A method and system for constructing a hosiery color matching database based on production data

Abstract

This invention relates to the field of sock color matching database construction technology, specifically a method and system for constructing a sock color matching database based on production data. The method includes: acquiring first production data and second market data for socks, and fusing them to form first sock fused data; identifying the fused data based on a sock color matching analysis model to obtain inventory feasibility scores, equipment compatibility scores, process stability scores, second color number profit scores, and second market value scores; obtaining a comprehensive production feasibility score and a comprehensive market feasibility score, and then generating a comprehensive color number feasibility score; constructing a multimodal attention dynamic weight network to analyze real-time production data and market data, obtaining updated production feasibility weights and market feasibility weights, and then obtaining the updated comprehensive color number feasibility score, thereby constructing the sock color matching database. This invention can improve production efficiency, market efficiency, and the level of intelligence in sock color matching database construction.

Description

Technical Field

[0001] This invention relates to the field of sock color matching database construction technology, specifically a method and system for constructing a sock color matching database based on production data. Background Technology

[0002] With the continuous development of industrial intelligence, traditional production models are gradually transforming into data-driven and intelligent decision-making. However, in the field of sock manufacturing, traditional sock color matching decisions usually rely on analysis of single-modal data, failing to fully integrate production and market data. This leads to discrepancies between color matching results and actual production needs, making it difficult to meet the diverse requirements of product quality and market demand. Consequently, production efficiency, market efficiency, and the level of intelligence in building a sock color matching database are all limited.

[0003] To address this, a method and system for constructing a sock color matching database based on production data is proposed. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for constructing a sock color matching database based on production data. This method involves acquiring and fusing first production data and second market data for socks to form first fused sock data. Based on a sock color matching analysis model, the fused data is used to identify inventory feasibility scores, equipment compatibility scores, process stability scores, second color number profit scores, and second market value scores. A comprehensive production feasibility score and a comprehensive market feasibility score are then obtained, which in turn generate a comprehensive color number feasibility score. A multimodal attention dynamic weight network is constructed to analyze real-time production and market data, obtaining updated production feasibility weights and market feasibility weights, thereby obtaining an updated comprehensive color number feasibility score and constructing the sock color matching database. This invention can improve production efficiency, market efficiency, and the level of intelligence in constructing the sock color matching database.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A method for constructing a sock color matching database based on production data includes:

[0007] S1. Obtain the first production data and the second market data of socks, and merge the first production data and the second market data to obtain the first sock fused data;

[0008] S2. Construct a sock color matching analysis model to identify the first sock fusion data and obtain inventory feasibility score, equipment compatibility score, process stability score, second color number profit score, and second market value score; based on the inventory feasibility score, equipment compatibility score, and process stability score, obtain a comprehensive production feasibility score; based on the second color number profit score and second market value score, obtain a comprehensive market feasibility score; based on the comprehensive production feasibility score, comprehensive market feasibility score, comprehensive production feasibility weight, and comprehensive market feasibility weight, obtain a comprehensive color number feasibility score;

[0009] S3. Construct a multimodal attention dynamic weight network to analyze real-time production data and real-time market data to obtain updated production feasibility weights and updated market feasibility weights; thereby obtaining an updated comprehensive color code feasibility score.

[0010] S4. Construct a sock color matching database based on the updated comprehensive color code feasibility score, first production data, and second market data.

[0011] Preferably, the first production data includes inventory data for each color yarn, historical color difference pass rate, number of associated equipment and failure rate of associated equipment, and customer order volume; the second market data includes historical sales data, historical return data, historical user feedback data, energy consumption cost per pair and selling price per pair for each color sock.

[0012] Preferably, the sock color matching analysis model includes a first input layer, a preprocessing layer, a sock color matching association feature extraction layer, a sock color matching association feature analysis layer, and a first output layer;

[0013] The first input layer is used to input the first sock fusion data into the sock color matching analysis model;

[0014] The preprocessing layer is used to perform data cleaning and outlier handling on the first sock product fusion data;

[0015] The sock color matching association feature extraction layer is used to extract inventory feasibility score, equipment compatibility score, process stability score, second color number profit score, and second market value score; a comprehensive production feasibility score is obtained based on the inventory feasibility score, equipment compatibility score, and process stability score; a comprehensive market feasibility score is obtained based on the second color number profit score and second market value score; the sock color matching association feature analysis layer is used to obtain a comprehensive color number feasibility score based on the comprehensive production feasibility score and the comprehensive market feasibility score; the first output layer is used to output the comprehensive color number feasibility score.

