Product design element trend prediction method

Abstract

The invention discloses a method for predicting the popularity trend of product design elements. The method comprises the steps of data collection, data preprocessing, data simulation model construction, product design element community construction, community parameter optimization and popularity trend prediction. The invention relates to the technical field of fashion trend prediction, in particular to a product design element fashion trend prediction method, which constructs a data simulation model by determining an objective function, designing a hyperbolic association activation function and defining a loss function, and improves the accuracy and reliability of fashion trend prediction. An element community is constructed and designed by defining an association distance, calculating an enhanced association degree, dividing community members and updating a community center, so that community division better conforms to actual conditions and requirements; parameter searching is carried out by designing a parameter searching dynamic radius, a direction adjusting factor, a region choice function and a global fine searching function, the parameter searching capability is improved, and searching is more flexible and targeted.

Description

Technical Field

[0001] The present invention relates to the technical field of fashion trend prediction, and in particular to a method for predicting the fashion trend of product design elements. Background Art

[0002] A method for predicting the popular trends of product design elements is to combine artificial intelligence technology and data processing technology to build a product design element data community, analyze data, capture the connection between data, and predict the popular trends of product design elements. However, the traditional popular trend prediction method model has the problem of low accuracy and reliability of prediction due to insufficient data quality and richness; the traditional method of building a product design element community has the problem of difficulty in accurately reflecting the relationship between elements, and the community division does not meet the actual situation and needs; the traditional parameter search algorithm has the problem of low search flexibility and pertinence, low probability of searching for the optimal parameters and slow speed. Summary of the invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a method for predicting the popular trends of product design elements. In view of the low accuracy and reliability of the prediction caused by insufficient data quality and richness in the traditional popular trend prediction method model, the present invention determines the objective function, designs the hyperbolic association activation function, defines the loss function of the generator and the discriminator, and allows the generator and the discriminator to compete and improve each other through alternating training, thereby enriching the data set and improving the accuracy and reliability of the popular trend prediction; in view of the problem that the traditional method of building a product design element community is difficult to accurately reflect the relationship between elements, and the community division does not meet the actual situation and needs, the present invention defines the association distance, calculates the enhanced association degree, divides the community members, and updates the community center to construct the design element community, more accurately reflects the relationship between elements, and makes the community division more in line with the actual situation and needs; in view of the low flexibility and pertinence of the search in the traditional parameter search algorithm, the low probability and slow speed of searching for the optimal parameters, the present invention performs parameter search by designing the parameter search dynamic radius, the direction adjustment factor, the regional decision function and the global fine search function, so as to make the search more flexible and pertinent, improve the probability and speed of finding the optimal parameters, and better control the search process and results.

[0004] The technical solution adopted by the present invention is as follows: a method for predicting the popular trend of product design elements, the method comprising the following steps:

[0005] Step S1: data collection;

[0006] Step S2: data preprocessing;

[0007] Step S3: constructing a data simulation model;

[0008] Step S4: Building a product design element community;

[0009] Step S5: Optimizing community parameters;

[0010] Step S6: Predicting fashion trends.

[0011] Furthermore, in step S1, the data collection is to collect descriptive data, social media data, market statistics data, consumer survey data and popularity data of product design elements; the descriptive data includes the color, material, style, shape and size of the product design elements; the design media data includes the views, likes, comments and sharing data of the product design element topic; the market statistics data includes the product sales volume, sales, market share and growth rate data of the product design elements; the consumer survey data includes consumers' preferences, needs and opinions on design elements; the popularity data includes the popularity intensity and popularity duration of the product design elements.

[0012] Further, in step S2, the data preprocessing verifies the data, identifies erroneous, duplicated and missing data, modifies erroneous data, removes duplicated data, fills in missing data, normalizes the data, and unifies the unit and range of the data;

[0013] Furthermore, in step S3, the building of the data simulation model specifically includes the following steps:

[0014] Step S31: Design the model structure. The data simulation model consists of two modules: a generator and a discriminator. The generator receives noise data and outputs simulation data. The discriminator discriminates the simulation data output by the generator and outputs the discrimination result of the simulation data.

