Garment trend prediction and analysis method based on reinforcement learning

Abstract

The invention discloses a garment trend prediction and analysis method based on reinforcement learning, and relates to the technical field of garment trend prediction based on reinforcement learning, and the method comprises the steps: collecting multi-modal data, carrying out the preprocessing, and generating multi-modal garment data; performing trend scene evolution on the multi-modal clothing data through a trend evolution learning model to obtain a trend evolution strategy, and generating a trend development path; optimizing a production plan, inventory distribution and logistics scheduling according to the innovation style data set in combination with historical sales data and a market demand prediction result, and generating an inventory production plan; and collecting market response feedback data, performing inventory and production adjustment in combination with the inventory production plan, obtaining an inventory adjustment suggestion, identifying the difference between the inventory and the market demand, and generating a trend analysis report. According to the overall scheme, through dynamic prediction and real-time adjustment, efficient response of market demands is ensured, and the production efficiency and market adaptability of the clothing industry are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of reinforcement learning clothing trend prediction, in particular to a clothing trend prediction and analysis method based on reinforcement learning. BACKGROUND

[0002] With the continuous development of artificial intelligence technology, especially the wide application of deep learning and reinforcement learning, the trend prediction and analysis method of the clothing industry has also been improved; in recent years, many studies have begun to explore the use of machine learning-based methods, especially deep learning and reinforcement learning technology, to improve the accuracy and adaptability of prediction through automated data analysis.

[0003] Although the trend prediction method based on artificial intelligence improves the prediction accuracy to a certain extent, the existing technology still faces some challenges; traditional reinforcement learning models usually rely on a single data source dataset, and cannot effectively integrate heterogeneous data from multiple dimensions, such as visual images, consumer behavior data and market environment information. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides a clothing trend prediction and analysis method based on reinforcement learning, which solves the problem of being unable to effectively integrate multi-modal heterogeneous data.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] The present application provides a clothing trend prediction and analysis method based on reinforcement learning, which includes,

[0008] Collecting multi-modal data and preprocessing to generate multi-modal clothing data;

[0009] Evolution of trend learning model for multi-modal clothing data trend scenario evolution, get trend evolution strategy, and generate trend development path;

[0010] Input the trend development path into the trend prediction model, and perform trend prediction and style generation to output an innovative style dataset;

[0011] According to the innovative style dataset, combining historical sales data and market demand prediction results, optimizing production plan, inventory allocation and logistics scheduling, generating inventory production plan;

[0012] Collect market reaction feedback data, combine inventory production plan to adjust inventory and production, get inventory adjustment suggestion, and identify the gap between inventory and market demand, generate trend analysis report.

[0013] As a preferred embodiment of the reinforcement learning-based clothing trend prediction and analysis method of the present invention, the specific steps for collecting multimodal data and preprocessing it are as follows:

[0014] Visual image data of clothing is acquired using a high-definition camera, and friction sound data when clothing comes into contact with the human body is acquired using a microphone to generate a comprehensive clothing dataset.

[0015] The comprehensive clothing dataset is cleaned and denoised to generate a cleaned comprehensive clothing dataset.

[0016] The Z-score normalization method was used to transform the cleaned clothing dataset into a standard normal distribution, generating a standardized clothing dataset.

[0017] Deep features are extracted from the visual image data in the standardized clothing dataset, and all image features are aggregated to generate an image feature dataset.

[0018] As a preferred embodiment of the reinforcement learning-based clothing trend prediction and analysis method of the present invention, the generation of multimodal clothing data refers to constructing a graph structure by combining a standardized clothing dataset and an image feature dataset according to temporal order and spatial relationship, and then using a graph convolutional network for data fusion.

[0019] As a preferred embodiment of the reinforcement learning-based clothing trend prediction and analysis method of the present invention, the specific steps for obtaining trend evolution strategies by performing trend scenario evolution on multimodal clothing data through a trend evolution learning model are as follows.

[0020] Multimodal clothing data is input into the trend evolution learning model to learn the long-term dependencies of clothing trends and generate a trained trend evolution learning model.

[0021] The standardized clothing dataset is input into the trained trend evolution learning model, and the time series of the standardized clothing dataset is subjected to trend scenario evolution to extract long-term dependencies and trend change patterns, thereby generating a trend evolution strategy.

[0022] As a preferred embodiment of the reinforcement learning-based clothing trend prediction and analysis method of the present invention, the specific steps for generating the trend development path are as follows:

[0023] Based on the trend evolution strategy, a multi-objective optimization algorithm is used to evaluate the trend accuracy and compare the market adaptability of multiple trend paths in the trained trend evolution learning model, and generate preliminary trend path data.

[0024] By combining preliminary trend path data with historical sales data, and evaluating the market adaptability and sales potential of different preliminary trend path data, a market adaptability score is generated.

[0025] According to the market adaptability score, the trend paths with high market adaptability are screened from the preliminary trend path data to generate screened preliminary trend paths;

[0026] The market adaptability of the screened preliminary trend paths is adjusted to generate trend development paths.

