A product marketing trend prediction system for evaluating cultural creativity

Abstract

The present invention relates to the field of product marketing technology, and specifically to a product marketing trend prediction system for evaluating cultural creativity, which comprises a data collection module, an integrated multi-model module and an integrated decision module, wherein: the data collection module collects data on cultural creativity products, including historical market data, social media data and web page visual data; the integrated multi-model module predicts the future market sales of each product based on the historical market data; predicts product acceptance based on social media data; predicts highly accepted visual elements based on web page visual data; the integrated decision module predicts the future market sales of products in order of product acceptance from high to low, selects products with high product acceptance but low future market sales, extracts the visual elements corresponding to the products, compares them with the highly accepted visual elements in a convolutional neural network algorithm model, and recalculates the web page interface of the product and optimizes the processing based on the comparison results.

Description

Technical Field

[0001] The present invention relates to the technical field of product marketing, and in particular to a product marketing trend prediction system for evaluating cultural creativity. Background Art

[0002] Traditional evaluation and forecasting systems for cultural and creative product marketing trends often rely on simple statistical models and historical trend analysis, which show obvious deficiencies in dealing with complex market dynamics and diversified consumer preferences. Such systems are usually unable to effectively integrate and analyze large amounts of data from different channels, such as social media sentiment analysis and web page visual data analysis, which limits their ability in terms of prediction accuracy and market insight.

[0003] Due to the lack of advanced algorithms and machine learning technology, traditional systems perform poorly in identifying visual elements that are highly correlated with product sales, and are unable to automatically extract and analyze key visual features in images such as color, layout, and graphic elements. This processing method not only increases uncertainty in product design and marketing strategies, but may also lead to unscientific marketing decisions and ineffective prediction of market trends, making it difficult for products to attract target customer groups and reduce market competitiveness. Taking "high-end handicrafts" as an example, suppose there is an exquisite ceramic vase that is highly appreciated by art collectors and designers on social media. According to social media data, the acceptance of this vase is very high. However, historical market sales data show that although it is loved by professionals, ordinary consumers rarely buy it due to its high price and low practicality. From the perspective of web page visual elements, high-end handicrafts such as ceramic flower porcelain are popular in the market but have low sales, which may be because their online display fails to fully demonstrate their value. Because high-definition pictures, elegant layouts, appropriate color usage, and highly interactive designs can better highlight their artistic and craftsmanship value, enhance user experience and purchase intention. Through the careful application of these visual strategies, consumers' purchasing motivation can be effectively improved, thereby potentially increasing sales.

[0004] In addition, traditional systems that lack dynamic optimization and real-time feedback mechanisms cannot quickly adapt to rapid changes in the market and consumer behavior, which is particularly unfavorable in the fast-paced and ever-changing cultural and creative market, making it difficult for companies to maintain their competitive advantage and innovation capabilities in a highly competitive environment. Summary of the invention

[0005] The purpose of the present invention is to provide a product marketing trend prediction system for evaluating cultural creativity to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above-mentioned purpose, the present invention provides the following technical solution: a product marketing trend prediction system for evaluating cultural creativity, characterized in that it includes a data collection module, an integrated multi-model module and an integrated decision module, wherein:

[0007] The data collection module collects data on cultural and creative products, including historical market data, social media data, and web page visual data;

[0008] The integrated multi-model module uses a support vector machine algorithm model to predict the future market sales of each product based on the historical market data; uses a sentiment dictionary to predict product acceptance based on the social media data; and uses a convolutional neural network algorithm model to predict highly accepted visual elements based on the webpage visual data.

[0009] The integrated decision module predicts the future market sales of the products in order of the product acceptance from high to low, selects products with high product acceptance but low future market sales, extracts the visual elements corresponding to the products as the current visual elements, and compares the current visual elements with the high acceptance visual elements in the convolutional neural network algorithm model, wherein:

[0010] If not, the product's web interface will be redesigned based on the highly accepted visual elements.

