coefficient is the production stage ranked first among the M production stages, and determine that the second production stage corresponding to the maximum correlation coefficient is the production stage ranked second among the M production stages. 5×5 ) in the maximum correlation coefficient ρ(y i ,y j ), then the first production stage is y i , the second production stage is y j .

[0122] After the electronic device determines the second production stage, the row with the same column number as the second production stage is determined from the correlation coefficient matrix, and the second production stage corresponding to the maximum correlation coefficient in the row is determined as the target production stage. For example, the maximum correlation coefficient ρ(y j ,y k ), then the column where the value is located is the third production stage y k .

[0123] It should be noted that the column order of the target production stage is not equal to the row order of the first production stage, and the column order of the target production stage is not equal to the column order of the second production stage. k In , k≠i and k≠j.

[0124] The column order is used to represent the column where the production stage is located in the correlation coefficient matrix, and the row order is used to represent the row where the production stage is located in the correlation coefficient matrix.

[0125] For example, the arrangement order of the M production stages determined above may specifically be any one of the following cases (1) and (2).

[0126] As shown in Table 2 below, x1 to x5 in the table are the product cost prices corresponding to M production stages, and the values ​​in brackets are the correlation coefficients between any two production stages among the M production stages.

[0127] Case (1): Determine the maximum correlation coefficient in the correlation coefficient matrix, ρ(x1, x3) in Table 2 below. The maximum correlation coefficient ρ(x1, x3) in the correlation coefficient matrix for the M production stages is 70, corresponding to the product cost prices x1 and x3. Accordingly, the production stages corresponding to the maximum correlation coefficient are y1 (array process stage) and y3 (cell stage).

[0128] The electronic device uses y1 as the first-ranked production stage among the M production stages, and y3 as the second-ranked production stage among the M production stages.

[0129] Among them, the column number of the second-ranked production stage y3 is 3, so the electronic device selects the maximum correlation coefficient from the row with the same value as the column number, that is, selects the maximum correlation coefficient from the 3rd row.

[0130] The maximum correlation coefficient ρ(x3,x1) in row 3 is 70, corresponding to the product cost prices x3 and x1. Accordingly, the production stages y3 and y1 are associated with the maximum correlation coefficient in row 3, but the order of y3 and y1 is already determined. Therefore, after eliminating this coefficient, the maximum correlation coefficient in row 3 is ρ(x3,x4), which is 67, corresponding to the production stages y3 and y4. For electronic devices, y4 (the module process stage) is the third production stage in the order.

[0131] Among them, the column number of the production stage y4 ranked third is 4. Then, the electronic device selects the maximum correlation coefficient from the row with the same value as the column number, that is, selects the maximum correlation coefficient from the 4th row.

[0132] The maximum correlation coefficient ρ(x4, x5) in row 4 is 69, corresponding to the product cost prices x4 and x5. Accordingly, the production stages y4 and y5 are associated with the maximum correlation coefficient in row 4. Therefore, the electronic device ranks y5 (the packaging process) as the fourth production stage in the sorting order.

[0133] Among them, the column number of the fourth production stage y5 is 5. Then, the electronic device selects the maximum correlation coefficient from the row with the same value as the column number, that is, selects the maximum correlation coefficient from the 5th row.

[0134] The maximum correlation coefficient ρ(x5,x4) in row 5 is 69, corresponding to the product cost prices x5 and x4. Accordingly, the production stages y5 and y4 are associated with the maximum correlation coefficient in row 5, but the order of y5 and y4 is already determined. Therefore, after eliminating this coefficient, the maximum correlation coefficient in row 5 is determined to be ρ(x5,x2), which is 68, corresponding to the production stages y5 and y2. Electronic devices place y2 (the color film (Cf) process) as the fifth production stage in the order.

[0135] In summary, the order of the M production stages for an electronic device is: y1, y3, y4, y5, y2. In other words, the order of the M production stages is: array process, cell process, module process, packaging process, and color film (Cf) process.

[0136] Table 2

[0137] It should be noted that, as shown in Table 2 above, the maximum correlation coefficients in the correlation coefficient matrix of the M production stages include ρ(x1, x3) and ρ(x3, x1). The above case (1) provides a detailed description of the case where ρ(x1, x3) is determined as the maximum correlation coefficient in the correlation coefficient matrix. The following details the determination of ρ(x3, x1) as the maximum correlation coefficient in the correlation coefficient matrix.

[0138] Case (2): Determine the maximum correlation coefficient in the correlation coefficient matrix, ρ(x3, x1) in Table 2. The maximum correlation coefficient ρ(x3, x1) in the correlation coefficient matrix for the M production stages is 70, corresponding to the product cost prices x3 and x1. Accordingly, the production stages corresponding to the maximum correlation coefficient are y3 (cell stage) and y1 (array process stage).

[0139] The electronic device uses y3 (cell stage) as the first production stage among M production stages, and y1 (array process stage) as the second production stage among M production stages.

[0140] The column number of the second-ranked production stage y1 is 1, and the electronic device selects the maximum correlation coefficient from the row with the same value as the column number, that is, selects the maximum correlation coefficient from the first row.