[0027] As a preferred scheme of the clothing trend prediction and analysis method based on reinforcement learning, the trend development paths are input into a trend prediction model, and trend prediction and style generation are performed to output an innovative style dataset, and the specific steps are as follows,

[0028] The trend development paths are input into a trend prediction model, and historical trend change data and market data are combined to predict future trend evolution paths by using a time series analysis method to generate trend prediction data;

[0029] The trend prediction data are combined with market demand and popular elements, and the sales potential and market adaptability of each style are analyzed to generate an innovative style dataset.

[0030] As a preferred scheme of the clothing trend prediction and analysis method based on reinforcement learning, the trend development paths are input into a trend prediction model, and historical trend change data and market data are combined to predict future trend evolution paths by using a time series analysis method to generate trend prediction data;

[0031] The trend prediction data are combined with market demand and popular elements, and the sales potential and market adaptability of each style are analyzed to generate an innovative style dataset.

[0032] The production plan, inventory allocation and logistics scheduling are extracted from the original inventory production plan;

[0033] The production demand is combined with the existing production capacity and working hours to optimize the production sequence and resource allocation in the production plan to generate an optimized production plan;

[0034] The inventory demand and the existing inventory situation are combined to optimize the inventory allocation proportion of each interval in the inventory allocation to generate an optimized inventory allocation;

[0035] The logistics demand, existing inventory and transportation capacity are combined to optimize the distribution route and sequence in the logistics scheduling to generate an optimized logistics scheduling.

[0036] As a preferred scheme of the clothing trend prediction and analysis method based on reinforcement learning, the trend development paths are input into a trend prediction model, and historical trend change data and market data are combined to predict future trend evolution paths by using a time series analysis method to generate trend prediction data;

[0037] As a preferred scheme of the clothing trend prediction and analysis method based on reinforcement learning, wherein: the market reaction feedback data is collected, combined with the inventory production plan to adjust the inventory and production, and the inventory adjustment suggestion is obtained, and the specific steps are as follows,

[0038] The market reaction feedback data, sales channel data and consumer research results are collected and data cleaning is performed to generate feedback cleaning data.

[0039] The feedback cleaning data is combined with the inventory production plan, and the inventory allocation and inventory production plan are adjusted through the particle swarm optimization algorithm to generate inventory production adjustment data.

[0040] The inventory production adjustment data is combined with the market demand and the current inventory situation, and the difference analysis method is used to quantitatively analyze the gap between the inventory quantity, demand quantity and inventory production plan, identify the inventory insufficient area, and obtain the inventory adjustment suggestion.

[0041] As a preferred scheme of the clothing trend prediction and analysis method based on reinforcement learning, wherein: the market reaction feedback data is collected, combined with the inventory production plan to adjust the inventory and production, and the inventory adjustment suggestion is obtained, and the specific steps are as follows,

[0042] The inventory adjustment suggestion is combined with the market demand and the inventory production plan, the influence of inventory adjustment on the supply chain is evaluated, and the optimization effect of inventory adjustment on demand satisfaction and cost benefit is measured to generate an inventory adjustment report.

[0043] The inventory adjustment report is combined with the historical sales data, market demand prediction results and inventory adjustment suggestion, and the influence of inventory adjustment on sales trend and market demand fluctuation is evaluated to generate a trend analysis report.

[0044] The present application has the following advantages: by collecting and preprocessing multi-modal clothing data, using a trend evolution learning model to extract long-term trend changes, providing accurate guidance for subsequent trend path generation; combining historical sales data and market demand prediction, optimizing production planning, inventory allocation and logistics scheduling, improving the efficiency of the supply chain, and adjusting inventory and production through market feedback data to generate a trend analysis report, thereby realizing fine production and inventory management; the overall scheme ensures efficient response to market demand through dynamic prediction and real-time adjustment, improving the production efficiency and market adaptability of the clothing industry. BRIEF DESCRIPTION OF DRAWINGS

[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0046] Fig. 1 Flow chart of the clothing trend prediction and analysis method based on reinforcement learning.

[0047] Fig. 2 Flow chart of data collection and preprocessing.

[0048] Fig. 3 Flow chart of trend scenario evolution.

[0049] Fig. 4 Flow chart of supply chain optimization. DETAILED DESCRIPTION

[0050] In order to make the above objectives, features and advantages of the present application more apparent, specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0051] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be apparent to one skilled in the art that the present application can be practiced without the specific details and other implementations can be employed. Therefore, the present application is not limited to the details described herein and can be practiced with variations that are within the scope and spirit of the claims.

[0052] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. The "in one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments.

[0053] Reference Figs. 1-4 For one embodiment of the present application, the embodiment provides a clothing trend prediction and analysis method based on reinforcement learning, comprising the following steps:

[0054] S1, collect multi-modal data and perform preprocessing to generate multi-modal clothing data.