[0011] If they are consistent, the convolutional neural network algorithm in the integrated multi-model module is optimized, the high acceptance visual elements are re-predicted, and the high acceptance visual elements before and after optimization are compared. If they are consistent, the product is marked and sent to the management personnel as an abnormal result.

[0012] As a further improvement of the technical solution, the integrated multi-model module includes a future market sales forecasting unit, which uses a support vector machine algorithm to predict future market sales based on historical market data, specifically including:

[0013] Data preparation: Extract features from historical market data, including but not limited to time series features, prices, promotional activities, actual sales, and divide the data into training data and prediction data;

[0014] Feature selection: Use correlation analysis methods to reduce feature dimensions and select the most representative features;

[0015] Support vector machine model training: define the loss function as ,in Represents the loss function, which is used to calculate and express the gap between the predicted value of the support vector machine model and the actual value; Represents the actual value, the actual sales value obtained from the data; represents the predicted value of the support vector machine model, The support vector machine model learns by inputting training data The prediction function obtained; The input features of the training data representing the model are extracted from historical market data; the radial basis function is selected as the kernel function, and cross-validation is used to determine the optimal model parameters;

[0016] Future market sales forecast: Use the trained support vector machine model to predict the forecast data, and the prediction result is the future market sales.

[0017] As a further improvement of the present technical solution, the integrated multi-model module includes a product acceptance prediction unit, which uses a sentiment dictionary to analyze consumer sentiment and predict product acceptance, predicts text reviews about products from social media data, performs text preprocessing on the text reviews by standardizing, removing stop words and extracting stems, calculates sentiment scores using the sentiment dictionary, and takes products as units, statistics and averages the sentiment scores of each product to convert them into product acceptance.

[0018] As a further improvement of the technical solution, the integrated multi-model module includes a visual element prediction unit, which uses a convolutional neural network algorithm model to analyze the correlation between product visual elements and current sales levels based on web page visual data, and identifies highly accepted visual elements, specifically including:

[0019] Data preparation: Collect web page visual data and the current sales level of related products as training data. The current sales level includes high sales, medium sales and low sales. The web page visual data includes but is not limited to product pictures, colors and layouts, and convert the web page visual data into a format suitable for convolutional neural network processing;

[0020] Selection and configuration of convolutional neural network: Select the VGGNet convolutional neural network model to configure the network's layers and parameters, and build a deep network by repeatedly using small-sized convolution kernels and pooling layers, including input layers, multiple convolutional layers, pooling layers, and fully connected layers;

[0021] Model training: The input data is web page visual data, and the current sales level is used as a label; the network is trained using labeled training data;

[0022] Feature extraction and analysis: Use trained convolutional neural networks to process the web visual data of each product. When processing images, convolutional neural networks extract deep features of the entire image or part of the image. In convolutional neural networks, each layer extracts different characteristics of the input data. As data is passed through the network, primary features are combined into more complex features.

[0023] Prediction of highly acceptable visual elements: By extracting features and analyzing the output of the convolutional neural network model, we determine which feature appears most frequently in the images of high-selling products and use this feature as a highly acceptable visual element. Highly acceptable visual elements include but are not limited to color combinations, shapes, and image arrangement methods.

[0024] As a further improvement of the technical solution, the integrated decision module includes a consistency calculation unit, which receives the future market sales volume prediction unit and the product acceptance prediction unit respectively predicting the future market sales volume and product acceptance, and compares the product features and visual elements, specifically including:

[0025] Sort all products from high to low according to product acceptance;

[0026] For each product, compare its future market sales and product acceptance, and select products with high acceptance but low future predicted sales;

[0027] For products with high acceptance but low sales forecast, the current visual elements of the product are extracted, and the webpage visual data corresponding to the product is input into the visual element prediction unit, and the highly accepted visual elements of the product are output, and similarity comparison is performed, as follows:

[0028] The cosine similarity formula is used to evaluate the similarity between the two sets of feature vectors of the current visual element and the highly accepted visual element:

[0029] ,in and are two sets of eigenvectors, represents vector dot product, and is the Euclidean norm of the vector;

[0030] Threshold judgment: Set a threshold. If the similarity of two feature vectors is higher than this threshold, they are considered consistent; if it is lower than this threshold, they are considered inconsistent.