[0015] Step S32: confirm the optimization target, which is expressed as follows:

[0016] ;

[0017] Among them, G represents the generator, D represents the discriminator, The objective function representing the performance of the reaction generator G and the discriminator D, It means that the objective function is maximized by adjusting the discriminator and minimized by adjusting the parameters of the generator. E[·] represents the average value, x represents the real data, represents the sampling symbol, represents the probability distribution of real data, z represents noise data, represents the probability distribution of the noise data, represents the logarithm, D(x) represents the discriminant result of the real data, Represents the discriminator's judgment result on the data generated by the generator through noise data;

[0018] Step S33: Design a hyperbolic association activation function, which is expressed as follows:

[0019] ;

[0020] in, represents the hyperbolic association activation function, represents the independent variable of the hyperbolic association activation function, represents the hyperbolic sine function, represents the hyperbolic cosine function;

[0021] Step S34: Construct a generator, select the hyperbolic correlation activation function as the activation function of the generator output layer, and control the output range between -1 and 1, as shown below:

[0022] ;

[0023] in, represents the output of the generator output layer, represents the weight of the generator output layer, represents the input of the generator output layer, represents the bias of the generator output layer;

[0024] Step S35: define the loss function, which is expressed as follows:

[0025] ;

[0026] ;

[0027] in, is the loss value of the generator, which represents the average of the negative logarithm of the probability that the discriminator judges the generated data as false. is the loss value of the discriminator;

[0028] Step S36: Alternate training, first fix the parameters of the generator to maximize the performance of the discriminator, then fix the parameters of the discriminator to minimize the loss of the generator, and repeat the training process to allow the generator and the discriminator to compete and train with each other;

[0029] Step S37: Data enrichment: using the generator to generate simulation data, performing quality inspection on the simulation data, and finally adding the simulation data to the original data to expand and enrich the original data.

[0030] Furthermore, in step S4, the building of the product design element community specifically includes the following steps:

[0031] Step S41: Community initialization, setting the number of communities; randomly selecting a data sample as the first community center; calculating the Euclidean distance between each remaining sample in the data set and the selected community center, obtaining the farthest distance value from the community center, and taking half of the farthest distance value; calculating the number of data points whose distance from the community center is greater than half of the farthest distance value and the number of data points whose distance from the community center is less than half of the farthest distance value, randomly selecting the next community center in the area with fewer data points, and repeating the process until all initial community centers are determined;

[0032] Step S42: define the association distance, which is expressed as follows:

[0033] ;

[0034] Among them, m represents the data sample, c represents the community center, represents the association distance between the data sample and the community center, mx represents the horizontal coordinate of the data sample, my represents the vertical coordinate of the data sample, cx represents the horizontal coordinate of the community center, and cy represents the vertical coordinate of the community center. Indicates taking the absolute value, arc indicates the inverse function sign, sin(·) indicates the sine function, cos(·) indicates the cosine function, Indicates square root;

[0035] Step S43: Calculate the enhanced correlation, which is expressed as follows:

[0036] ;

[0037] Among them, i and j represent the indexes of data samples and community centers respectively. represents the enhanced association between the i-th data sample and the j-th community center, represents the i-th data sample, represents the jth community center, v represents the number of community centers, represents the reinforcement factor, represents the association distance between the i-th data sample and the j-th community center;

[0038] Step S44: community member division, calculating the degree of reinforcement correlation between each data sample and each community center, and assigning the data sample to the community to which the community center with the highest degree of reinforcement correlation with the data sample belongs;

[0039] Step S45: Update the community center, as shown below:

[0040] ;

[0041] in, represents the updated community center, n represents the total number of data samples, Represents the horizontal coordinate of the i-th data sample;

[0042] Step S46: Community formation, repeat steps S44-S45 until the community converges.