[0141] The maximum correlation coefficient ρ(x1,x3) in row 1 is 70, corresponding to the product cost prices x1 and x3. Accordingly, the production stages y1 and y3 are associated with the maximum correlation coefficient in row 1, but the order of y1 and y3 is already determined. Therefore, after eliminating this coefficient, the maximum correlation coefficient in row 1 is determined to be ρ(x1,x2), which is 60, corresponding to the production stages y1 and y2. Electronic devices have y2 (the color film (Cf) process) as the third production stage in the order.

[0142] Among them, the column number of the production stage y2 ranked third is 2. Then, the electronic device selects the maximum correlation coefficient from the row with the same value as the column number, that is, selects the maximum correlation coefficient from the second row.

[0143] The maximum correlation coefficient ρ(x2,x5) in row 2 is 68, corresponding to the product cost prices x2 and x5. Accordingly, the production stages y2 and y5 are associated with the maximum correlation coefficient in row 2. Therefore, the electronic device ranks y5 (the packaging process) as the fourth production stage in the ranking order.

[0144] Among them, the column number of the fourth production stage y5 is 5. Then, the electronic device selects the maximum correlation coefficient from the row with the same value as the column number, that is, selects the maximum correlation coefficient from the 5th row.

[0145] The maximum correlation coefficient ρ(x5,x4) in row 5 is 69, corresponding to the product cost prices x5 and x4. Accordingly, the production stages y5 and y4 correspond to the maximum correlation coefficient in row 5. Electronic devices have y4 (the module process stage) as the fifth production stage in the order.

[0146] In summary, the order of the M production stages for an electronic device is: y3, y1, y2, y5, y4. In other words, the order of the M production stages is: cell stage, array process stage, color film (Cf) process stage, packaging process stage, and module process stage.

[0147] It can be understood that the higher the correlation coefficient between any two production stages in the embodiment of the present application, the greater the correlation and mutual influence between the two production stages. Therefore, the electronic device can determine the arrangement order of the M production stages through the correlation coefficient matrix of the M production stages, thereby improving the subsequent prediction results of the product.

[0148] S202. Based on the arrangement order of the M production stages, determine the data set of the neural network model corresponding to each production stage in the M production stages.

[0149] In the embodiments related to the present application, one production stage corresponds to one neural network model, which means that the neural network model corresponding to each production stage can be different. Accordingly, the training set and test set of the neural network model for each production stage are also different. Among them, the data set corresponding to the Nth production stage in the order of arrangement includes product parameters of the same type of products, product cost prices of the same type of products in the Nth production stage, and reference price values ​​predicted by the neural network model corresponding to all production stages in the order of arrangement before the Nth production stage; wherein M is greater than or equal to N, and M and N are positive integers.

[0150] In the embodiment of the present application, the dataset includes a training set and a test set. The training set may include product parameters of the same type of products and reference price values ​​predicted by the neural network model for all production stages prior to the Nth production stage. The test set may include the product cost price for the same type of products at the Nth production stage.

[0151] After the electronic device determines the arrangement order of the M production stages of the target product, the electronic device may determine a training set and a test set corresponding to different production stages.

[0152] For example, the product parameter is N. For example, the order of the M production stages of the target product is y1, y3, y4, y5, and y2, i.e., the array process stage, the cell stage, the module process stage, the packaging process stage, and the color film (Cf) process stage.

[0153] Then, the training set of the neural network model corresponding to the first-ranked array process stage y1 is the product parameter N, while the training set of the neural network model corresponding to the second-ranked cell stage y3 is the product parameter and the reference price value x1 of the array process stage y1.

[0154] Correspondingly, the training set of the neural network model corresponding to the third-ranked module (Module) process stage y4 is the product parameters, the reference price value x3 of the box (Cell) stage y3, and the reference price value x1 of the array (Array) process stage y1.

[0155] The training set of the neural network model corresponding to the fourth-ranked packaging process stage y5 is the product parameters, the reference price value x1 of the array process stage y1, the reference price value x3 of the cell stage y3, and the reference price value x4 of the module process stage y4.

[0156] The training set for the neural network model corresponding to the fifth-ranked color film (Cf) process stage y2 includes product parameters, the reference price value x1 of the array process stage y1, the reference price value x3 of the cell process stage y3, the reference price value x4 of the module process stage y4, and the reference price value x5 of the packaging process stage y5.

[0157] For example, the training set of the neural network model for different production stages is D. Among them, the training set D1 of the neural network model corresponding to the array process stage y1, the training set D2 of the neural network model for the cell process stage y3, the training set D3 of the neural network model corresponding to the module process stage y4, the training set D4 of the neural network model corresponding to the packaging process stage y5, and the training set D5 of the neural network model corresponding to the color film (Cf) process stage y2 are shown in Table 3 below.

[0158] Table 3

[0159] Accordingly, different production stages have different test sets, as shown in Table 4 below.

[0160] Table 4

[0161] S203. For each production stage, iteratively train the neural network model of the production stage based on the data set of the production stage until the loss value of the neural network model is lower than a preset threshold, and use the neural network model as the product price prediction model of the production stage.