[0031] As a further improvement of the technical solution, the integrated decision module includes a decision unit, which makes a decision based on the consistency result of the consistency calculation unit, wherein if the current visual element is inconsistent with the high acceptance visual element, the web page interface of the product is redesigned based on the high acceptance visual element, specifically including:

[0032] List the key features of highly receptive visual elements;

[0033] Use quantitative metrics to describe these visual elements, including:

[0034] Color analysis: Extract the color histogram of the highly acceptable image and calculate the main color ratio. ;

[0035] Layout analysis: Identify layout structures through image segmentation techniques, including but not limited to partition proportions and element alignment;

[0036] Visual design principle formulation: formulate design principles based on the results of color and layout analysis;

[0037] Design Implementation: Apply design principles and quantitative metrics extracted from high-perception visual elements to design web pages, including:

[0038] Color instantiation: Adjust the CSS style of the web interface to set the color code to match the most prominent color of the color histogram in the high-reception visual element;

[0039] Layout optimization: Adjust the content layout through CSS and HTML to ensure that the layout is consistent with the layout indicators of high-acceptance products.

[0040] As a further improvement of the present technical solution, when the consistency result in the decision unit is inconsistent, the convolutional layer of the convolutional neural network algorithm model in the visual element prediction unit is adjusted, and the current number of filters in the convolutional layer is adjusted according to the formula: new number of filters = current number of filters × adjustment factor; and the optimized convolutional neural network algorithm model is used to re-identify the highly acceptable visual elements, and compare the highly acceptable visual elements before and after the optimization. If they are relatively consistent, the product is marked as abnormal and sent to the management personnel as an abnormal result.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] 1. This product marketing trend prediction system for evaluating cultural and creative products evaluates and optimizes the marketing strategy of cultural and creative products by comprehensively utilizing data analysis and model prediction. First, the data collection module is used to collect multi-dimensional data including historical market data, social media data, and web page visual data, which provides a basis for in-depth understanding of market trends and consumer preferences; with the help of the integrated multi-model module, the system predicts the future market sales of products through the support vector machine algorithm, and uses the sentiment dictionary analysis tool to extract consumer emotions from social media data to evaluate product acceptance, and applies convolutional neural networks to analyze web page visual data to identify visual elements related to high sales. Through the integrated application of these models, the system can analyze the market potential and audience acceptance of products from multiple angles, providing a scientific basis for decision-making.

[0043] 2. This product marketing trend prediction system for evaluating cultural creativity uses convolutional neural networks to analyze and predict highly acceptable visual elements. The convolutional neural network algorithm conducts in-depth analysis of web page visual data and utilizes powerful image processing capabilities to automatically identify and extract key visual features, including color, shape, pattern, and texture, which are usually important factors in consumers' visual appeal to products. Through classification training of visual data of high, medium, and low-selling products, the convolutional neural network model can learn which visual features are closely related to the sales volume of products. In addition, by analyzing the visual elements of new products with trained models, their acceptance and possible sales performance can be predicted. This data-based approach reduces uncertainty before market promotion, enabling companies to optimize visual presentation in the early stages of product design, making it more likely to attract target customer groups and enhance the market competitiveness of products.