[0043] Furthermore, in step S5, the optimization of community parameters specifically includes the following steps:

[0044] Step S51: Search initialization, determine the search parameters, including the number of communities and the reinforcement factor, create a parameter search space, generate an initial parameter search point set, and set the similarity between the data point and other data points in the same community as the parameter optimization degree, as shown below:

[0045] ;

[0046] Among them, opt represents the degree of parameter optimization, It means calculating the sum of the average distances from all data sample points to all data sample points in the same community. It means calculating the minimum sum of the average distances of all data sample points to all data sample points in other communities, and max(·) means taking the maximum value;

[0047] Step S52: Design parameter search dynamic radius, expressed as follows:

[0048] ;

[0049] Where t represents the number of parameter searches, represents the dynamic radius of parameter search during the tth parameter search, represents a cyclic linear increasing number between 0.5 and 2, r1 and r2 represent random numbers between 0 and 1, represents pi, Rw represents a random number between 0.5 and 2;

[0050] Step S53: Design the direction adjustment factor, which is expressed as follows:

[0051] ;

[0052] ;

[0053] in, and They represent the first direction adjustment factor and the second direction adjustment factor in the t-th parameter search, respectively. Indicates the current maximum parameter search dynamic radius. represents the hyperbolic sine function, represents the hyperbolic cosine function, Indicates taking the absolute value;

[0054] Step S54: Design a region selection function, which is expressed as follows:

[0055] ;

[0056] in, Indicates the location of the parameter point searched by the regional decision function during the t+1th parameter search. represents the current optimal parameter position, r3 represents a random number between 0 and 1, It represents the average position of all parameter points searched by the regional decision function during the tth parameter search. Indicates the position of the parameter point searched by the regional decision function during the tth parameter search;

[0057] Step S55: Design a global fine search function, which is expressed as follows:

[0058] ;

[0059] in, Indicates the position of the parameter point searched by the global fine search function during the t+1th parameter search, r4 represents a random number between 0 and 1, Indicates the position of the parameter point searched by the global fine search function during the tth parameter search;

[0060] Step S56: parameter optimization search, set the parameter optimization degree threshold and the maximum number of parameter searches, perform position searches of the parameter search point set using the regional decision function and the global fine search function in turn, calculate the parameter optimization degree of the searched parameter points, and increase the number of searches by one; during the search process, if there is a parameter point whose optimization degree is greater than the optimization degree threshold, stop the search, and output the parameters of the parameter point with the highest optimization degree at this time as the community optimization parameters; if the maximum number of parameter searches is reached, search again; otherwise, continue the search.

[0061] Furthermore, in step S6, the prediction of popular trends is to collect descriptive data, social media data, market statistics, and consumer survey data of the product design elements to be predicted, match the data to the constructed product design element community, obtain the popularity data of the product design elements in the same community to predict the popular trends of the product design elements to be predicted, and continuously collect new data to update and improve the community data.

[0062] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0063] (1) In order to address the problem of low prediction accuracy and reliability in traditional trend prediction methods due to insufficient data quality and richness, this scheme determines the objective function, designs the hyperbolic correlation activation function, defines the loss functions of the generator and discriminator, and allows the generator and discriminator to compete and improve with each other through alternating training, thereby enriching the data set and improving the accuracy and reliability of trend prediction.

[0064] (2) The traditional method of building a product design element community has the problem that it is difficult to accurately reflect the relationship between elements and the community division does not conform to the actual situation and needs. This solution builds a design element community by defining the association distance, calculating the enhanced association degree, dividing the community members, and updating the community center. It can more accurately reflect the relationship between elements and make the community division more in line with the actual situation and needs.