[0016] Preferably, the multimodal attention dynamic weight network includes a second input layer, a multimodal attention layer, a dynamic fusion layer, and a second output layer;

[0017] The second input layer is used to input real-time production data and real-time market data into the multimodal attention dynamic weight network;

[0018] The multimodal attention layer is used to acquire production attention heads and market attention heads;

[0019] The dynamic fusion layer is used to analyze production attention head and market attention head based on a gating mechanism to obtain updated production feasibility weight and updated market feasibility weight.

[0020] The second output layer is used to output updated production feasibility weights and updated market feasibility weights.

[0021] Preferably, the process of obtaining the updated comprehensive color number feasibility score is as follows: weighting the updated production feasibility weight, the comprehensive production feasibility score, the updated market feasibility weight, and the comprehensive market feasibility score, and then multiplying the weighted value with the adjustment factor.

[0022] A system for building a sock color matching database based on production data includes:

[0023] The data acquisition and fusion module is used to acquire the first production data and the second market data of socks, and to fuse the first production data and the second market data to obtain the first sock fused data;

[0024] The comprehensive color number scoring and analysis module is used to construct a sock color matching analysis model to identify the first set of sock fusion data and obtain inventory feasibility score, equipment compatibility score, process stability score, second color number profit score, and second market value score; based on the inventory feasibility score, equipment compatibility score, and process stability score, a comprehensive production feasibility score is obtained; based on the second color number profit score and second market value score, a comprehensive market feasibility score is obtained; and based on the comprehensive production feasibility score, comprehensive market feasibility score, comprehensive production feasibility weight, and comprehensive market feasibility weight, a comprehensive color number feasibility score is obtained.

[0025] The dynamic weight update module constructs a multimodal attention dynamic weight network to analyze real-time production data and real-time market data, and obtains updated production feasibility weights and updated market feasibility weights; thereby obtaining an updated comprehensive color code feasibility score.

[0026] The dynamic sock color matching database construction module is used to build a sock color matching database based on the updated comprehensive color code feasibility score, primary production data, and secondary market data.

[0027] Preferably, the first production data includes inventory data for each color yarn, historical color difference pass rate, number of associated equipment and failure rate of associated equipment, and customer order volume; the second market data includes historical sales data, historical return data, historical user feedback data, energy consumption cost per pair and selling price per pair for each color sock.

[0028] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0029] 1. This invention obtains first production data and second market data of socks, and merges the first production data and the second market data to obtain first sock fused data; by comprehensively analyzing the production data and market data, it can not only improve the overall guarantee of sock color matching analysis regarding production efficiency and market efficiency, but also help improve the intelligent level of sock color matching database construction.

[0030] 2. This invention identifies the first set of fused sock data by constructing a sock color matching analysis model, obtaining inventory feasibility score, equipment compatibility score, process stability score, second color number profit score, and second market value score; based on the inventory feasibility score, equipment compatibility score, and process stability score, a comprehensive production feasibility score is obtained; based on the second color number profit score and second market value score, a comprehensive market feasibility score is obtained; and based on the comprehensive production feasibility score, comprehensive market feasibility score, comprehensive production feasibility weight, and comprehensive market feasibility weight, a comprehensive color number feasibility score is obtained. This not only enhances the comprehensive assurance of sock color matching analysis regarding production efficiency and market efficiency, but also helps to improve the intelligence level of sock color matching database construction.