[0044] 3. This product marketing trend prediction system for evaluating cultural creativity further optimizes the market performance of the product by comparing the current visual elements of the product with the highly acceptable visual elements identified by the convolutional neural network model in the integrated decision-making module. When it is found that the current visual elements are inconsistent with the highly acceptable visual elements, the decision-making unit will guide the redesign of the web page interface and use data-driven methods to accurately adjust the color, layout, and graphic elements, etc., so that it is more in line with the market-verified high-acceptance model. This not only improves the visual appeal of the product, but also directly enhances the market competitiveness of the product. When the consistency results show consistency, the system will also optimize the convolutional neural network algorithm model to ensure the continuous progress of the model and adapt to market changes, and enhance the model's prediction accuracy and applicability. This process ensures that enterprises can continue to use the latest market data to make decisions, optimize product design, increase sales and market share, so as to maintain competitiveness and promote innovation in the cultural and creative industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a schematic diagram of the overall module of the present invention;

[0046] Figure 2 It is a schematic diagram of an integrated multi-model module unit of the present invention;

[0047] Figure 3 It is a schematic diagram of the integrated decision module unit of the present invention.

[0048] In the figure: 100, data collection module; 200, integrated multi-model module; 201, future market sales prediction unit; 202, product acceptance prediction unit; 203, visual element prediction unit; 300, integrated decision module; 301, consistency calculation unit; 302, decision unit. DETAILED DESCRIPTION

[0049] The following will be combined with the accompanying drawings in the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] See also Figure 1-3 The present invention provides a technical solution: a product marketing trend prediction system for evaluating cultural creativity, including a data collection module 100, an integrated multi-model module 200 and an integrated decision module 300.

[0051] The data collection module 100 collects data on cultural and creative products, including historical market data, social media data and web page visual data, wherein the historical market data includes but is not limited to the product's previous sales records and price changes; social media data uses social platforms to collect discussions, comments, likes, etc. about the product, and social media data reflects the public's emotional tendencies and attitudes towards the product; web page visual data collects visual materials such as pictures, advertisements, videos, etc. of the product, and is used to analyze the relationship between the product's visual presentation and market performance.

[0052] The future market sales volume prediction unit 201 in the integrated multi-model module 200 uses a support vector machine algorithm to predict future market sales volume based on historical market data, specifically including:

[0053] Data preparation: Extract features from historical market data, including time series features (such as month, seasonality), price, promotional activities, actual sales, etc., and divide the data into training data and prediction data;

[0054] Feature selection: Use correlation analysis methods to reduce feature dimensions and select the most representative features;

[0055] Data standardization: standardize the feature values ​​to eliminate the deviation caused by different magnitudes;

[0056] Support vector machine model training: define the loss function as ,in Represents the loss function, which is used to calculate and express the gap between the predicted value of the support vector machine model and the actual value; Represents the actual value, that is, the actual sales value obtained from the data; represents the predicted value of the support vector machine model, where The support vector machine model learns by inputting training data The prediction function obtained; Represents the training data input features of the model, which are extracted from historical market data, such as time series features, prices, promotional activities, etc.; selects radial basis function as the kernel function, and uses cross-validation to determine the optimal model parameters (such as penalty coefficient C, kernel function parameters);

[0057] Future market sales forecast: Use the trained support vector machine model to predict the forecast data, and the prediction result is the future market sales.

[0058] The product acceptance prediction unit 202 in the integrated multi-model module 200 uses the sentiment dictionary to analyze consumer sentiment and predict product acceptance, specifically including:

[0059] Data collection: predicting textual reviews about products from social media data;

[0060] Text preprocessing: including standardization, stop word removal, stemming, etc.

[0061] Sentiment analysis: sentiment scores are calculated using sentiment dictionaries. The sentiment tendency (positive, neutral, negative) of each comment is calculated based on the sentiment dictionary score.

[0062] Aggregation and analysis: Taking products as units, count and average the sentiment scores of each product and convert them into product acceptance.