[0162] In an embodiment related to the present application, the loss value of the product price prediction model is lower than a preset threshold; the loss value is used to characterize the error value of the test set and the output result; the output result is the reference price value of the production stage predicted by the neural network model based on the training set; the product price prediction model is used to predict the reference price value of the target production stage.

[0163] In an embodiment of the present application, the electronic device can input a training set into a neural network model, determine an output result of the neural network model, and determine a loss value of the neural network model based on the output result and the test set.

[0164] Furthermore, the electronic device determines whether the loss value of the neural network model is lower than a preset threshold. When the loss value of the neural network model is lower than the preset threshold, the electronic device may use the neural network model as a product price prediction model.

[0165] If the loss value of the neural network model is higher than the preset threshold, the electronic device adjusts the model parameters of the neural network model according to the loss value, uses the neural network model with adjusted parameters as the current neural network model, and iteratively executes the above steps until the loss value of the neural network model is lower than the preset threshold.

[0166] For example, taking the arrangement order in case (1) as an example, the product price prediction model for determining the production stage can specifically be any one of the following cases (3) and (4).

[0167] Case (3): When the loss value of the neural network model is higher than a preset threshold, the electronic device adjusts the model parameters of the neural network model.

[0168] In the embodiment of the present application, as shown in Table 5 below, the training set D1 of the array process stage y1 is the product parameter N. The electronic device inputs the product parameter N into the neural network model corresponding to the process stage and determines the output result (reference price value) x1 of the neural network model. Then, the electronic device can compare the output result (reference price value) x1 with the test set Substitute the following formula 8 to determine the loss value E1 of the neural network model. The specific implementation can be:

[0169] Furthermore, if the electronic device determines that the loss value E1 of the neural network model is lower than the preset threshold value S1, the electronic device may use the neural network model as the product price prediction model h1 of the array process stage y1.

[0170] If the electronic device determines that the loss value E1 of the neural network model is higher than the preset threshold S1, the electronic device adjusts the model parameters of the neural network model, uses the neural network model after adjusting the parameters as the current neural network model, and iteratively executes the above steps until the loss value of the neural network model is lower than the preset threshold, and determines the product price prediction model h1 of the array process stage y1.

[0171] Correspondingly, the training set D2 of the cell stage y3 is the product parameter N and the output result (reference price value) x1 of the array process stage y1.

[0172] It can be understood that the product price forecasting model for each subsequent production stage is determined in the same manner as in the above case (3).

[0173] Table 5

[0174] Case (4): When the loss value of the neural network model is higher than a preset threshold, the electronic device adjusts the model parameters and training set of the neural network model.

[0175] Taking the array process stage y1 in the above case (3) as an example, combined with the above Table 5, the training set D1 of the array process stage y1 is the product parameter N. The electronic device inputs the product parameter N into the neural network model corresponding to the process stage and determines the output result (reference price value) x1 of the neural network model. Then, the electronic device can compare the output result (reference price value) x1 with the test set (product cost price) Substitute into the above formula 8 to determine the loss value E1 of the neural network model.

[0176] If the electronic device determines that the loss value E1 of the neural network model is higher than the preset threshold S1, the electronic device adjusts the model parameters and training set of the neural network model and uses the neural network model after adjusting the parameters as the current neural network model.

[0177] Then, as shown in Table 6 below, the electronic device can input the adjusted training set N' into the neural network model with adjusted parameters to determine the output result (reference price value) x1' of the neural network model. Then, according to the output result (reference price value) x1 and the test set Determine the loss value.

[0178] Until the electronic device determines that the loss value of the neural network model is lower than the preset threshold, the product price prediction model Y1 of the array process stage y1 is determined.

[0179] It can be understood that since the training set N of the first-ranked array process stage y1 is adjusted to N′, the corresponding training set of the second-ranked cell stage y3 can also be adjusted to N′x1, and the same method as the array process stage y1 is adopted in subsequent model training.

[0180] Table 6

[0181] Based on the above technical solution, the electronic device in the embodiment of the present application determines the order of the production stages by determining the correlation coefficient matrix of the M production stages, fully considering the correlation between any two production stages in the M production stages. It is understood that the higher the correlation coefficient between any two production stages, the greater the correlation and mutual influence between the two production stages.

[0182] Since the prices between two adjacent production stages affect each other, the electronic device can determine that the training set corresponding to the Nth production stage among the M production stages is the product parameters and the reference price value of the production stage before the Nth production stage, so as to further improve the accuracy of model training.

[0183] The product cost prediction methods in the following embodiments can all be implemented in the electronic device 100 having the above hardware structure. The product cost prediction methods in the embodiments of the present application are described below.

[0184] In an embodiment of the present application, the product cost prediction method can be applied to price prediction for display screens. As shown in Figure 3, the price prediction method may include steps S301-S303. Steps S301 may also be referred to as the "determining input parameters" process, S302 may be referred to as the "determining reference price values ​​for each production stage" process, and S303 may be referred to as the "determining predicted price values ​​for the target product" process. Steps S301-S303 are described in detail below.