[0031] 3. This invention constructs a multimodal attention dynamic weight network to analyze real-time production data and real-time market data, obtaining updated production feasibility weights and updated market feasibility weights; thereby obtaining an updated comprehensive color feasibility score; based on the updated comprehensive color feasibility score, the first production data, and the second market data, a sock color matching database is constructed; through real-time analysis of sock color matching, this not only improves the comprehensive guarantee of sock color matching analysis regarding production efficiency and market efficiency, but also helps to improve the intelligence level of sock color matching database construction. Attached Figure Description

[0032] Figure 1 A flowchart illustrating a method for constructing a sock color matching database based on production data, provided in an embodiment of the present invention;

[0033] Figure 2 This is a schematic diagram of the structure of a sock color matching analysis model provided in an embodiment of the present invention;

[0034] Figure 3This is a schematic diagram of a sock color matching database construction system based on production data, provided as an embodiment of the present invention. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] Example 1

[0037] To effectively improve the intelligence level of sock color matching database construction for sock manufacturer A, a sock color matching database construction method based on production data was applied.

[0038] Reference Figure 1 A flowchart illustrating a method for constructing a sock color matching database based on production data, provided in an embodiment of the present invention, includes:

[0039] S1. Obtain the first production data and the second market data of socks, and merge the first production data and the second market data to obtain the first sock fused data;

[0040] Furthermore, the first production data includes inventory data for each color of yarn, historical color difference pass rate, number of associated equipment and failure rate of associated equipment, and customer order volume; the second market data includes historical sales data, historical return data, historical user feedback data, energy consumption cost per pair, and selling price per pair for each color of socks. The historical user feedback data includes user ratings and user reviews;

[0041] Methods for aligning and integrating user ratings and review data:

[0042] Data structuring: Each user rating and comment data is structured according to the "color code-rating-comment" triple to ensure that each comment data is associated with a specific color code.

[0043] Data grouping and aggregation: Group user ratings and comments according to color codes, and create a corresponding rating set and comment set for each color code.

[0044] Time series alignment: Add timestamps to each rating and comment data, align the data in chronological order, and ensure the time relevance of ratings and comments.

[0045] User ID mapping: The user ID mapping mechanism ensures that the ratings and comments of the same color number by the same user can be matched one by one.

[0046] Multi-dimensional fusion: A multi-dimensional tensor fusion method is adopted to convert user rating data (quantitative data) and user comment data (text data) into a unified vector space, which facilitates subsequent model processing; the fusion formula is:

[0047] ;

[0048] in, For the first Data integration of user feedback for each color shade For the first Structured data of user ratings for each color. For the first Structured data of user reviews for each color shade This is the balance coefficient, with a value range of [0,1].

[0049] Furthermore, the first set of sock product fusion data is as follows:

[0050] ;

[0051] in, First-hand sock product data integration; This represents the first production data; This indicates data from the second market.

[0052] This embodiment acquires first production data and second market data for socks, and merges the first production data and the second market data to obtain first sock fused data. By comprehensively analyzing the production data and market data, it can not only improve the overall guarantee of sock color matching analysis regarding production efficiency and market efficiency, but also help improve the intelligence level of sock color matching database construction.

[0053] S2. Construct a sock color matching analysis model to identify the first sock fusion data and obtain inventory feasibility score, equipment compatibility score, process stability score, second color number profit score, and second market value score; based on the inventory feasibility score, equipment compatibility score, and process stability score, obtain a comprehensive production feasibility score; based on the second color number profit score and second market value score, obtain a comprehensive market feasibility score; based on the comprehensive production feasibility score, comprehensive market feasibility score, comprehensive production feasibility weight, and comprehensive market feasibility weight, obtain a comprehensive color number feasibility score;

[0054] Furthermore, the sock color matching analysis model includes a first input layer, a preprocessing layer, a sock color matching association feature extraction layer, a sock color matching association feature analysis layer, and a first output layer;

[0055] The first input layer is used to input the first sock fusion data into the sock color matching analysis model;

[0056] The preprocessing layer is used to perform data cleaning and outlier handling on the first sock product fusion data;

[0057] Specifically, this includes: 1) The data cleaning process for the first batch of sock product fusion data:

[0058] Data integrity check: Detects missing fields in the merged data, including all dimensions of production and market data.

[0059] Data consistency verification: Ensure that data from different sources are fully aligned in terms of color code identification and time dimension.

[0060] Data redundancy elimination: Remove duplicate records generated during the fusion process.

[0061] Time series synchronization: Ensure that all time series data are aligned on the time axis for easy subsequent analysis.

[0062] 2) Structured outlier handling of the first batch of sock product fusion data:

[0063] Structural inconsistency detection: Identifies structural differences between different data sources, such as mismatched field formats.