[0055] It should be noted that by using a high-definition camera to collect visual image data of the garment and a microphone to collect friction sound data when the garment is in contact with the human body, a comprehensive garment data set is generated; based on the comprehensive garment data set, data cleaning and denoising are performed to generate a cleaned comprehensive garment data set; the Z-score standardization method is used to standardize the cleaned comprehensive garment data set to generate a standardized garment data set; the depth image features are extracted from the visual image data in the standardized garment data set and feature aggregation processing is performed to generate an image feature data set; the standardized garment data set and the image feature data set are constructed into a graph structure according to the time sequence and spatial relationship, and data fusion processing is performed, so that the feature information of the standardized garment data set and the image feature data set is consistently expressed and forms a correlated expression feature in the constructed graph structure, generating a multi-modal garment data.

[0056] S1.1, a high-definition camera is used to collect visual image data of the garment, and a microphone is used to collect friction sound data when the garment is in contact with the human body, to generate a comprehensive garment data set.

[0057] It should be noted that the high-definition camera is used to take pictures of the garment to collect visual image data of the garment, ensuring the clarity and detail of each frame of image; the microphone is used to capture the friction sound data generated when the garment is in contact with the human body, recording the frequency and intensity of the sound signal. The collected visual image data and friction sound data are combined to generate a comprehensive garment data set.

[0058] S1.2, data cleaning and denoising are performed on the comprehensive garment data set to generate a cleaned comprehensive garment data set.

[0059] It should be noted that the missing value processing is performed on the visual image data and the friction sound data to fill in the missing data that may be generated during the collection process; the noise in the visual image data is filtered out to ensure that the image details are clear and free of unnecessary interference; the background noise in the friction sound data is suppressed to retain the sound features related to the contact between the garment and the human body, generating a cleaned comprehensive garment data set.

[0060] S1.3, the Z-score standardization method is used to convert the cleaned comprehensive garment data set into a standard normal distribution to generate a standardized garment data set.

[0061] It should be noted that the Z-score standardization method is used to calculate the mean and standard deviation of each feature in the cleaned comprehensive garment data set; for each data point in each feature, the data point is subtracted from the mean of the corresponding feature, and the result is divided by the standard deviation, so that the data is converted into a standard normal distribution, ensuring that the mean of each feature is 0 and the standard deviation is 1. After this conversion, all data in the cleaned comprehensive garment data set will be standardized to generate a standardized garment data set.

[0062] S1.4. Perform deep feature extraction on the visual image data in the standardized clothing dataset, and aggregate all image features to generate an image feature dataset.

[0063] It should be noted that a convolutional neural network is used to perform multi-level feature learning on each image, extracting local and global features at different levels. Through convolutional and pooling layers, basic features such as edges, textures, and shapes, as well as higher-level structural features, are captured. During feature extraction, non-linear transformations are applied to the image features, making them richer and more discriminative. All extracted image features are then aggregated to form a unified image feature dataset.

[0064] S1.5 Construct a graph structure by combining the standardized clothing dataset and the image feature dataset according to time order and spatial relationship, and use a graph convolutional network to fuse the data to generate multimodal clothing data.

[0065] It should be noted that time alignment and spatial matching are performed on each data point in the standardized clothing dataset and each feature point in the image feature dataset to ensure that the data at corresponding timestamps and spatial locations can be associated in the graph structure. Nodes represent data at each time point and spatial location, while edges represent the relationships and dependencies between data points. The constructed graph structure effectively links the information in the standardized clothing dataset and the image feature dataset, providing a foundation for data fusion. A graph convolutional network is used to process the constructed graph structure, enabling information propagation and feature learning through the connections between nodes, thereby achieving deep fusion of the standardized clothing dataset and the image feature dataset to generate multimodal clothing data.

[0066] S2. Use a trend evolution learning model to perform trend scenario evolution on multimodal clothing data, obtain trend evolution strategies, and generate trend development paths.

[0067] It should be noted that existing methods typically rely on traditional statistical models to predict apparel trends, often analyzing only historical sales data, market demand, or fashion trends, lacking the integration and prediction of evolving trends from multimodal data (such as images, sound, and consumer behavior data). Traditional methods struggle to capture the complex relationships and dynamic changes between multidimensional and multi-source data.

[0068] This invention utilizes a trend evolution learning model, employing deep learning and time series analysis to perform trend scenario evolution on multimodal clothing data. This method can simultaneously process multimodal data from visual, auditory, and kinematic sources, predicting multiple possible trend paths through model evolution and generating the optimal trend development path based on actual conditions. This approach not only considers the combined influence of multiple data sources but also dynamically adjusts trend prediction strategies, improving the accuracy and flexibility of predictions.

[0069] S2.1 Input multimodal clothing data into the trend evolution learning model to learn the long-term dependencies of clothing trends and generate a trained trend evolution learning model.

[0070] It should be noted that visual image data, friction sound data, and other relevant data from multimodal clothing data are integrated to ensure temporal alignment and feature matching between different data sources. This data is then used to train a trend evolution learning model. By learning the long-term dependencies of clothing trends, the trend evolution learning model can capture time-series patterns and trend change regularities in the data. During training, the trend evolution learning model iterates through multiple rounds, continuously adjusting parameters to optimize its predictive ability for clothing trend evolution, resulting in a post-trained trend evolution learning model.