[0065] (3) To address the problems of low flexibility and targeting, low probability of finding the optimal parameters and slow speed in traditional parameter search algorithms, this scheme designs parameter search dynamic radius, direction adjustment factor, regional decision function and global fine search function to perform parameter search, making the search more flexible and targeted, improving the probability and speed of finding the optimal parameters, and better controlling the search process and results. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 A schematic diagram of a flow chart of a method for predicting the popular trend of product design elements provided by the present invention;

[0067] Figure 2 is a schematic flow chart of step S3;

[0068] Figure 3 is a schematic flow chart of step S4;

[0069] Figure 4 It is a flowchart diagram of step S5.

[0070] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0071] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0072] In the description of the present invention, it should be understood that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0073] Example 1, see Figure 1 The present invention provides a method for predicting the popular trend of product design elements, the method comprising the following steps:

[0074] Step S1: Data collection, collecting descriptive data of product design elements, social media data, market statistics, consumer survey data, and popularity data;

[0075] Step S2: Data preprocessing, checking the data, identifying and processing errors, duplications and missing data, and normalizing the data;

[0076] Step S3: construct a data simulation model, and train the data simulation model by designing a model structure, defining an objective function, designing a hyperbolic association activation function, and defining loss functions of a generator and a discriminator;

[0077] Step S4: constructing a product design element community, by initializing the community, defining the association distance, calculating the enhanced association degree, dividing the community members, and updating the community center to construct the product design element community;

[0078] Step S5: Searching for optimal parameters, performing parameter optimization search by searching initialization, designing parameter search dynamic radius, designing direction adjustment factor, designing area selection function, and designing global fine search function;

[0079] Step S6: Predicting popular trends, collecting data on product design elements to be predicted, matching the data to the constructed product design element community, and obtaining popularity data of product design elements in the same community to predict popular trends for the product design elements to be predicted.

[0080] Example 2, see Figure 1This embodiment is based on the above embodiment. In step S1, the data collection is to collect descriptive data, social media data, market statistics data, consumer survey data and popularity data of product design elements; the descriptive data includes the color, material, style, shape and size of the product design elements; the design media data includes the views, likes, comments and sharing data of the product design element topic; the market statistics data includes the product sales volume, sales, market share and growth rate data of the product design elements; the consumer survey data includes consumers' preferences, needs and opinions on design elements; the popularity data includes the popularity intensity and popularity duration of the product design elements.

[0081] Example 3, see Figure 1 This embodiment is based on the above embodiment. In step S2, the data preprocessing verifies the data, identifies erroneous, duplicated and missing data, modifies erroneous data, removes duplicated data, fills in missing data, normalizes the data, and unifies the unit and range of the data.

[0082] Example 4, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In step S3, building a data simulation model specifically includes the following steps:

[0083] Step S31: Design the model structure. The data simulation model consists of two modules: a generator and a discriminator. The generator receives noise data and outputs simulation data. The discriminator discriminates the simulation data output by the generator and outputs the discrimination result of the simulation data.

[0084] Step S32: confirm the optimization target, which is expressed as follows:

[0085] ;

[0086] Among them, G represents the generator, D represents the discriminator, The objective function representing the performance of the reaction generator G and the discriminator D, It means that the objective function is maximized by adjusting the discriminator and minimized by adjusting the parameters of the generator. E[·] represents the average value, x represents the real data, represents the sampling symbol, represents the probability distribution of real data, z represents noise data, represents the probability distribution of the noise data, represents the logarithm, D(x) represents the discriminant result of the real data, Represents the discriminator's judgment result on the data generated by the generator through noise data;

[0087] Step S33: Design a hyperbolic association activation function, which is expressed as follows:

[0088] ;

[0089] in, represents the hyperbolic association activation function, represents the independent variable of the hyperbolic association activation function, represents the hyperbolic sine function, represents the hyperbolic cosine function;

[0090] Step S34: Construct a generator, select the hyperbolic correlation activation function as the activation function of the generator output layer, and control the output range between -1 and 1, as shown below:

[0091] ;

[0092] in, represents the output of the generator output layer, represents the weight of the generator output layer, represents the input of the generator output layer, represents the bias of the generator output layer;

[0093] Step S35: define the loss function, which is expressed as follows:

[0094] ;

[0095] ;

[0096] in, is the loss value of the generator, which represents the average of the negative logarithm of the probability that the discriminator judges the generated data as false. is the loss value of the discriminator;

[0097] Step S36: Alternate training, first fix the parameters of the generator to maximize the performance of the discriminator, then fix the parameters of the discriminator to minimize the loss of the generator, and repeat the training process to allow the generator and the discriminator to compete and train with each other;

[0098] Step S37: Data enrichment: using the generator to generate simulation data, performing quality inspection on the simulation data, and finally adding the simulation data to the original data to expand and enrich the original data.

[0099] By performing the above operations, in order to address the problem of low prediction accuracy and reliability in traditional trend prediction methods due to insufficient data quality and richness, this solution determines the objective function, designs the hyperbolic correlation activation function, defines the loss functions of the generator and discriminator, and allows the generator and discriminator to compete and improve with each other through alternating training, thereby enriching the data set and improving the accuracy and reliability of trend prediction.

[0100] Example 5, see Figure 1 and Figure 3 This embodiment is based on the above embodiment. In step S4, the construction of the product design element community specifically includes the following steps:

[0101] Step S41: Community initialization, setting the number of communities; randomly selecting a data sample as the first community center; calculating the Euclidean distance between each remaining sample in the data set and the selected community center, obtaining the farthest distance value from the community center, and taking half of the farthest distance value; calculating the number of data points whose distance from the community center is greater than half of the farthest distance value and the number of data points whose distance from the community center is less than half of the farthest distance value, randomly selecting the next community center in the area with fewer data points, and repeating the process until all initial community centers are determined;

[0102] Step S42: define the association distance, which is expressed as follows:

[0103] ;

[0104] Among them, m represents the data sample, c represents the community center, represents the association distance between the data sample and the community center, mx represents the horizontal coordinate of the data sample, my represents the vertical coordinate of the data sample, cx represents the horizontal coordinate of the community center, and cy represents the vertical coordinate of the community center. Indicates taking the absolute value, arc indicates the inverse function sign, sin(·) indicates the sine function, cos(·) indicates the cosine function, Indicates square root;

[0105] Step S43: Calculate the enhanced correlation, which is expressed as follows:

[0106] ;

[0107] Among them, i and j represent the indexes of data samples and community centers respectively. represents the enhanced association between the i-th data sample and the j-th community center, represents the i-th data sample, represents the jth community center, v represents the number of community centers, represents the reinforcement factor, represents the association distance between the i-th data sample and the j-th community center;

[0108] Step S44: community member division, calculating the degree of reinforcement correlation between each data sample and each community center, and assigning the data sample to the community to which the community center with the highest degree of reinforcement correlation with the data sample belongs;

[0109] Step S45: Update the community center, as shown below:

[0110] ;

[0111] in, represents the updated community center, n represents the total number of data samples, Represents the horizontal coordinate of the i-th data sample;

[0112] Step S46: Community formation, repeat steps S44-S45 until the community converges.

[0113] By performing the above operations, the traditional method of building a product design element community has the problem that it is difficult to accurately reflect the relationship between elements and the community division does not meet the actual situation and needs. This solution builds a design element community by defining the association distance, calculating the enhanced association degree, dividing the community members, and updating the community center, which more accurately reflects the relationship between elements and makes the community division more in line with the actual situation and needs.