[0064] Relationship verification: Verify whether the mapping relationship between color code, production data, and market data is reasonable.

[0065] Anomaly Dimension Handling: Correcting dimensional misalignments that occur during the fusion process, such as some color codes lacking corresponding market data.

[0066] 3) Multi-dimensional outlier detection and correction:

[0067] Multivariate outlier detection: Using Mahalanobis distance to identify outliers in a multidimensional feature space and capture comprehensive anomalies caused by correlations between variables.

[0068] Analysis of covariance: Analyzes the correlation between production indicators and market indicators to identify data points that do not conform to historical correlation patterns.

[0069] Anomaly pattern recognition: Clustering algorithms are used to identify anomaly patterns in fused data, such as production data for certain color numbers that are seriously inconsistent with market feedback.

[0070] 4) Global normalization processing: Normalize the merged overall data.

[0071] Among them, (1) the user rating data cleaning process includes:

[0072] Missing value handling: The missing scores are handled by the mean imputation method, that is, the missing values ​​are filled with the mean of the historical scores of the color number.

[0073] Outlier detection: Box plot method is used to detect outlier scores. Q1 is defined as the 25th percentile of the score data, Q3 is the 75th percentile, and IQR = Q3-Q1. Scores less than (Q1-1.5IQR) or greater than (Q3+1.5IQR) are considered outliers. IQR represents the interquartile range of the score data.

[0074] Outlier handling: The Winsorizing method is used to replace outliers with the closest non-outliers to maintain the data distribution characteristics.

[0075] (2) User comment data cleaning process:

[0076] Null value handling: Remove comment data that has no substantial content.

[0077] Text standardization: The comment text is preprocessed by word segmentation, stop word removal, and stemming.

[0078] Spam filtering: Use rule-based filters to remove irrelevant comments such as advertisements and attacks.

[0079] Sentiment polarity anomaly detection: The sentiment analysis model is used to detect whether the sentiment of the comment is consistent with the corresponding rating. If they are inconsistent, they are marked or adjusted.

[0080] (3) Cleaning of production and inventory data:

[0081] Outlier detection: The Z-score method is used to detect outliers in numerical data such as inventory levels and equipment failure rates. |Z|>3 is considered an anomaly.

[0082] Time series data anomaly detection: using the sliding window method and ARIMA model to detect abnormal fluctuations in time series data.

[0083] Data consistency verification: Check the logical relationships between various data fields, such as ensuring that the inventory quantity is not less than the order quantity.

[0084] The sock color matching association feature extraction layer is used to extract inventory feasibility score, equipment compatibility score, process stability score, second color number profit score, and second market value score; based on the inventory feasibility score, equipment compatibility score, and process stability score, a comprehensive production feasibility score is obtained; based on the second color number profit score and second market value score, a comprehensive market feasibility score is obtained; the sock color matching association feature analysis layer is used to obtain a comprehensive color number feasibility score based on the comprehensive production feasibility score and the comprehensive market feasibility score; the first output layer is used to output the comprehensive color number feasibility score.

[0085] The sock color matching feature extraction layer adopts a neural network structure that combines a multilayer perceptron and an attention mechanism, as detailed below:

[0086] 1) The network structure includes:

[0087] Input layer: Receives preprocessed fused data; First hidden layer: 128 neurons, using the ReLU activation function; Attention layer: Multi-head self-attention mechanism, 8 attention heads; Second hidden layer: 64 neurons, using the ReLU activation function; Output layer: 5 neurons, corresponding to 5 different ratings.

[0088] 2) Feature extraction parameters include: learning rate of 0.001, batch size of 64, optimizer of Adam optimizer, loss function of mean squared error, attention head dimension of 16, and dropout rate of 0.2 to prevent overfitting.

[0089] 3) Feature mapping matrix and forward propagation process:

[0090] The feature mapping matrix is ​​a set of weight matrices that transform input data into different feature spaces. In this embodiment, the feature mapping process involves multiple transformation steps. First, the input data is transformed into a 128-dimensional hidden space through the first weight matrix W_1, capturing the primary features of the data. Then, these features are weighted and recombined through the attention mechanism weight matrix W_attn to highlight the relationships between key features. Next, the attention output is mapped to a 64-dimensional refined feature space through the second weight matrix W_2. Finally, these refined features are transformed into five target scoring metrics through the output weight matrix W_out.