[0071] It should also be noted that the trend evolution learning model is built upon existing deep learning models, such as recurrent neural networks (RNNs), long short-term memory networks (LSTMs), and time series analysis methods. It effectively captures long-term dependencies and trend change patterns in data, and is particularly suitable for processing time-dependent multimodal data. The trend evolution learning model further introduces a multimodal data fusion mechanism, combining visual image data, friction sound data, and consumer behavior data to enhance its predictive ability for trend evolution, thus more accurately reflecting the dynamic changes and multidimensional influencing factors of clothing trends.

[0072] S2.2 Input the standardized clothing dataset into the trained trend evolution learning model, and perform trend scenario evolution on the time series of the standardized clothing dataset to extract long-term dependencies and trend change patterns, and generate trend evolution strategies.

[0073] It should be noted that the data at each time point in the standardized clothing dataset should be arranged chronologically to maintain the integrity of the time series. In the trained trend evolution learning model, this time series data is used for trend scenario evolution analysis. The trained trend evolution learning model extracts the changing patterns of clothing trends by learning long-term dependencies in historical data. During the evolution process, the model captures the evolutionary patterns of clothing trends by identifying the correlations and pattern changes between different time points, and generates trend evolution strategies.

[0074] S2.3. Based on the trend evolution strategy, a multi-objective optimization algorithm is used to evaluate the trend accuracy and compare the market adaptability of multiple trend paths in the trained trend evolution learning model, and generate preliminary trend path data.

[0075] It should be noted that multiple trend paths are selected as optimization targets. These trend paths originate from the output of the post-trained trend evolution learning model and represent the possible evolutionary directions of clothing trends under different scenarios. The predicted data for each trend path is compared with historical trend data, and statistical measures such as mean squared error (MSE) and correlation coefficient are used to measure the difference between the two, thereby judging the accuracy of the prediction results. A small correlation coefficient indicates that the predicted trend path is closer to historical trends, indicating high accuracy. The post-trained trend evolution learning model also compares the market adaptability of each trend path. Multiple market factors are considered for training the post-trained trend evolution learning model, such as changes in market demand, changes in consumer behavior, and fluctuations in market trends. By analyzing the impact of these factors on the trend paths, the actual applicability under different market conditions is evaluated. The post-trained trend evolution learning model calculates the adaptability of each trend path in the future market environment based on historical data of consumer preferences, changes in market supply and demand, industry trends, and competitor dynamics. For example, by analyzing whether a certain trend path can meet the changing needs of a specific consumer group, the applicability assessment result of the path in the market is obtained, generating preliminary trend path data.

[0076] The fitness expression for each trend path in the future market environment is calculated as follows:

[0077] ;

[0078] in: For the first The adaptability of the trend path; The number of time periods to be evaluated indicates the prediction period length of the trend path; This is a time point index, representing each specific moment in the evaluation; For the first Trend path at time point Changes in standardized market demand; For the first Trend path at time point Standardized changes in consumer behavior; For the first Trend path at time point Standardized market trend fluctuations; This indicates changes in market demand; This indicates changes in consumer behavior; This indicates market trend fluctuations.

[0079] It should be noted that the standardized change in market demand is obtained by standardizing market demand change data, resulting in a change relative to the average value; the standardized change in consumer behavior is obtained by standardizing consumer behavior data, resulting in a change relative to the average value; the standardized market trend fluctuation is obtained by standardizing market trend fluctuation data, resulting in a change relative to the average value; the change in market demand is determined by analyzing the importance of the impact of market demand change on the trend path; the change in consumer behavior is determined by analyzing the degree of impact of consumer behavior change on the trend path; and the market trend fluctuation is determined by assessing the degree of impact of market trend fluctuation on the trend path.

[0080] S2.4 Combine the preliminary trend path data with historical sales data, and evaluate the market adaptability and sales potential of different preliminary trend path data to generate a market adaptability score.

[0081] It should be noted that each preliminary trend path data point in the preliminary trend path data is time-aligned with the relevant sales records in the historical sales data to ensure temporal consistency between each preliminary trend path data point and its corresponding historical sales data. By calculating the market adaptability of each trend path within its corresponding time period, its performance in historical sales data is evaluated, and the relationship between the trend path and market demand is identified. The degree of matching between the trend path and historical sales data is quantified, and the potential performance of each trend path in the future market is assessed based on sales potential. Based on the evaluation results, a market adaptability score is generated for each trend path.

[0082] S2.5. Based on the market adaptability score, select trend paths with high market adaptability from the preliminary trend path data and generate the selected preliminary trend paths.

[0083] It should be noted that all preliminary trend paths are sorted from highest to lowest based on market adaptability scores, and the data of the preliminary trend paths with the highest scores are selected. These high-scoring trend paths are further evaluated in terms of their performance against historical sales data and market demand to ensure they reflect strong market adaptability and sales potential. The selected trend paths with high market adaptability are then integrated to generate the filtered preliminary trend paths.

[0084] S2.6. Adjust the initial trend path after screening to adapt to the market and generate a trend development path.