[0114] Example 6, see Figure 1 and Figure 4 This embodiment is based on the above embodiment. In step S5, the optimization of community parameters specifically includes the following steps:

[0115] Step S51: Search initialization, determine the search parameters, including the number of communities and the reinforcement factor, create a parameter search space, generate an initial parameter search point set, and set the similarity between the data point and other data points in the same community as the parameter optimization degree, as shown below:

[0116] ;

[0117] Among them, opt represents the degree of parameter optimization, It means calculating the sum of the average distances from all data sample points to all data sample points in the same community. It means calculating the minimum sum of the average distances of all data sample points to all data sample points in other communities, and max(·) means taking the maximum value;

[0118] Step S52: Design parameter search dynamic radius, expressed as follows:

[0119] ;

[0120] Where t represents the number of parameter searches, represents the dynamic radius of parameter search during the tth parameter search, represents a cyclic linear increasing number between 0.5 and 2, r1 and r2 represent random numbers between 0 and 1, represents pi, Rw represents a random number between 0.5 and 2;

[0121] Step S53: Design the direction adjustment factor, which is expressed as follows:

[0122] ;

[0123] ;

[0124] in, and They represent the first direction adjustment factor and the second direction adjustment factor in the t-th parameter search, respectively. Indicates the current maximum parameter search dynamic radius. represents the hyperbolic sine function, represents the hyperbolic cosine function, Indicates taking the absolute value;

[0125] Step S54: Design a region selection function, which is expressed as follows:

[0126] ;

[0127] in, Indicates the location of the parameter point searched by the regional decision function during the t+1th parameter search. represents the current optimal parameter position, r3 represents a random number between 0 and 1, It represents the average position of all parameter points searched by the regional decision function during the tth parameter search. Indicates the position of the parameter point searched by the regional decision function during the tth parameter search;

[0128] Step S55: Design a global fine search function, which is expressed as follows:

[0129] ;

[0130] in, Indicates the position of the parameter point searched by the global fine search function during the t+1th parameter search, r4 represents a random number between 0 and 1, Indicates the position of the parameter point searched by the global fine search function during the tth parameter search;

[0131] Step S56: parameter optimization search, set the parameter optimization degree threshold and the maximum number of parameter searches, perform position searches of the parameter search point set using the regional decision function and the global fine search function in turn, calculate the parameter optimization degree of the searched parameter points, and increase the number of searches by one; during the search process, if there is a parameter point whose optimization degree is greater than the optimization degree threshold, stop the search, and output the parameters of the parameter point with the highest optimization degree at this time as the community optimization parameters; if the maximum number of parameter searches is reached, search again; otherwise, continue the search.

[0132] By performing the above operations, in order to address the problems of low search flexibility and targeting, low probability and slow speed of searching for optimal parameters in traditional parameter search algorithms, this scheme designs parameter search dynamic radius, direction adjustment factor, regional decision function and global fine search function to perform parameter search, making the search more flexible and targeted, improving the probability and speed of finding the optimal parameters, and better controlling the search process and results.

[0133] Embodiment 7, see Figure 1 This embodiment is based on the above embodiment. In step S6, the prediction of popular trends is to collect descriptive data, social media data, market statistics, and consumer survey data of the product design elements to be predicted, match the data to the constructed product design element community, obtain the popularity data of the product design elements in the same community to predict the popular trends of the product design elements to be predicted, and continuously collect new data to update and improve the community data.

[0134] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0135] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0136] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. A method for predicting the popular trend of product design elements, characterized in that: The method comprises the following steps: Step S1: Data collection, collecting descriptive data of product design elements, social media data, market statistics, consumer survey data, and popularity data; Step S2: Data preprocessing, checking the data, identifying and processing errors, duplications and missing data, and normalizing the data; Step S3: construct a data simulation model, and train the data simulation model by designing a model structure, defining an optimization target, designing a hyperbolic association activation function, and defining loss functions of a generator and a discriminator; Step S4: constructing a product design element community, by initializing the community, defining the association distance, calculating the enhanced association degree, dividing the community members, and updating the community center to construct the product design element community; Step S5: Searching for optimal parameters, performing parameter optimization search by searching initialization, designing parameter search dynamic radius, designing direction adjustment factor, designing area decision function, and designing global fine search function; Step S6: Predicting fashion trends, collecting data on product design elements to be predicted, matching the data to the constructed product design element community, and obtaining popularity data of product design elements in the same community to predict fashion trends for the product design elements to be predicted.