[0091] 4) The forward propagation process describes the complete computational flow of data from input to output. The input data is first multiplied by W_1 and a bias b_1 is added. Then, a non-linear transformation is generated through the ReLU activation function, forming the first hidden layer output H_1. These outputs are then fed into a multi-head attention layer, where each attention head focuses on the importance of different feature dimensions and the results are fused to generate the attention output H_attn. The attention output then undergoes a linear transformation and ReLU activation in the second hidden layer to generate refined features H_2. Finally, H_2 is transformed through the output layer and activated by the Sigmoid activation function, producing five scoring metrics ranging from [0,1].

[0092] This structural design effectively captures the complex nonlinear relationships and inter-feature interactions in sock color matching data, thereby enabling accurate predictions of multi-dimensional scores such as inventory feasibility and equipment compatibility. In particular, the introduction of an attention mechanism allows the model to adaptively focus on the most relevant feature combinations in different scenarios, improving the accuracy and interpretability of the scores.

[0093] Furthermore, the inventory feasibility score is as follows:

[0094] ;

[0095] in, Indicates the first Inventory feasibility score for each yarn color number; Indicates the first Inventory quantity of each yarn color number; Indicates customer order volume; Indicates An exponential function with base 0;

[0096] The device compatibility score is:

[0097] ;

[0098] in, Indicates the first Equipment compatibility rating for each yarn color number; Indicates the first Failure rate of associated equipment for each yarn color number; Indicates the first The number of devices associated with each yarn color number; Indicates the type of yarn color;

[0099] Furthermore, the process stability score is:

[0100] ;

[0101] in, Indicates the first Process stability rating for each yarn color number; Indicates the first Historical color difference pass rate for each yarn color number;

[0102] Furthermore, the comprehensive production feasibility score is as follows:

[0103] ;

[0104] in, Indicates the first Comprehensive production feasibility score for each yarn color number; Indicates the first Inventory feasibility rating coefficient for each yarn color number; Indicates the first Equipment compatibility rating coefficient for each yarn color number; Indicates the first The process stability rating coefficient for each yarn color number;

[0105] Furthermore, the profit rating for the second color is:

[0106] ;

[0107] in, Indicates the first Profit rating for the second color of each yarn color; Indicates the price for odd or even numbers; Indicates the energy consumption cost of single and double production;

[0108] Furthermore, the second market value score is obtained through sales trend score, return trend score, and user feedback score;

[0109] The sales trend score is obtained through exponential smoothing. Specifically, this involves splitting historical sales data chronologically to obtain first historical sales data and second historical sales data, with the first historical sales data occurring before the second historical sales data. Then, exponential smoothing is used to calculate smoothed sales volume sequences for both the first and second historical sales data. Finally, the smoothed sales volume of the last day is taken from each smoothed sales volume sequence to obtain the first smoothed sales volume and the second smoothed sales volume. The sales trend score is then derived from the first smoothed sales volume and the second smoothed sales volume.

[0110] The sales trend score is:

[0111] ;

[0112] in, Indicates the first Sales trend rating for each yarn color number; This represents the first smoothed sales volume; This indicates the second smoothed sales volume;

[0113] The return trend score is obtained through exponential smoothing. Specifically, this involves splitting historical return data chronologically to obtain first historical return data and second historical return data, where the first historical return data occurs before the second historical return data. Then, exponential smoothing is used to calculate the smoothed return volume sequence for both the first and second historical return data. Based on each smoothed return volume sequence, the smoothed return volume for the last day is taken to obtain the first smoothed return volume and the second smoothed return volume. Finally, the return trend score is obtained based on the first and second smoothed return volumes. The return trend score is as follows:

[0114] ;

[0115] in, Indicates the first Return trend rating for each yarn color number; Indicates the second smooth return quantity; Indicates the first smooth return quantity;

[0116] Furthermore, user rating coefficients are obtained by averaging and normalizing user ratings; user comments are analyzed using NLP to obtain user comment coefficients, which reflect user satisfaction; and user feedback ratings are obtained based on user rating coefficients and user comment coefficients.