[0185] S301. Based on the arrangement order of the M production stages of the target product, determine the input parameters of the product price prediction model corresponding to each of the M production stages.

[0186] In this embodiment of the present application, each production stage corresponds to a product price prediction model. The input parameters of the product price prediction model corresponding to the Nth production stage include the product parameters of the target product and the reference price values ​​predicted by the product cost prediction models corresponding to all production stages prior to the Nth production stage. Where M is greater than or equal to N, and M and N are positive integers.

[0187] In the embodiment of the present application, the order of the M production stages of the target product still follows the order of arrangement in the above-mentioned S201 (1), i.e., y1, y3, y4, y5, y2. Among them, y1 is the array process stage, y3 is the cell stage, y4 is the module process stage, y5 is the packaging process stage, and y2 is the color film (Cf) process stage.

[0188] For example, as shown in Table 7 below, since there are no other production stages before the first ranked array process stage y1, the input parameters of the price prediction model h1 corresponding to the array process stage y1 are the product parameters x q .

[0189] Correspondingly, the input parameter of the price prediction model h3 corresponding to the second-ranked cell stage y3 is the product parameter x q , and the reference price value of the array process stage y1 The input parameters of the price prediction model h4 corresponding to the process stage y4 of the third ranked module are product parameters x q , Reference price value of y1 in the Array process stage And the reference price value of the cell stage y3

[0190] Table 7

[0191] It is understandable that in order to improve the accuracy of model training, it is necessary to determine the sorting order of the M production stages of the target product before model training. Therefore, when predicting the price of the target product, the sorting order of the M production stages of the target product previously determined is still used to further improve the accuracy of the predicted price of the target product.

[0192] It should be noted that the order of determining the M production stages of the target product can be referred to the embodiment in the above S201, which will not be described in detail here.

[0193] S302. For each production stage, input the input parameters of the production stage into the product price prediction model corresponding to the production stage, determine the reference price value predicted for the production stage, and obtain the reference price value for each of the M production stages.

[0194] In conjunction with the example in S301, for the array process stage y1, the electronic device converts the input parameter x q Input the corresponding price prediction model to obtain the reference price value of the array process stage y1

[0195] Furthermore, for the cell stage y3, the electronic device converts the input parameter x q ,as well as Output the corresponding price prediction model to obtain the reference price value of the cell stage y3

[0196] By analogy, the reference price value predicted for each of the M production stages is obtained.

[0197] S303: Determine a predicted price value of the target product based on the reference price value of each of the M production stages.

[0198] In an embodiment of the present application, the predicted price value of the target product is the sum of the reference price values ​​of each production stage in the M production stages.

[0199] Optionally, the price prediction model may be an Xgboost model, and the Xgboost prediction method corresponding to the model may be replaced by a regression algorithm such as random forest, ridge regression, SVR, or regression tree.

[0200] Based on the technical solution of Figure 3 above, the embodiment of the present application constructs an arrangement order of production stages based on a correlation coefficient matrix, and each time the reference price of the next production stage is predicted, the reference price values ​​of all production stages before this production stage are added as features to the input set, and the Xgboost algorithm is used to predict the reference price value of each production stage, thereby avoiding the problem of not considering the certain relationship between the prices of multiple stages and the sequential nature of the algorithm, which in turn leads to low accuracy in predicting prices of products at different stages.

[0201] In the embodiment of the present application, after predicting the target product's price, the predicted price can be further compared with the price of the reference product with the highest attribute similarity to verify the feasibility and reliability of the predicted price of the target product. As shown in FIG4 , the product cost prediction method further includes the following steps S401-S403.

[0202] S401: Obtain product parameters of a target product and product parameters of multiple reference products.

[0203] In the embodiment of the present application, the electronic device can obtain product parameters of similar products (reference products) and product parameters of the target product. The product parameters include but are not limited to: product category, product size, product form, and resolution.

[0204] For example, the following table 8 shows the reference products. It should be noted that the reference products obtained are not limited to the reference products A and B in the following table 8. Similarly, the device references corresponding to different reference products are not limited to the characteristic attributes shown in Table 8.

[0205] Table 8

[0206] S402 : For each reference product, determine the similarity between the reference product and the target product based on product parameters of the target product and the reference product, so as to obtain the similarity between each reference product in the plurality of reference products and the target product.

[0207] In an embodiment of the present application, the electronic device may calculate the similarity between the target product and the reference product based on the product parameters of the target product and the reference product, and the contribution of the product parameters.

[0208] Exemplarily, for each reference product, the electronic device can determine the same / equal product parameters between the reference product and the target product, and use the sum of the contribution values ​​of the same product parameters as the numerator and the sum of the contribution values ​​of all product parameters as the denominator to obtain the similarity between the reference product and the target product.

[0209] It should be noted that the sum of the contributions of all product parameters may be the sum of the contributions of all product parameters of the reference product, or the sum of the contributions of all product parameters of the reference product and the contributions of the product parameters of the target product.

[0210] Among them, the contribution of product parameters refers to the contribution of product parameters to the prediction results, which is used to evaluate the influence of product parameters on the prediction results (price). It can be understood as the change in the prediction result (price) when a product parameter is updated from the current value to another value.