[0063] The visual element prediction unit 203 in the integrated multi-model module 200 uses a convolutional neural network algorithm model to analyze the correlation between the product visual elements and the current sales level according to the webpage visual data, and identifies the highly accepted visual elements, specifically including:

[0064] Data preparation: Collect web page visual data and the current sales level of related products as training data. The current sales level includes high sales, medium sales and low sales. Web page visual data includes but is not limited to product images, colors and layouts, etc. Convert these visual data into a format suitable for convolutional neural network processing, such as adjusting the images to the same resolution and size;

[0065] Selection and configuration of convolutional neural network: Select the VGGNet convolutional neural network model to configure the network's layers and parameters, and build a deep network by repeatedly using small-sized convolution kernels and pooling layers, including the input layer: accepting raw image input; multiple convolution layers: each convolution layer includes a series of convolution operations and nonlinear activation functions; pooling layer: used to reduce the spatial dimension of the feature map, using maximum pooling; fully connected layer: used to convert the feature map output by the convolution layer into a prediction result;

[0066] Model training: The input data is web page visual data, and the current sales level is used as a label. The network is trained using labeled training data to optimize network weights so that the model can predict the current sales level based on the input web page visual data.

[0067] Feature extraction and analysis: Use the previously trained convolutional neural network to process the webpage visual data of each product. When processing images, the convolutional neural network extracts deep features of the entire image or part of the image. These features include color, shape, pattern, texture, etc. In the convolutional neural network, each layer (especially the convolutional layer and pooling layer) extracts different characteristics of the input data. As the data is transmitted in the network, primary features such as edges and corners are gradually combined into more complex features. These complex features can describe important visual patterns in the image;

[0068] Prediction of highly acceptable visual elements: By extracting features and analyzing the output of the convolutional neural network model, determine which features are related to the current sales of the product. If a certain feature appears most frequently in the images of high-selling products, the feature is regarded as a highly acceptable visual element. Highly acceptable visual elements may include specific color combinations (such as warm colors may be more popular than cool colors), specific shapes or graphics (such as circles may be more attractive to consumers than sharp-angle graphics), or specific ways of arranging images (such as simple and elegant layouts are more popular than crowded and complex layouts).

[0069] Through the above steps, a convolutional neural network is used to analyze the relationship between the visual data of the web page and the current sales level, and to identify the visual elements that have the greatest impact on the current sales level.

[0070] The integrated multi-model module 200 provides a comprehensive approach to enhance the market performance and acceptance of cultural and creative products by integrating various units of market sales forecasting, product acceptance analysis and high-acceptance visual element prediction. First, the future market sales forecasting unit 201 uses a support vector machine to analyze historical market data, and helps companies develop more effective production and market strategies through accurate prediction models. Secondly, the product acceptance forecasting unit 202 evaluates the public's emotional response to the product through sentiment analysis of social media data, enabling companies to adjust product features to better meet consumer needs. Finally, the visual element prediction unit 203 analyzes the product's web page visual data through a convolutional neural network, identifies those visual features associated with high sales, and provides a basis for the design team to create a product appearance that is more attractive to consumers.

[0071] The consistency calculation unit 301 in the integrated decision module 300 receives the future market sales volume prediction unit 201 and the product acceptance prediction unit 202 respectively predicting the future market sales volume and product acceptance, and compares the product features and visual elements, specifically including:

[0072] Sort all products by product acceptance from high to low in order to identify products to focus on and products that may require market strategy adjustments;

[0073] For each product, compare its future market sales and product acceptance, and select products with high acceptance but low future sales forecasts, because these products may have potential that the market has not fully utilized. The selection steps are as follows:

[0074] Set thresholds: Set a minimum threshold for the acceptance score (e.g., 0.8), indicating the minimum social media positive reviews that the product needs to achieve; set a maximum threshold for predicted sales (e.g., 30% of the market average predicted sales), indicating that the predicted sales of the product must not be higher than this value;

[0075] Build a decision tree:

[0076] Step 1: Check whether the acceptance score of each product is greater than or equal to the minimum acceptance score threshold. Products that meet the conditions will enter the next step of inspection, and those that do not meet the conditions will be directly excluded. Step 2: For the products screened in the first step, further check whether their predicted sales are less than the maximum threshold of predicted sales. Only when the predicted sales of the product are lower than this threshold, the product will be marked as "high acceptance but low future market sales";

[0077] For products with high acceptance but low sales forecast, the current visual elements of the product are extracted, and the webpage visual data corresponding to the product is input into the visual element prediction unit 203, and the highly accepted visual elements of the product are output, and similarity comparison is performed, as follows:

[0078] The cosine similarity formula is used to evaluate the similarity between the two sets of feature vectors of the current visual element and the highly accepted visual element:

[0079] ,in and are two sets of eigenvectors, represents vector dot product, and is the Euclidean norm of the vector;

[0080] Threshold judgment: Set a threshold, for example, 0.8, which means that if the similarity of two feature vectors is higher than this threshold, they are considered consistent; if it is lower than this threshold, they are considered inconsistent.

[0081] The decision unit 302 in the integrated decision module 300 makes a decision according to the consistency result of the consistency calculation unit 301, which specifically includes:

[0082] If the current visual elements are inconsistent with the highly acceptable visual elements, redesign the product's web interface based on the highly acceptable visual elements, including:

[0083] List the key features of highly receptive visual elements: color combinations, layout methods, graphic elements, etc.

[0084] Use quantitative metrics to describe these visual elements, including:

[0085] Color analysis: Extract the color histogram of the highly acceptable image and calculate the main color ratio. ;

[0086] Layout analysis: Identify layout structures, such as partition ratios, element alignment, etc., through image segmentation technology;

[0087] Develop visual design principles: Develop design principles based on the results of color and layout analysis. For example, if the analysis shows that warm colors are more popular, choose warm colors as the main color when designing; if the data shows that simple layouts are more attractive to consumers, reduce clutter and use large areas of white space when designing;

[0088] Design implementation: Apply the above design principles and quantitative indicators extracted from high-acceptance visual elements to design web pages, including:

[0089] Color instantiation: Adjust the CSS style of the web interface to set the color code to match the most prominent color of the color histogram in the high-reception visual element;

[0090] Layout optimization: Adjust the content layout through CSS and HTML to ensure that the layout is consistent with the layout indicators of high-acceptance products.

[0091] Through the above steps, we can ensure that the product visual design is closer to the highly accepted visual elements that have been verified by the market, thereby improving the product's acceptance and sales in the market. This method is not only based on data-driven decision-making, but also has strong adaptability and can be adjusted for different products and market environments.

[0092] If the current visual element is consistent with the high acceptance visual element, the convolution layer of the convolution neural network algorithm model in the visual element prediction unit 203 is adjusted to better capture the features, and the current number of filters in the convolution layer is adjusted according to the formula: new number of filters = current number of filters × adjustment factor, so as to optimize the convolution neural network algorithm model; the highly acceptance visual elements are re-identified using the optimized convolution neural network algorithm model, and the highly acceptance visual elements before and after the optimization are compared. If they are consistent, those products that still show high acceptance but low predicted sales even after the convolution neural network optimization are marked as abnormal, and sent to management personnel as abnormal results.

[0093] The consistency calculation unit 301 and the decision unit 302 in the integrated decision module 300 provide a systematic and data-driven approach to improving the market performance and visual appeal of products by accurately matching consumer preferences and market demands; the consistency calculation unit 301 first receives data obtained from the future market sales forecast unit 201 and the product acceptance forecast unit 202, and based on these data, evaluates and compares the market sales forecasts and consumer acceptance of different products. In particular, the unit reveals market potential that may not be fully utilized by sorting products by acceptance and identifying those products that have high acceptance but low sales forecasts; this approach enables companies to analyze and adjust strategies with a clear goal, focusing on products that have potential but have not yet maximized their market influence.