[0117] In this embodiment, the user comments use the SnowNLP library in NLP as a specific technical means. By calling its sentiment analysis function, a sentiment quantification value in the range of [0, 1] is generated for each comment.

[0118] The user feedback rating is:

[0119] ;

[0120] in, Indicates the first User feedback ratings for each yarn color number; This represents the influencing factor of user rating coefficients; This represents the user rating coefficient; This represents the influence factor of user review coefficient; Indicates the user review coefficient;

[0121] Furthermore, the second market value score is:

[0122] ;

[0123] in, Indicates the first Second market value rating for each yarn color number; Indicates the first Influencing factors for sales trend scores of individual yarn color numbers; Indicates the first Return trend rating influencing factors for individual yarn color numbers; Indicates the first Factors influencing user feedback ratings for each yarn color;

[0124] Furthermore, the comprehensive market feasibility score is as follows:

[0125] ;

[0126] in, Indicates the first A comprehensive market feasibility score for each yarn color number; This indicates the profit rating weight for the second color number; This indicates the weighting of the second market value score;

[0127] Furthermore, the feasibility score for the comprehensive color code is as follows:

[0128] ;

[0129] in, Indicates the first A comprehensive feasibility score for each yarn color number; Indicates factors affecting production feasibility; Indicates market feasibility weight;

[0130] This embodiment identifies the first set of fused sock data by constructing a sock color matching analysis model, obtaining inventory feasibility score, equipment compatibility score, process stability score, second color number profit score, and second market value score; based on the inventory feasibility score, equipment compatibility score, and process stability score, a comprehensive production feasibility score is obtained; based on the second color number profit score and second market value score, a comprehensive market feasibility score is obtained; and based on the comprehensive production feasibility score, comprehensive market feasibility score, comprehensive production feasibility weight, and comprehensive market feasibility weight, a comprehensive color number feasibility score is obtained. This not only enhances the comprehensive assurance of sock color matching analysis regarding production efficiency and market efficiency, but also helps to improve the intelligence level of sock color matching database construction.

[0131] S3. Construct a multimodal attention dynamic weight network to analyze real-time production data and real-time market data to obtain updated production feasibility weights and updated market feasibility weights; thereby obtaining an updated comprehensive color code feasibility score.

[0132] Furthermore, the multimodal attention dynamic weight network includes a second input layer, a multimodal attention layer, a dynamic fusion layer, and a second output layer;

[0133] The second input layer is used to input real-time production data and real-time market data into the multimodal attention dynamic weight network;

[0134] The multimodal attention layer is used to acquire production attention heads and market attention heads;

[0135] The dynamic fusion layer is used to analyze production attention head and market attention head based on a gating mechanism to obtain updated production feasibility weight and updated market feasibility weight.

[0136] The second output layer is used to output updated production feasibility weights and updated market feasibility weights.

[0137] Furthermore, the multimodal attention layer obtains real-time inventory feasibility scores, real-time equipment compatibility scores, real-time process stability scores, color number profit scores, sales trend scores, return trend scores, and user feedback scores based on real-time production data and real-time market data; and constructs production vectors and market vectors; and obtains production attention heads and market attention heads based on production vectors and market vectors.

[0138] The production vector is:

[0139] ;

[0140] in, Represents the production vector; This indicates a real-time inventory feasibility score. This indicates a real-time device compatibility score. This indicates the real-time process stability score;

[0141] The market vector is:

[0142] ;

[0143] in, Represents the market vector; This indicates the real-time color code profit rating; This indicates a real-time sales trend score; Indicates a real-time return trend score; This indicates real-time user feedback ratings;

[0144] The production focus is:

[0145] ;

[0146] in, Indicates the production focus; Indicates the activation function; , , Represents the query, key, and value matrix of the production vector; Indicates the vector dimension;

[0147] The market focus is:

[0148] ;

[0149] in, This indicates market attention; Indicates the activation function; , , Represents the query, key, and value matrix of the market vector; Indicates the vector dimension;

[0150] Furthermore, the dynamic fusion layer obtains updated production feasibility weights and updated market feasibility weights through a gating mechanism;