[0211] In the embodiment of the present application, the contribution of each product parameter is built into the model (Xgboost). Optionally, as shown in Figure 5, the SHAP tool can also perform a contribution analysis of product parameters. The SHAP tool's contribution analysis includes but is not limited to: refresh rate-whether it is a touch product, real size-refresh rate, refresh rate-color gamut, display mode / technology-refresh rate, real size-type, whether it is a touch product-demand, refresh rate-brightness, product form-refresh rate, display mode / technology-color gamut, resolution details-color gamut, refresh rate-glass cutting, product form-color gamut, resolution details-brightness, real size-resolution details, product form-brightness, real size-color gamut, product form, product form-glass cutting, product form-bonding process, and resolution details-refresh rate.

[0212] S403: Determine the reference product with the highest similarity to the target product among the multiple reference products as the first product.

[0213] In the embodiment of the present application, the price value of the first product is used to determine whether there is an abnormality in the price value of the target product.

[0214] For example, the electronic device uses reference product A, which has the highest similarity, as the first product and determines the price of the first product. Furthermore, because reference product A is the product with the highest similarity to the target product, the electronic device can compare the price of reference product A with the predicted price of the target product.

[0215] If the difference between the price of reference product A and the predicted price of the target product is greater than a first threshold, the electronic device determines that the price of the target product is abnormal, i.e., the predicted price of the target product is too high. The electronic device can adjust the parameters in the price prediction model and re-predict the price of the target product. Alternatively, the electronic device can lower the predicted price of the target product.

[0216] When the difference between the price value of reference product A and the predicted price value of the target product is greater than the second threshold value and less than the first threshold value, it is determined that the price value of the target product is within the normal range and can be adopted.

[0217] If the difference between the price of reference product A and the predicted price of the target product is less than the second threshold, the electronic device determines that the target product's price is abnormal, i.e., the predicted price of the target product is too low. Similar to the treatment described above for a predicted price of the target product that is too high, the electronic device can adjust the parameters in the price prediction model and re-predict the target product's price. Alternatively, the predicted price of the target product can be increased.

[0218] Based on the technical solution of Figure 4 above, the electronic device in the embodiment of the present application determines that the reference product with the highest similarity to the target product is the first product. Based on the price of the first product, the price value of the target product predicted by the price prediction model can be further verified to check the feasibility of the predicted price value of the target product.

[0219] In the embodiment of the present application, the display panel product consists of several different production stages. For example, the array process stage, the cell stage, the module process stage, the packaging process stage, the color film (Cf) process stage, etc. Since the product planner needs to focus on the reference price value (cost price) of each production stage of the target product when quoting the target product proposed by the customer, it is necessary to make a price forecast for each of the M production stages of the target product based on the product parameters of similar products and with the help of big data mining technology, and provide reference product information similar to the target product.

[0220] As shown in FIG6 , the model training method of the embodiment of the present application is summarized below.

[0221] For each production stage, the electronic device obtains the product parameters (requirement specifications) of the same type of products and preprocesses them, namely data processing, feature engineering, etc., to determine the product parameters and one or more predicted reference price values ​​before the production stage, and inputs the product parameters (requirement specifications) and one or more predicted reference price values ​​before the production stage into the neural network model for iterative training to determine the product cost prediction model corresponding to each production stage.

[0222] In the examples of this application, product parameters for similar products include, but are not limited to, product category, size, form, and resolution. Data processing includes, but is not limited to, removing outliers, correcting erroneous values, and supplementing missing values. Feature engineering includes, but is not limited to, categorical feature encoding, high-variance feature screening, and key attribute feature combination.

[0223] As shown in FIG7 , the product cost prediction method of the embodiment of the present application is summarized as follows.

[0224] For each production stage, the electronic device can determine the product parameters and one or more predicted reference price values ​​before the production stage, and then input them into the price prediction model to obtain the reference price value of each production stage in the M production stages (estimated BOM cost of each production stage), and then determine the predicted price value of the target product (estimated overall BOM cost). In addition, the price prediction model also has built-in feature contribution analysis (the contribution of multiple reference products) and reference product similarity to facilitate the subsequent evaluation of the predicted price value of the target product.

[0225] Among them, BOM cost refers to the standard cost of raw materials, parts, tools and labor required in the product manufacturing process.

[0226] In the embodiment of the present application, the prediction methods used by the price prediction model include but are not limited to LightGBM, Xgboost, ridge regression, SVR, and regression tree.

[0227] It should be pointed out that the various embodiments of the present application can refer to each other, for example, the same or similar steps, method embodiments, system embodiments and device embodiments can refer to each other without limitation.

[0228] In the embodiment of the present application, the electronic device can be divided into functional modules or functional units according to the above method examples. For example, each functional module or functional unit can be divided according to each function, or two or more functions can be integrated into one processing module. The above-mentioned integrated module can be implemented in the form of hardware or in the form of software functional modules or functional units. Among them, the division of modules or units in the embodiment of the present application is schematic and is only a logical function division. There may be other division methods in actual implementation.