[0085] It should be noted that the preliminary trend path after screening will be compared and analyzed with current market demand data to identify changing trends in market demand. Based on consumer preferences, demand fluctuations, and other market factors reflected in the market demand data, the preliminary trend path will be appropriately revised to ensure it better reflects the actual market environment. During the adjustment process, different dimensions of market demand data, such as seasonal fluctuations and changes in consumer behavior, will be considered to dynamically optimize the trend path and improve its market adaptability. The adjusted trend path will serve as the basis for the final trend development path.

[0086] S3. Input the trend development path into the trend prediction model, perform trend prediction and style generation, and output an innovative style dataset.

[0087] S3.1 Input the trend development path into the trend prediction model, and combine it with historical trend change data and market data. Use time series analysis to predict the future trend evolution path and generate trend prediction data.

[0088] It should be noted that the optimized trend development path is combined with historical trend change data and market data to ensure that all data are aligned on the timeline and that the spatial relationships between data points are consistent. Time series analysis methods are used to model the combined data. By analyzing the fluctuations, cyclical changes, and patterns in historical trends and market data, the trend prediction model can learn the patterns of trend evolution and predict future trend changes. Through training and iteration, the trend prediction model can generate trend prediction data based on patterns in historical data and current market changes.

[0089] It should also be noted that time series analysis is a statistical method used to analyze datasets with a time sequence, primarily aimed at identifying patterns such as trends, seasonality, periodicity, and randomness in the data. Through time series analysis, historical data can be modeled, the time dependencies of the data can be extracted, and these patterns can be used to predict future data.

[0090] Trend forecasting models are built upon existing time series analysis models, such as Autoregressive Integrated Moving Average (ARIMA), Long Short-Term Memory (LSTM) networks, and deep learning models like the Transformer. They are capable of handling trend, seasonal, and cyclical changes in time series data, capturing long-term dependencies. By combining historical trend change data with market demand data, trend forecasting models can learn and predict the evolution path of future trends.

[0091] S3.2 Combine trend forecast data with market demand and popular elements, and analyze the sales potential and market adaptability of each style to generate an innovative style dataset.

[0092] It should be noted that each trend path in the trend forecast data is matched with demand fluctuations in the market demand data. By calculating the correlation and fit between trend paths and market demand, the potential demand for each trend path in the current market is assessed. For example, statistical indicators such as correlation coefficients and mean squared error are used to measure the degree of matching between trend forecast data and market demand data. Combining popular element data with trend forecast data, and comparing the performance of popular elements in the current market, identifies which popular elements are consistent with changes in market demand and consumer preferences. This process can be achieved through retrospective evaluation of historical sales data, consumer behavior analysis, and popular trends. Through this analysis and comparison, it is possible to determine which combinations of trend paths and popular elements have high sales potential in the future market, generating an innovative style dataset.

[0093] The expression for calculating the correlation and fit between trend paths and market demand is:

[0094] ;

[0095] in: For the first Trend path and the first The Pearson correlation coefficient between market demand; For the first Trend path at time point Trend forecast data; For the first Market demand at a given time point Demand data on; For the first The average trend value of the trend path at all points in time; For the first The average demand value of the market at all points in time; This refers to the number of time periods, specifically the number of time points for trend forecast data and market demand data.

[0096] S4. Based on the innovative style dataset, combined with historical sales data and market demand forecasts, optimize production planning, inventory allocation, and logistics scheduling to generate an inventory production plan.

[0097] It should be noted that existing methods typically use historical sales data and market demand forecasts as a basis to optimize production planning, inventory allocation, and logistics scheduling. These methods generally rely on static data analysis, which may overlook the impact of rapidly changing market demands and emerging styles, and cannot fully integrate innovative style datasets with multi-dimensional demand forecast information.

[0098] This invention combines innovative style datasets, historical sales data, and market demand forecasts, employing a more sophisticated optimization algorithm to dynamically adjust production plans, inventory allocation, and logistics scheduling. This method can process multimodal data in real time, accurately identify the market potential of each innovative style, and thus optimize inventory and production plans. Unlike traditional methods, this invention effectively integrates the latest market trends and forecasts, flexibly adjusting to different demand changes to achieve comprehensive optimization of production, inventory, and logistics.

[0099] S4.1 Combine the innovative style dataset with historical sales data and market demand forecasts, and predict the production demand, inventory demand, and logistics demand for each innovative style.

[0100] It should be noted that the correlation coefficient between the sales volume and market demand of each innovative style is calculated to identify which styles have performed well in the past market, and the relationship between the sales patterns of these styles and market demand is analyzed. For example, by comparing sales data during different seasons and promotional periods, factors influencing sales volume can be identified, thereby determining potential best-selling styles in the future market. When combining market demand forecasts with the innovative style dataset, the future demand volume in the market demand forecast is compared with the design features and positioning of each innovative style to assess which innovative styles are more in line with future market trends. Specific methods include using trend forecasting models to assess the demand change trend of each style, and using regression analysis, difference analysis, and other methods to identify which styles have greater potential for demand growth in the future market.