[0151] The weight for the feasibility of updating production is:

[0152] ;

[0153] in, This indicates an update to the production feasibility weight; Indicates the gating factor; Indicates the production focus; This indicates market attention;

[0154] The gating factor is:

[0155] ;

[0156] in, Indicates the gating factor; Represents the weight matrix, used for weighting... and Perform a linear transformation; Indicates the bias term;

[0157] The updated market feasibility weight is:

[0158] ;

[0159] This embodiment improves the real-time performance of comprehensive color feasibility scoring by analyzing production and market data in real time. Real-time analysis of sock color matching not only enhances the comprehensive assurance of sock color matching analysis regarding production and market efficiency, but also improves the intelligence level of sock color matching database construction.

[0160] Furthermore, the process of obtaining the updated comprehensive color code feasibility score is as follows: The updated production feasibility weight, comprehensive production feasibility score, updated market feasibility weight, and comprehensive market feasibility score are weighted, and the weighted value is multiplied by an adjustment factor to obtain the updated comprehensive color code feasibility score.

[0161] ;

[0162] in, This indicates an update to the overall color designation feasibility score; This indicates an update to the production feasibility weight; This indicates an update to the market feasibility weight; Indicates the first Comprehensive production feasibility score for each color number; Indicates the first A comprehensive market feasibility score for each color number; Indicates the regulating factor;

[0163] S4. Construct a sock color matching database based on the updated comprehensive color code feasibility score, first production data, and second market data;

[0164] Furthermore, the sock color matching database includes fields for comprehensive production feasibility score, comprehensive market feasibility score, comprehensive color number feasibility score, production feasibility weight, market feasibility weight, historical production data, historical market data, real-time production data, and real-time market data.

[0165] This embodiment constructs a multimodal attention dynamic weight network to analyze real-time production data and real-time market data, obtaining updated production feasibility weights and updated market feasibility weights; thereby obtaining an updated comprehensive color feasibility score; based on the updated comprehensive color feasibility score, the first production data and the second market data, a sock color matching database is constructed; through real-time analysis of sock color matching, not only can the comprehensive guarantee of sock color matching analysis regarding production efficiency and market efficiency be improved, but it is also conducive to improving the intelligent level of sock color matching database construction.

[0166] This implementation acquires and merges primary and secondary market data for socks to form primary sock fusion data. Based on a sock color matching analysis model, the fusion data is used to identify inventory feasibility scores, equipment compatibility scores, process stability scores, secondary color number profit scores, and secondary market value scores. A comprehensive production feasibility score and a comprehensive market feasibility score are obtained, which in turn generate a comprehensive color number feasibility score. A multimodal attention dynamic weight network is constructed to analyze real-time production and market data, resulting in updated production feasibility weights and market feasibility weights, leading to an updated comprehensive color number feasibility score, and ultimately, the construction of a sock color matching database. This invention can improve production efficiency, market efficiency, and the intelligence level of sock color matching database construction.

[0167] To verify the effectiveness of the sock color matching database construction method based on production data provided in this embodiment, different methods were applied to the color matching process of sock manufacturer A. The product qualification rate and market revenue of sock manufacturer A under different methods were compared and analyzed with the case of no method application. The specific results are shown in Table 1.

[0168] Table 1. Comparison of Product Qualification Rate and Market Revenue of Sock Manufacturer A under Different Methods

[0169]

[0170] Method 1 is a method for constructing a sock color matching database based on production data provided in this embodiment; Method 2 is based on Method 1 without considering production data analysis; Method 3 is based on Method 1 without considering market data analysis.

[0171] As shown in Table 1, the method for constructing a sock color matching database based on production data provided in this embodiment has a certain degree of effectiveness. It can ensure both production quality and market revenue, comprehensively guaranteeing the production efficiency and market efficiency of sock manufacturer A, and thus has a certain degree of effectiveness.