[0229] As shown in FIG8 , it is a schematic structural diagram of a product cost prediction device provided in an embodiment of the present application. The product cost prediction device includes: a processing unit 801 and an acquisition unit 802 .

[0230] The processing unit 801 is configured to determine the input parameters of the product price prediction model corresponding to each of the M production stages based on the arrangement order of the M production stages of the target product.

[0231] Among them, each production stage corresponds to a product price prediction model, and the input parameters of the product price prediction model corresponding to the Nth production stage include the product parameters of the target product, and the reference price values ​​predicted by the product cost prediction models corresponding to all production stages arranged in an order before the Nth production stage; among them, M is greater than or equal to N, and M and N are positive integers.

[0232] The processing unit 801 is also configured to input the input parameters of the production stage into the product price prediction model corresponding to the production stage for each production stage, determine the reference price value predicted by the production stage, and obtain the reference price value of each production stage in the M production stages.

[0233] The processing unit 801 is further configured to determine a predicted price value of the target product based on the reference price value of each production stage in the M production stages.

[0234] In a possible implementation, the acquiring unit 802 is configured to acquire product parameters of a target product and product parameters of a plurality of reference products.

[0235] The processing unit 801 is further configured to determine, for each product, a similarity between the reference product and the target product based on the product parameters of the target product and the product parameters of the reference product, so as to obtain a similarity between each reference product in the plurality of reference products and the target product;

[0236] The processing unit 801 is further configured to determine a reference product with the highest similarity to the target product among the multiple reference products as the first product; the price value of the first product is used to determine whether there is an abnormality in the price value of the target product.

[0237] In one possible implementation, the product price prediction model corresponding to each production stage is different.

[0238] In one possible implementation, the predicted price value of the target product is the sum of the reference price values ​​of each production stage in the M production stages.

[0239] In one possible implementation, the product parameters include at least one of the following: product category; product size; product form; resolution.

[0240] As shown in FIG9 , which is a structural diagram of a model training device provided in an embodiment of the present application, the model training device includes: a processing unit 901 and an acquisition unit 902 .

[0241] The processing unit 901 is configured to determine an arrangement order of the M production stages of the target product based on the product cost prices of the M production stages, wherein the product cost prices of the M production stages are the actual product cost prices of the same type of products of the target product in the corresponding production stages.

[0242] The processing unit 901 is further configured to determine a data set of a neural network model corresponding to each of the M production stages based on the arrangement order of the M production stages.

[0243] Each production stage corresponds to a neural network model. The dataset corresponding to the Nth production stage includes product parameters of the same type of product, the product cost price of the same type of product in the Nth production stage, and the reference price values ​​predicted by the neural network model for all production stages before the Nth production stage. M is greater than or equal to N, and M and N are positive integers.

[0244] The processing unit 901 is also configured to iteratively train the neural network model of the production stage according to the data set of the production stage for each production stage until the loss value of the neural network model is lower than a preset threshold, and use the neural network model as the product price prediction model of the production stage.

[0245] In one possible implementation, processing unit 901 is specifically configured to determine, based on the product cost prices of the M production stages, a correlation coefficient matrix for the M production stages using a covariance algorithm, and determine an order of the M production stages based on the maximum correlation coefficient in the correlation coefficient matrix. The correlation coefficient matrix includes multiple correlation coefficients, and any correlation coefficient is determined based on the product cost prices of any two production stages among the M production stages.

[0246] In one possible implementation, processing unit 901 is specifically configured to determine the first and second production stages corresponding to the maximum correlation coefficient, where the first production stage is the production stage ranked first among the M production stages, and the second production stage is the production stage ranked second among the M production stages. A target production stage is determined, where the target production stage is the production stage after the second production stage. The row order of the target production stage in the correlation coefficient matrix is ​​the column order of the second production stage in the correlation coefficient matrix, and the correlation coefficient between the second production stage and the target production stage is the largest correlation coefficient in the row order. Based on the first production stage, the second production stage, and the target production stage, the order of the M production stages is determined.

[0247] In one possible implementation, the column order of the target production stage is not equal to the row order of the first production stage, and the column order of the target production stage is not equal to the column order of the second production stage. The column order represents the column in the correlation coefficient matrix, and the row order represents the row in the correlation coefficient matrix.

[0248] In one possible implementation, the product parameters include at least one of the following: product category; product size; product form; resolution.

[0249] In one possible implementation, the processing unit 901 is further configured to obtain original product parameters of products of the same type, perform data processing on the original product parameters, and determine product parameters of products of the same type. The data processing includes data preprocessing and feature engineering.

[0250] In one possible implementation, data preprocessing includes at least one of eliminating, correcting, and supplementing abnormal parameters within similar product parameters. Feature engineering is the process of converting raw product parameters of similar products after data preprocessing into product parameters for similar products. Feature engineering includes at least one of the following: feature screening; feature construction; feature encoding; and feature combination.