[0094] Furthermore, by comparing and analyzing the visual elements of these products with high acceptance but low sales forecasts with known high acceptance visual elements, the consistency calculation unit 301 uses cosine similarity to evaluate the consistency between the two. If inconsistency is found, the decision unit 302 will intervene and put forward specific design improvement suggestions based on the data. For example, by analyzing the key visual features such as the color histogram and layout structure of the highly accepted visual elements, and using these features as the basis for the design of new products, it is ensured that the new design is more in line with the acceptance standards verified by the market. This comprehensive adjustment not only improves the visual appeal of the product, but also enhances market competitiveness and consumers' willingness to buy. If the current visual elements are consistent with the highly accepted visual elements, the convolutional neural network model will be further optimized to ensure that the most effective visual features are continuously captured. This reflects a dynamic improvement process that accurately responds to changes in the market and consumers, so that product design and market strategies are always kept in the best state.

[0095] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A product marketing trend prediction system for evaluating cultural creativity, characterized in that: The system comprises a data collection module (100), an integrated multi-model module (200) and an integrated decision module (300), wherein: The data collection module (100) collects data related to cultural and creative products, including historical market data, social media data, and web page visual data; The integrated multi-model module (200) uses a support vector machine algorithm model to predict the future market sales of each product based on the historical market data; uses a sentiment dictionary to predict product acceptance based on the social media data; and uses a convolutional neural network algorithm model to predict highly accepted visual elements based on the webpage visual data; the integrated multi-model module (200) includes a visual element prediction unit (203), and the visual element prediction unit (203) uses a convolutional neural network algorithm model to analyze the relationship between product visual elements and current sales levels based on the webpage visual data, and identifies highly accepted visual elements, specifically including: Data preparation: Collect web page visual data and the current sales level of related products as training data. The current sales level includes high sales, medium sales and low sales. The web page visual data includes but is not limited to product pictures, colors and layouts, and convert the web page visual data into a format suitable for convolutional neural network processing; Selection and configuration of convolutional neural network: Select the VGGNet convolutional neural network model to configure the network's layers and parameters, and build a deep network by repeatedly using small-sized convolution kernels and pooling layers, including input layers, multiple convolutional layers, pooling layers, and fully connected layers; Model training: The input data is web page visual data, and the current sales level is used as a label; the network is trained using labeled training data; Feature extraction and analysis: Use trained convolutional neural networks to process the web visual data of each product. When processing images, convolutional neural networks extract deep features of the entire image or part of the image. In convolutional neural networks, each layer extracts different characteristics of the input data. As data is passed through the network, primary features are combined into more complex features. Prediction of highly acceptable visual elements: By extracting features and analyzing the output of the convolutional neural network model, we determine which feature appears most frequently in the images of high-selling products and use it as a highly acceptable visual element. Highly acceptable visual elements include but are not limited to color combinations, shapes, and image arrangement methods. The integrated decision module (300) comprises a consistency calculation unit (301), wherein the consistency calculation unit (301) receives the future market sales volume prediction unit (201) and the product acceptance prediction unit (202) to respectively predict the future market sales volume and the product acceptance, and compares the product features and visual elements; The integrated decision module (300) predicts the future market sales of the products in order of the product acceptance from high to low, selects products with high product acceptance but low future market sales, extracts the visual elements corresponding to the products as the current visual elements, and compares the current visual elements with the high acceptance visual elements in the convolutional neural network algorithm model, wherein the integrated decision module (300) includes a decision unit (302), and the decision unit (302) makes a decision based on the consistency result of the consistency calculation unit (301), wherein if the current visual element is inconsistent with the high acceptance visual element, the web page interface of the product is redesigned based on the high acceptance visual element, specifically including: Sort all products from high to low according to product acceptance; For each product, compare its future market sales and product acceptance, and select products with high acceptance but low future predicted sales; For products with high acceptance but low sales forecast, the current visual elements of the product are extracted, and the webpage visual data corresponding to the product is input into the visual element prediction unit (203), and the highly accepted visual elements of the product are output, and similarity comparison is performed, as follows: The cosine similarity formula is used to evaluate the similarity between the two sets of feature vectors of the current visual element and the highly accepted visual element: ,in and are two sets of eigenvectors, represents vector dot product, and is the Euclidean norm of the vector; Threshold judgment: Set a threshold. If the similarity of two feature vectors is higher than this threshold, they are considered consistent; if it is lower than this threshold, they are considered inconsistent. If not, the product's web interface will be redesigned based on the highly accepted visual elements, including: List the key features of highly receptive visual elements; Use quantitative metrics to describe these visual elements, including: Color analysis: Extract the color histogram of the highly acceptable image and calculate the main color ratio. ; Layout analysis: Identify layout structures through image segmentation techniques, including but not limited to partition proportions and element alignment; Visual design principle formulation: formulate design principles based on the results of color and layout analysis; Design Implementation: Apply design principles and quantitative metrics extracted from high-perception visual elements to design web pages, including: Color instantiation: Adjust the CSS style of the web interface to set the color code to match the most prominent color of the color histogram in the high-reception visual element; Layout optimization: adjust the content layout through CSS and HTML to ensure that the layout is consistent with the layout indicators of high-acceptance products; If they are consistent, the convolutional neural network algorithm in the integrated multi-model module (200) is optimized, the high acceptance visual elements are re-predicted, and the high acceptance visual elements before and after the optimization are compared. If they are consistent, the product is marked and sent to the management personnel as an abnormal result.