[0101] S4.2 Extract production plans, inventory allocation, and logistics scheduling from the original inventory production plan.

[0102] It should be explained that, based on the data in the original inventory production plan, the production demand, production sequence, and specific time arrangements for each production cycle of each product are identified. Inventory allocation information is extracted from the original inventory production plan, the gap between current inventory and predicted demand is analyzed, the allocation ratio for each inventory range is determined, and an appropriate inventory quantity is allocated to each range. The logistics scheduling information in the original inventory production plan is analyzed to extract the transportation route, delivery method, and delivery time arrangement for each product, ensuring that the logistics plan matches the production plan and inventory allocation to avoid supply delays or inventory shortages. Through the extraction and analysis of this data, production plans, inventory allocation, and logistics scheduling are generated.

[0103] S4.3 Combine production demand with existing production capacity and working hours to optimize the production sequence and resource allocation in the production plan, and generate an optimized production plan.

[0104] It should be explained that the production quantity of each product should be extracted from the production demand and compared with the existing production capacity to assess whether the production task can be completed within the specified time. Considering the existing work schedule, the required work hours for each production stage should be analyzed to ensure that production tasks are rationally arranged within the time constraints. The production sequence in the production plan should be adjusted to ensure that the order of each production stage minimizes downtime and changeover costs. Based on the analysis results of existing production capacity and work hours, resource allocation should be optimized to ensure that the raw materials, equipment, and labor required for production are optimally configured, avoiding resource waste and production bottlenecks, and generating an optimized production plan.

[0105] S4.4 Combine the inventory demand and the existing inventory situation to optimize the inventory allocation ratio of each interval in the inventory allocation and generate the optimized inventory allocation.

[0106] It should be explained that the demand for each product is extracted from the inventory demand and compared with the current inventory level to understand whether the current inventory can meet the predicted demand. Based on the demand and the current inventory level, the inventory pressure of each inventory range is analyzed to identify which ranges have insufficient inventory and which ranges may have excess inventory. The inventory allocation ratio of each range is optimized to ensure that the inventory in each range is reasonably allocated according to the demand and to avoid inventory backlog. In the optimization process, factors such as inventory turnover rate, storage costs, and delivery time also need to be considered to ensure that the inventory allocation plan can not only meet market demand but also maximize inventory efficiency and reduce costs, generating an optimized inventory allocation.

[0107] S4.5 Combine logistics demand, existing inventory and transportation capacity to optimize delivery routes and sequences in logistics scheduling and generate optimized logistics scheduling.

[0108] It should be explained that, based on logistics needs, the delivery demand for each order is determined, and the available inventory for each product is assessed against existing inventory levels. Factors such as vehicle capacity, transit time, and costs are analyzed in conjunction with transportation capacity to ensure each delivery is completed within the available capacity. Delivery routes and sequences in the logistics scheduling are adjusted to optimize delivery paths, reducing transit time and costs and ensuring efficient logistics operations. During optimization, the timeliness and regional distribution of deliveries are considered, and the priority of each delivery task is rationally assigned to ensure that high-priority orders are delivered as quickly as possible, generating an optimized logistics schedule.

[0109] S4.6 Integrate the optimized production plan, optimized inventory allocation, and optimized logistics scheduling, and perform resource integration to generate an inventory production plan.

[0110] It should be noted that the production sequence, resource allocation, and production task arrangement in the optimized production plan should be compared and combined with the allocation ratio of each inventory interval in the optimized inventory allocation to ensure coordination between the production plan and inventory allocation. Combined with the delivery routes and sequences in the optimized logistics scheduling, the priority of each delivery task should be analyzed to ensure that production tasks and inventory allocation can be executed efficiently under the constraints of logistics scheduling. During the integration process, through resource integration, production resources, inventory, and logistics distribution are rationally arranged, enabling each production task to obtain the necessary resources in a timely manner and ensuring that logistics distribution can meet production needs in the shortest possible time, generating an inventory production plan.

[0111] S5. Collect market response feedback data, combine it with inventory and production plans to adjust inventory and production, obtain inventory adjustment suggestions, identify the gap between inventory and market demand, and generate trend analysis reports.

[0112] S5.1 Collect market response feedback data, sales channel data, and consumer survey results, and perform data cleaning to generate feedback cleaned data.

[0113] It should be noted that market feedback data is collected from various channels, including consumer reviews, purchasing trends, and product feedback; sales data is collected from major sales channels, recording sales volume, sales revenue, and promotional effectiveness across different channels; consumer surveys are used to collect data on consumer needs, preferences, and purchasing motivations, and this data is organized in a standardized format to ensure consistency and completeness of data sources. Data cleaning is performed to remove duplicates, missing values, and outliers, ensuring data accuracy and consistency, and generating cleaned feedback data.

[0114] S5.2 Combine the feedback cleaning data with the inventory production plan, and adjust the inventory allocation and inventory production plan through the particle swarm optimization algorithm to generate inventory production adjustment data.