[0251] In one possible implementation, the processing unit 901 is further configured to input the training set into the neural network model to determine the output result of the neural network model; determine the loss value of the neural network model based on the output result and the test set; determine whether the loss value of the neural network model is lower than a preset threshold; if not, adjust the model parameters of the neural network model according to the loss value; if so, use the neural network model as a product price prediction model.

[0252] When implemented by hardware, the acquisition unit 802 and the acquisition unit 902 in the embodiment of the present application can be integrated on the communication interface, and the processing unit 801 and the processing unit 901 can be integrated on the processor. The specific implementation is shown in FIG10 .

[0253] Figure 10 shows a possible structural diagram of the electronic device involved in the above embodiment. The electronic device includes: a processor 1002 and a communication interface 1003. The processor 1002 is used to control and manage the actions of the electronic device, for example, to execute the steps performed by the above processing unit 801 and the processing unit 901, and / or to execute other processes of the technology described in this document. The communication interface 1003 is used to support communication between the electronic device and other network entities, for example, to execute the steps performed by the above acquisition unit 802 and the acquisition unit 902. The electronic device may also include a memory 1001 and a bus 1004, and the memory 1001 is used to store program code and data of the electronic device.

[0254] Among them, the memory 1001 can be a memory in the electronic device, etc., and the memory can include a volatile memory, such as a random access memory; the memory can also include a non-volatile memory, such as a read-only memory, a flash memory, a hard disk or a solid-state drive; the memory can also include a combination of the above types of memory.

[0255] The processor 1002 may implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure herein. The processor may be a central processing unit, a general-purpose processor, a digital signal processor, an application-specific integrated circuit, a field-programmable gate array, or other programmable logic device, a transistor logic device, a hardware component, or any combination thereof. The processor may implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure herein. The processor may also be a combination that implements computing functions, such as a combination of one or more microprocessors, or a combination of a DSP and a microprocessor.

[0256] Bus 1004 may be an Extended Industry Standard Architecture (EISA) bus, etc. Bus 1004 may be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, FIG10 shows only one thick line, but this does not mean that there is only one bus or only one type of bus.

[0257] The device in FIG10 may also be a chip, which includes one or more (including two) processors 1002 and a communication interface 1003 .

[0258] Optionally, the chip further includes a memory 1005, which may include a read-only memory and a random access memory, and provides operation instructions and data to the processor 1002. A portion of the memory 1005 may also include a non-volatile random access memory (NVRAM).

[0259] In some embodiments, the memory 1005 stores the following elements, execution modules or data structures, or a subset thereof, or an extended set thereof.

[0260] In the embodiment of the present application, the corresponding operation is performed by calling the operation instruction stored in the memory 1005 (the operation instruction may be stored in the operating system).

[0261] Some embodiments of the present disclosure provide a computer-readable storage medium (e.g., a non-transitory computer-readable storage medium), which stores computer program instructions. When the computer program instructions are executed on a computer (e.g., a receiving node), the computer executes a synchronization method as in any of the above embodiments.

[0262] Exemplarily, the above-mentioned computer-readable storage media may include, but are not limited to: magnetic storage devices (e.g., hard disks, floppy disks, or magnetic tapes, etc.), optical disks (e.g., CDs (Compact Disks), DVDs (Digital Versatile Disks), etc.), smart cards, and flash memory devices (e.g., EPROMs (Erasable Programmable Read-Only Memory), cards, sticks, or key drives, etc.). The various computer-readable storage media described in the present disclosure may represent one or more devices and / or other machine-readable storage media for storing information. The term "machine-readable storage medium" may include, but is not limited to, wireless channels and various other media capable of storing, containing, and / or carrying instructions and / or data.

[0263] Some embodiments of the present disclosure further provide a computer program product, for example, stored on a non-transitory computer-readable storage medium. The computer program product includes computer program instructions that, when executed on a computer (e.g., a receiving node), cause the computer to perform the synchronization method described in the above embodiments.

[0264] Some embodiments of the present disclosure further provide a computer program. When the computer program is executed on a computer (eg, a receiving node), the computer program enables the computer to execute the synchronization method of the above embodiments.

[0265] The beneficial effects of the above-mentioned computer-readable storage medium, computer program product and computer program are the same as the beneficial effects of the synchronization method of some of the above-mentioned embodiments, and will not be repeated here.

[0266] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0267] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0268] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0269] The above are only specific embodiments of the present disclosure, but the scope of protection of the present disclosure is not limited thereto. Any changes or substitutions that a person skilled in the art can conceive within the technical scope disclosed in this disclosure should be included in the scope of protection of the present disclosure. Therefore, the scope of protection of the present disclosure should be based on the scope of protection of the claims.

Claims

1. A product cost forecasting method, wherein: The method comprises: Based on the order of the M production stages of the target product, the input parameters of the product price forecasting model corresponding to each of the M production stages are determined; each production stage corresponds to one product price forecasting model, and the input parameters of the product price forecasting model corresponding to the Nth production stage in the order of arrangement include the product parameters of the target product and the reference price values ​​predicted by the product cost forecasting models corresponding to all production stages in the order of arrangement before the Nth production stage; wherein M is greater than or equal to N, and M and N are positive integers; For each production stage, inputting the input parameters of the production stage into a product price prediction model corresponding to the production stage, determining a reference price value predicted for the production stage, and thereby obtaining a reference price value for each of the M production stages; Based on the reference price value of each production stage in the M production stages, a predicted price value of the target product is determined.