2. The method for predicting the popular trend of product design elements according to claim 1, characterized in that: In step S3, the construction of the data simulation model specifically includes the following steps: Step S31: Design the model structure. The data simulation model consists of two modules: a generator and a discriminator. The generator receives noise data and outputs simulation data. The discriminator discriminates the simulation data output by the generator and outputs the discrimination result of the simulation data. Step S32: confirm the optimization target, which is expressed as follows: ； Among them, G represents the generator, D represents the discriminator, The objective function representing the performance of the reaction generator G and the discriminator D, It means that the objective function is maximized by adjusting the discriminator and minimized by adjusting the parameters of the generator. represents the average value, x represents the real data, represents the sampling symbol, represents the probability distribution of real data, z represents noise data, represents the probability distribution of the noise data, represents the logarithm, D(x) represents the discriminant result of the real data, Represents the discriminator's judgment result on the data generated by the generator through noise data; Step S33: Design a hyperbolic association activation function, which is expressed as follows: ； in, represents the hyperbolic association activation function, represents the independent variable of the hyperbolic association activation function, represents the hyperbolic sine function, represents the hyperbolic cosine function; Step S34: Construct a generator, select the hyperbolic correlation activation function as the activation function of the generator output layer, and control the output range between -1 and 1, as shown below: ； in, represents the output of the generator output layer, represents the weight of the generator output layer, represents the input of the generator output layer, represents the bias of the generator output layer; Step S35: define the loss function, which is expressed as follows: ； ； in, is the loss value of the generator, which represents the average of the negative logarithm of the probability that the discriminator judges the generated data as false. is the loss value of the discriminator; Step S36: Alternate training, first fix the parameters of the generator to maximize the performance of the discriminator, then fix the parameters of the discriminator to minimize the loss of the generator, and repeat the training process to allow the generator and the discriminator to compete and train with each other; Step S37: Data enrichment: using the generator to generate simulation data, performing quality inspection on the simulation data, and finally adding the simulation data to the original data to expand and enrich the original data.

3. The method for predicting the popular trend of product design elements according to claim 1, characterized in that: In step S4, the construction of the product design element community specifically includes the following steps: Step S41: Community initialization, setting the number of communities; randomly selecting a data sample as the first community center; calculating the Euclidean distance between each remaining sample in the data set and the selected community center, obtaining the farthest distance value from the community center, and taking half of the farthest distance value; calculating the number of data points whose distance from the community center is greater than half of the farthest distance value and the number of data points whose distance from the community center is less than half of the farthest distance value, randomly selecting the next community center in the area with fewer data points, and repeating the process until all initial community centers are determined; Step S42: define the association distance, which is expressed as follows: ； Among them, m represents the data sample, c represents the community center, represents the association distance between the data sample and the community center, mx represents the horizontal coordinate of the data sample, my represents the vertical coordinate of the data sample, cx represents the horizontal coordinate of the community center, and cy represents the vertical coordinate of the community center. Indicates taking the absolute value, arc indicates the inverse function sign, sin(·) indicates the sine function, cos(·) indicates the cosine function, Indicates square root; Step S43: Calculate the enhanced correlation, which is expressed as follows: ； Among them, i and j represent the indexes of data samples and community centers respectively. represents the enhanced association between the i-th data sample and the j-th community center, represents the i-th data sample, represents the jth community center, v represents the number of community centers, represents the reinforcement factor, represents the association distance between the i-th data sample and the j-th community center; Step S44: community member division, calculating the degree of reinforcement correlation between each data sample and each community center, and assigning the data sample to the community to which the community center with the highest degree of reinforcement correlation with the data sample belongs; Step S45: Update the community center, as shown below: ； in, represents the updated community center, n represents the total number of data samples, Represents the horizontal coordinate of the i-th data sample; Step S46: Community formation, repeat steps S44-S45 until the community converges.