[0172] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for constructing a sock color matching database based on production data, characterized in that, include: S1. Obtain first production data and second market data for socks, and merge the first production data and second market data to obtain first sock fused data; the first production data includes inventory data of yarn for each color number, historical color difference pass rate, number of associated equipment and failure rate of associated equipment, and customer order volume; the second market data includes historical sales data, historical return data, historical user feedback data, energy consumption cost per pair and selling price per pair for each color number of socks; S2. Construct a sock color matching analysis model to identify the first sock fusion data and obtain inventory feasibility score, equipment compatibility score, process stability score, second color profit score, and second market value score; A comprehensive production feasibility score is obtained based on inventory feasibility score, equipment compatibility score, and process stability score; a comprehensive market feasibility score is obtained based on second color number profit score and second market value score; a comprehensive color number feasibility score is obtained based on comprehensive production feasibility score, comprehensive market feasibility score, comprehensive production feasibility weight, and comprehensive market feasibility weight; the sock color matching analysis model includes a first input layer, a preprocessing layer, a sock color matching association feature extraction layer, a sock color matching association feature analysis layer, and a first output layer; The first input layer is used to input the first sock fusion data into the sock color matching analysis model; The preprocessing layer is used to perform data cleaning and outlier handling on the first sock product fusion data; The sock color matching association feature extraction layer is used to extract inventory feasibility score, equipment compatibility score, process stability score, second color number profit score, and second market value score. A comprehensive production feasibility score is obtained based on inventory feasibility score, equipment compatibility score, and process stability score; a comprehensive market feasibility score is obtained based on second color number profit score and second market value score; the sock color matching correlation feature analysis layer is used to obtain a comprehensive color number feasibility score based on the comprehensive production feasibility score and comprehensive market feasibility score; the first output layer is used to output the comprehensive color number feasibility score. S3. Construct a multimodal attention dynamic weight network to analyze real-time production data and real-time market data to obtain updated production feasibility weights and updated market feasibility weights; thereby obtaining an updated comprehensive color code feasibility score. The multimodal attention dynamic weight network includes a second input layer, a multimodal attention layer, a dynamic fusion layer, and a second output layer; The second input layer is used to input real-time production data and real-time market data into the multimodal attention dynamic weight network; The multimodal attention layer is used to acquire production attention heads and market attention heads; The dynamic fusion layer is used to analyze production attention head and market attention head based on a gating mechanism to obtain updated production feasibility weight and updated market feasibility weight. The second output layer is used to output updated production feasibility weights and updated market feasibility weights; S4. Construct a sock color matching database based on the updated comprehensive color code feasibility score, first production data, and second market data.

2. The method for constructing a sock color matching database based on production data according to claim 1, characterized in that: The process of obtaining the feasibility score for updating the comprehensive color code is as follows: weighting the updated production feasibility weight, the comprehensive production feasibility score, the updated market feasibility weight, and the comprehensive market feasibility score, and then multiplying the weighted value with the adjustment factor.

3. A system for constructing a sock color matching database based on production data, the system being used to execute a method for constructing a sock color matching database based on production data as described in any one of claims 1 to 2, characterized in that, include: The data acquisition and fusion module is used to acquire the first production data and the second market data of socks, and to fuse the first production data and the second market data to obtain the first sock fused data; The comprehensive color number scoring and analysis module is used to build a sock color matching analysis model to identify the first sock product fusion data and obtain inventory feasibility score, equipment compatibility score, process stability score, second color number profit score, and second market value score. A comprehensive production feasibility score is obtained based on inventory feasibility score, equipment compatibility score, and process stability score; a comprehensive market feasibility score is obtained based on second color number profit score and second market value score; and a comprehensive color number feasibility score is obtained based on the comprehensive production feasibility score, comprehensive market feasibility score, comprehensive production feasibility weight, and comprehensive market feasibility weight. The dynamic weight update module constructs a multimodal attention dynamic weight network to analyze real-time production data and real-time market data, and obtains updated production feasibility weights and updated market feasibility weights; thereby obtaining an updated comprehensive color code feasibility score. The dynamic sock color matching database construction module is used to build a sock color matching database based on the updated comprehensive color code feasibility score, primary production data, and secondary market data.

4. The sock color matching database construction system based on production data according to claim 3, characterized in that, The first production data includes inventory data for each color of yarn, historical color difference pass rate, number of associated equipment and failure rate of associated equipment, and customer order volume; the second market data includes historical sales data, historical return data, historical user feedback data, energy consumption cost per pair and selling price per pair for each color of socks.

Images