2. The method according to claim 1, wherein The method further comprises: Obtaining product parameters of the target product and product parameters of multiple reference products; For each product, based on the product parameters of the target product and the product parameters of the reference product, determine the similarity between the reference product and the target product, so as to obtain the similarity between each reference product in the plurality of reference products and the target product; A reference product among the multiple reference products that has the highest similarity to the target product is determined as the first product.

3. The method according to claim 2, wherein: The price value of the first product is used to determine whether there is an abnormality in the price value of the target product.

4. The method according to any one of claims 1 to 3, wherein: The product price prediction model corresponding to each production stage is different.

5. The method according to any one of claims 1 to 4, wherein: The predicted price value of the target product is the sum of the reference price values ​​of each production stage in the M production stages.

6. The method according to any one of claims 1 to 5, wherein: The product parameters include at least one of the following: Product categories; Product size; Product form; Resolution.

7. A method for training a product cost prediction model, wherein: The method comprises: Determine the order of the M production stages of the target product based on the product cost prices of the M production stages; the product cost prices of the M production stages are the actual product cost prices of the same type of products as the target product in the corresponding production stages; Based on the arrangement order of the M production stages, determine the data set of the neural network model corresponding to each production stage in the M production stages; one production stage corresponds to one neural network model, and the arrangement order is Nth production stage. The data set corresponding to the segment includes product parameters of the same type of product, the product cost price of the same type of product at the Nth production stage, and the reference price values ​​predicted by the neural network model corresponding to all production stages arranged in the order before the Nth production stage; where M is greater than or equal to N, and M and N are positive integers; For each production stage, the neural network model of the production stage is iteratively trained according to the data set of the production stage until the loss value of the neural network model is lower than a preset threshold, and the neural network model is used as the product price prediction model of the production stage.

8. The method according to claim 7, wherein: The step of determining the arrangement order of the M production stages of the target product based on the product cost prices of the M production stages includes: Based on the product cost prices of the M production stages, determining a correlation coefficient matrix of the M production stages using a covariance algorithm; wherein the correlation coefficient matrix includes a plurality of correlation coefficients, and any one of the correlation coefficients is determined based on the product cost prices of any two production stages among the M production stages; The arrangement order of the M production stages is determined based on the maximum correlation coefficient in the correlation coefficient matrix.

9. The method according to claim 8, wherein The determining the arrangement order of the M production stages based on the maximum correlation coefficient in the correlation coefficient matrix includes: Determine a first production stage and a second production stage corresponding to the maximum correlation coefficient, where the first production stage is the production stage ranked first among the M production stages, and the second production stage is the production stage ranked second among the M production stages; Determine a target production stage, where the target production stage is a production stage after the second production stage, the row order of the target production stage in the correlation coefficient matrix is ​​the column order of the second production stage in the correlation coefficient matrix, and the correlation coefficient between the second production stage and the target production stage is the largest correlation coefficient in the row order; An arrangement order of the M production stages is determined based on the first production stage, the second production stage, and the target production stage.

10. The method according to claim 9, wherein: The column order of the target production stage is not equal to the row order of the first production stage, and the column order of the target production stage is not equal to the column order of the second production stage; the column order is used to characterize the column in which the production stage is located in the correlation coefficient matrix, and the row order is used to characterize the row in which the production stage is located in the correlation coefficient matrix.

11. The method according to any one of claims 7 to 10, wherein: The product parameters include at least one of the following: Product categories; Product size; Product form; Resolution.

12. The method according to any one of claims 7 to 11, wherein: The method further comprises: Obtaining original product parameters of the same type of product; Data processing is performed on the original product parameters to determine product parameters of the same type of products; the data processing includes data preprocessing and feature engineering.

13. The method according to claim 12, wherein: The data preprocessing includes at least one of eliminating, correcting, and supplementing abnormal parameters among the product parameters of the same type of products; The feature engineering is a process of converting the original product parameters of the same type of products after data preprocessing into the product parameters of the same type of products.

14. The method according to claim 12 or 13, wherein: The feature engineering includes at least one of the following: Feature screening; Feature construction; Feature encoding; Feature combination.

15. An electronic device, wherein: include: processor and memory; The memory is used to store computer programs or instructions, and the processor is used to run the computer programs or instructions to implement the product cost prediction method as described in any one of claims 1-6, or to implement the product cost prediction model training method as described in any one of claims 7-14.

16. A computer-readable storage medium, wherein: The computer-readable storage medium stores instructions. When a computer executes the instructions, the computer executes the product cost prediction method described in any one of claims 1 to 6, or implements the product cost prediction model training method described in any one of claims 7 to 14.

17. A computer program product, wherein The computer program product includes instructions. When the instructions are executed on a computer, the computer executes the product cost prediction method as described in any one of claims 1 to 6 above, or implements the product cost prediction model training method as described in any one of claims 7 to 14.

Images