[0115] It should be noted that by analyzing market demand changes, consumer preferences, and sales performance in the feedback cleansing data, potential inventory shortages in the inventory production plan are identified. The feedback cleansing data is then combined with the inventory demand, production capacity, and current inventory status in the inventory production plan to assess the rationality of inventory allocation and revise the inventory allocation in the production plan. An optimization algorithm is then used to optimize the existing inventory production plan based on the adjusted inventory allocation. The Particle Swarm Optimization algorithm simulates the flight process of particles in the solution space, continuously adjusting the inventory allocation scheme to find the optimal solution, ensuring that the inventory allocation can maximize the satisfaction of market demand, reduce inventory backlog and shortages, and generate inventory production adjustment data.

[0116] S5.3 Combine inventory production adjustment data with market demand and current inventory status, and use the difference analysis method to quantitatively analyze the gap between inventory, demand and inventory production plan, identify areas of insufficient inventory, and obtain inventory adjustment suggestions.

[0117] It should be explained that inventory allocation and production scheduling are adjusted through inventory production data, and this data is compared with the projected demand and current inventory levels in the market demand data. A variance analysis method is used to quantitatively analyze the gaps between inventory levels, demand, and inventory production plans. During the variance analysis, the gap between inventory levels and demand for each product is calculated to identify areas of excess or insufficient inventory, especially those where inventory levels are insufficient to meet market demand. Through quantitative analysis, the impact of these gaps on inventory production plans is further assessed to determine the inventory levels that need adjustment and to obtain inventory adjustment recommendations.

[0118] S5.4 Integrate inventory adjustment recommendations with market demand and inventory production plans, assess the impact of inventory adjustments on the supply chain, measure the optimization effect of inventory adjustments on demand satisfaction and cost-effectiveness, and generate an inventory adjustment report.

[0119] It should be noted that the difference between the inventory adjustment amount for each product and the predicted market demand is calculated. For example, variance analysis is used to assess whether the adjusted inventory can meet the expected market demand, ensuring that product supply is neither excessive nor insufficient. Based on this, and in conjunction with the production schedule in the inventory production plan, the impact of inventory adjustments on the execution of the production plan is analyzed, and whether inventory changes will lead to production delays or resource waste is examined. The overall impact of inventory adjustments on the supply chain is assessed through key links in the supply chain (such as production cycles and logistics delivery times), identifying the potential pressure that inventory adjustments may cause on other links in the supply chain. A cost-benefit analysis is conducted by comparing indicators such as demand satisfaction, inventory turnover rate, and storage costs before and after inventory adjustments to determine whether the inventory adjustments effectively improved demand satisfaction and reduced storage costs, generating an inventory adjustment report.

[0120] S5.5 combines inventory adjustment reports with historical sales data, market demand forecasts, and inventory adjustment recommendations, and assesses the impact of inventory adjustments on sales trends and market demand fluctuations, generating a trend analysis report.

[0121] It should be noted that the calculation of the difference between inventory levels before and after the adjustment and historical sales volume analyzes whether the inventory adjustment effectively replenished inventory shortages. By comparing inventory and sales data for different time periods (e.g., quarters and months), identify which adjustments have had a positive impact on sales fluctuations and which adjustments may have led to inventory mismatches. For example, identify inventory shortages and surpluses during seasonal sales fluctuations or promotional activities. Compare market demand forecasts with inventory changes in the inventory adjustment report to assess whether the adjusted inventory accurately reflects future market demand trends. By comparing sales fluctuations in historical data and market demand forecasts, identify whether inventory adjustments can effectively respond to future demand changes and prevent inventory buildup due to inaccurate demand forecasts. Through a comprehensive analysis of inventory adjustment recommendations with historical sales data and market demand forecasts, assess the long-term impact of inventory adjustments on sales trends. Identify which adjustments help stabilize sales trends, avoid imbalances between inventory and demand, and ensure the supply chain can flexibly respond to market demand fluctuations, generating a trend analysis report.

[0122] In summary, this invention achieves refined production and inventory management by: collecting and preprocessing multimodal apparel data; utilizing a trend evolution learning model to extract long-term trend changes, providing precise guidance for subsequent trend path generation; combining historical sales data and market demand forecasts to optimize production planning, inventory allocation, and logistics scheduling, thereby improving supply chain efficiency; and adjusting inventory and production based on market feedback data to generate trend analysis reports. The overall solution ensures efficient response to market demand through dynamic forecasting and real-time adjustments, enhancing the production efficiency and market adaptability of the apparel industry.