2. The product marketing trend prediction system for evaluating cultural creativity according to claim 1 is characterized in that: The integrated multi-model module (200) comprises a future market sales volume prediction unit (201), wherein the future market sales volume prediction unit (201) uses a support vector machine algorithm to predict future market sales volume based on historical market data, specifically comprising: Data preparation: Extract features from historical market data, including but not limited to time series features, prices, promotional activities, actual sales, and divide the data into training data and prediction data; Feature selection: Use correlation analysis methods to reduce feature dimensions and select the most representative features; Support vector machine model training: define the loss function as ,in Represents the loss function, which is used to calculate and express the gap between the predicted value of the support vector machine model and the actual value; Represents the actual value, the actual sales value obtained from the data; represents the predicted value of the support vector machine model, The support vector machine model learns by inputting training data The prediction function obtained; The input features of the training data representing the model are extracted from historical market data; the radial basis function is selected as the kernel function, and cross-validation is used to determine the optimal model parameters; Future market sales forecast: Use the trained support vector machine model to predict the forecast data, and the prediction result is the future market sales.

3. The product marketing trend prediction system for evaluating cultural creativity according to claim 2 is characterized in that: The integrated multi-model module (200) includes a product acceptance prediction unit (202), which uses a sentiment dictionary to analyze consumer sentiment and predict product acceptance, predicts text reviews about products from social media data, performs text preprocessing on the text reviews by standardizing, removing stop words and extracting stems, calculates sentiment scores using the sentiment dictionary, and takes products as units, calculates and averages the sentiment scores of each product, and converts them into product acceptance.

4. The product marketing trend prediction system for evaluating cultural creativity according to claim 3 is characterized in that: When the consistency result in the decision unit (302) is inconsistent, the convolutional layer of the convolutional neural network algorithm model in the visual element prediction unit (203) is adjusted to adjust the current number of filters in the convolutional layer, and the formula is new number of filters = current number of filters × adjustment factor; The optimized convolutional neural network algorithm model is used to re-identify the highly accepted visual elements and compare the highly accepted visual elements before and after the optimization. If they are consistent, the product is marked as abnormal and sent to the management staff as an abnormal result.

Images