4. The method for predicting the popular trend of product design elements according to claim 1, characterized in that: In step S5, the optimization of community parameters specifically includes the following steps: Step S51: Search initialization, determine the search parameters, including the number of communities and the reinforcement factor, create a parameter search space, generate an initial parameter search point set, and set the similarity between the data point and other data points in the same community as the parameter optimization degree, as shown below: ； Among them, opt represents the degree of parameter optimization, It means calculating the sum of the average distances from all data sample points to all data sample points in the same community. It means calculating the minimum sum of the average distances of all data sample points to all data sample points in other communities, and max(·) means taking the maximum value; Step S52: Design parameter search dynamic radius, expressed as follows: ； Where t represents the number of parameter searches, represents the dynamic radius of parameter search during the tth parameter search, represents a cyclic linear increasing number between 0.5 and 2, r1 and r2 represent random numbers between 0 and 1, represents pi, Rw represents a random number between 0.5 and 2; Step S53: Design the direction adjustment factor, which is expressed as follows: ； ； in, and They represent the first direction adjustment factor and the second direction adjustment factor in the t-th parameter search, respectively. Indicates the current maximum parameter search dynamic radius. represents the hyperbolic sine function, represents the hyperbolic cosine function, Indicates taking the absolute value; Step S54: Design a region selection function, which is expressed as follows: ； in, Indicates the location of the parameter point searched by the regional decision function during the t+1th parameter search. represents the current optimal parameter position, r3 represents a random number between 0 and 1, It represents the average position of all parameter points searched by the regional decision function during the tth parameter search. Indicates the position of the parameter point searched by the regional decision function during the tth parameter search; Step S55: Design a global fine search function, which is expressed as follows: ； in, Indicates the position of the parameter point searched by the global fine search function during the t+1th parameter search, r4 represents a random number between 0 and 1, Indicates the position of the parameter point searched by the global fine search function during the tth parameter search; Step S56: parameter optimization search, set the parameter optimization degree threshold and the maximum number of parameter searches, perform position searches of the parameter search point set using the regional decision function and the global fine search function in turn, calculate the parameter optimization degree of the searched parameter points, and increase the number of searches by one; during the search process, if there is a parameter point whose optimization degree is greater than the optimization degree threshold, stop the search, and output the parameters of the parameter point with the highest optimization degree at this time as the community optimization parameters; if the maximum number of parameter searches is reached, search again; otherwise, continue the search.

5. The method for predicting the popular trend of product design elements according to claim 1, characterized in that: In step S1, the data collection is to collect descriptive data, social media data, market statistics data, consumer survey data and popularity data of product design elements; the descriptive data includes the color, material, style, shape and size of the product design elements; the design media data includes the views, likes, comments and sharing data of the product design element topic; the market statistics data includes the product sales volume, sales, market share and growth rate data of the product design elements; the consumer survey data includes consumers' preferences, needs and opinions on design elements; the popularity data includes the popularity intensity and popularity duration of the product design elements.

6. The method for predicting the popular trend of product design elements according to claim 1, characterized in that: In step S2, the data preprocessing verifies the data, identifies erroneous, duplicated and missing data, modifies erroneous data, removes duplicated data, fills in missing data, normalizes the data, and unifies the unit and range of the data; In step S6, the prediction of popular trends is to collect descriptive data, social media data, market statistics, and consumer survey data of the product design elements to be predicted, match the data to the constructed product design element community, obtain the popularity data of the product design elements in the same community to predict the popular trends of the product design elements to be predicted, and continuously collect new data to update and improve the community data.

Images