[0123] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for clothing trend prediction and analysis based on reinforcement learning, characterized in that: The application relates to a trend prediction method and system. Collect multi-modal data and preprocess the data to generate multi-modal clothing data; Evolve the multi-modal clothing data through a trend evolution learning model to obtain a trend evolution strategy and generate a trend development path; Input the trend development path into a trend prediction model to perform trend prediction and style generation and output an innovative style dataset; Optimize production planning, inventory allocation and logistics scheduling according to the innovative style dataset, historical sales data and market demand prediction results to generate an inventory production plan; Collect market reaction feedback data, adjust the inventory and production according to the inventory production plan, obtain inventory adjustment suggestions, identify the gap between the inventory and market demand and generate a trend analysis report. 2.The reinforcement learning based garment trend prediction and analysis method of claim 1, wherein: The specific steps of collecting multi-modal data and preprocessing the data are as follows, Collect visual image data of the clothing by using a high-definition camera and collect friction sound data of the clothing when the clothing is in contact with the human body by using a microphone to generate a comprehensive clothing dataset; Clean and denoise the comprehensive clothing dataset to generate a cleaned comprehensive clothing dataset; Convert the cleaned comprehensive clothing dataset into a standard normal distribution by using a Z-score standardization method to generate a standardized clothing dataset; Extract image features from the visual image data in the standardized clothing dataset and gather all the image features to generate an image feature dataset. 3.The reinforcement learning based garment trend prediction and analysis method of claim 2, wherein: The multi-modal clothing data is constructed into a graph structure according to time sequence and spatial relationship and data fusion is performed by using a graph convolution network. 4.The reinforcement learning based garment trend prediction and analysis method of claim 3, wherein: The specific steps of evolving the multi-modal clothing data through the trend evolution learning model to obtain the trend evolution strategy are as follows, Input the multi-modal clothing data into the trend evolution learning model, learn the long-term dependence relationship of the clothing trend and generate a trained trend evolution learning model; Input the standardized clothing dataset into the trained trend evolution learning model, evolve the time series of the standardized clothing dataset and extract long-term dependence relationship and trend change mode to generate a trend evolution strategy. 5.The reinforcement learning based garment trend prediction and analysis method of claim 1, wherein: The specific steps of generating the trend development path are as follows, Based on the trend evolution strategy, evaluate the trend accuracy and compare the market adaptability of multiple trend paths in the trained trend evolution learning model by using a multi-objective optimization algorithm to generate preliminary trend path data; Combine the preliminary trend path data with historical sales data, evaluate the market adaptability and sales potential of different preliminary trend path data and generate market adaptability scores; Select trend paths with high market adaptability from the preliminary trend path data according to the market adaptability scores to generate screened preliminary trend paths; Adjust the market adaptability of the screened preliminary trend paths to generate a trend development path. 6.The reinforcement learning based garment trend prediction and analysis method of claim 5, wherein: The specific steps of inputting the trend development path into the trend prediction model, performing trend prediction and style generation and outputting the innovative style dataset are as follows, Input the trend development path into the trend prediction model, combine historical trend change data and market data, predict the future trend evolution path by using a time series analysis method and generate trend prediction data; The trend prediction data is combined with market demand and popular elements, and the sales potential and market adaptability of each style are analyzed to generate an innovative style dataset.

7. The reinforcement learning based garment trend prediction and analysis method of claim 6, wherein: According to the innovative style dataset, historical sales data and market demand prediction results are combined to optimize production planning, inventory allocation and logistics scheduling, and the specific steps are as follows, The innovative style dataset is combined with historical sales data and market demand prediction results to predict the production demand, inventory demand and logistics demand of each innovative style. The production plan, inventory allocation and logistics scheduling are extracted from the original inventory production plan. The production demand is combined with the existing production capacity and working hours to optimize the production sequence and resource allocation in the production plan, generating an optimized production plan. The inventory demand and existing inventory situation are combined to optimize the inventory allocation ratio of each interval in the inventory allocation, generating an optimized inventory allocation. The logistics demand, existing inventory and transportation capacity are combined to optimize the distribution route and sequence in the logistics scheduling, generating an optimized logistics scheduling.

8. The reinforcement learning based garment trend prediction and analysis method of claim 7, wherein: The generation of inventory production plan refers to the fusion of optimized production plan, optimized inventory allocation and optimized logistics scheduling, and resource integration. 9.The reinforcement learning based garment trend prediction and analysis method of claim 8, wherein: The market reaction feedback data is collected, combined with the inventory production plan for inventory and production adjustment, and the inventory adjustment suggestions are obtained, and the specific steps are as follows, The market reaction feedback data, sales channel data and consumer research results are collected and data cleaning is performed to generate feedback cleaning data. The feedback cleaning data is combined with the inventory production plan, and the inventory allocation and inventory production plan are adjusted through the particle swarm optimization algorithm to generate inventory production adjustment data. The inventory production adjustment data is combined with market demand and current inventory situation, and the difference analysis method is used to quantitatively analyze the gap between inventory quantity, demand quantity and inventory production plan, identify the inventory shortage area, and obtain inventory adjustment suggestions. 10.The reinforcement learning based garment trend prediction and analysis method of claim 9, wherein: The gap between inventory and market demand is identified, and a trend analysis report is generated, and the specific steps are as follows, The inventory adjustment suggestions are combined with market demand and inventory production plan to evaluate the impact of inventory adjustment on the supply chain, and measure the optimization effect of inventory adjustment on demand satisfaction and cost efficiency to generate an inventory adjustment report. The inventory adjustment report is combined with historical sales data, market demand prediction results and inventory adjustment suggestions, and the impact of inventory adjustment on sales trend and market demand fluctuation is evaluated to generate a trend analysis report.

Images