CN116993393A - Clothing production order management system and method thereof - Google Patents
Clothing production order management system and method thereof Download PDFInfo
- Publication number
- CN116993393A CN116993393A CN202310836182.5A CN202310836182A CN116993393A CN 116993393 A CN116993393 A CN 116993393A CN 202310836182 A CN202310836182 A CN 202310836182A CN 116993393 A CN116993393 A CN 116993393A
- Authority
- CN
- China
- Prior art keywords
- data
- historical sales
- vectors
- sales data
- training
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000004519 manufacturing process Methods 0.000 title claims abstract description 69
- 238000000034 method Methods 0.000 title claims description 32
- 238000007726 management method Methods 0.000 claims abstract description 54
- 239000013598 vector Substances 0.000 claims description 331
- 239000011159 matrix material Substances 0.000 claims description 54
- 230000006870 function Effects 0.000 claims description 36
- 230000004913 activation Effects 0.000 claims description 15
- 238000004140 cleaning Methods 0.000 claims description 15
- 239000004744 fabric Substances 0.000 claims description 14
- 239000000463 material Substances 0.000 claims description 12
- 230000009471 action Effects 0.000 claims description 10
- 239000003086 colorant Substances 0.000 claims description 8
- 230000002787 reinforcement Effects 0.000 claims description 8
- 238000011176 pooling Methods 0.000 claims description 7
- 230000008569 process Effects 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 2
- 238000004458 analytical method Methods 0.000 abstract description 17
- 230000008901 benefit Effects 0.000 abstract description 6
- 238000010586 diagram Methods 0.000 description 15
- 238000009826 distribution Methods 0.000 description 8
- 238000003062 neural network model Methods 0.000 description 7
- 230000003213 activating effect Effects 0.000 description 4
- 230000008859 change Effects 0.000 description 4
- 238000012821 model calculation Methods 0.000 description 4
- 230000003252 repetitive effect Effects 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 230000003042 antagnostic effect Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000013480 data collection Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 230000035945 sensitivity Effects 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- 238000003860 storage Methods 0.000 description 2
- -1 styles Substances 0.000 description 2
- 238000009472 formulation Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000005728 strengthening Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0202—Market predictions or forecasting for commercial activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
- G06N3/0455—Auto-encoder networks; Encoder-decoder networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/08—Logistics, e.g. warehousing, loading or distribution; Inventory or stock management
- G06Q10/087—Inventory or stock management, e.g. order filling, procurement or balancing against orders
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/04—Manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- Finance (AREA)
- Development Economics (AREA)
- Accounting & Taxation (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- Marketing (AREA)
- Mathematical Physics (AREA)
- Molecular Biology (AREA)
- Entrepreneurship & Innovation (AREA)
- General Business, Economics & Management (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Tourism & Hospitality (AREA)
- Human Resources & Organizations (AREA)
- Game Theory and Decision Science (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Manufacturing & Machinery (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application relates to the technical field of clothing production, and more particularly discloses a clothing production management system and a clothing production management method, which are used for predicting sales in future seasons by carrying out semantic modeling and predictive analysis on historical sales data so as to facilitate the establishment of reasonable inventory plans and purchasing plans. According to the scheme, through accurate sales prediction, the inventory can be better managed, excessive stock backlog is avoided or the situation that market demands cannot be met is avoided, and the stability and economic benefit of a supply chain are improved.
Description
Technical Field
The application relates to the technical field of clothing production, in particular to a clothing production management system and a clothing production management method.
Background
Due to the long production cycle of the garment, once the market demand is misjudged, stock backlog or energy production stagnation can be caused, and the supply chain stability is affected. If too many garments are produced, but sales are not smooth, stock backlog will be caused, and huge cost and fund pressure are brought to enterprises; too little inventory may result in production interruptions and customer orders that are not delivered in time.
In the production management of clothing, historical sales data is very important information, and consumer buying habits and favorites can be known through analysis of the historical sales data, and future sales and demand trends are predicted, so that inventory is managed.
Accordingly, a production management system and method for garments is desired.
Disclosure of Invention
The present application has been made to solve the above-mentioned technical problems. Embodiments of the present application provide a production management system of clothing and a method thereof, which predicts sales in future seasons by performing semantic modeling and predictive analysis on historical sales data so as to make reasonable inventory plans and purchase plans. According to the scheme, through accurate sales prediction, the inventory can be better managed, excessive stock backlog is avoided or the situation that market demands cannot be met is avoided, and the stability and economic benefit of a supply chain are improved.
Accordingly, according to one aspect of the present application, there is provided a production management system for apparel, comprising:
the historical data acquisition module is used for acquiring historical sales data, wherein the historical sales data comprises colors, styles and fabric materials;
the data cleaning module is used for carrying out data cleaning on the historical sales data to obtain a plurality of data items, wherein the plurality of data items comprise data attributes and data values;
the attribute embedding coding module is used for respectively passing the data attribute of each data item in the plurality of data items through the word embedding layer to obtain a plurality of data item attribute word embedding vectors;
The data adding module is used for respectively adding the data value of each data item in the plurality of data items to the tail of each data item attribute word embedding vector so as to obtain a plurality of data item embedding vectors;
a historical sales data semantic understanding module for inputting the plurality of data item embedded vectors into a context encoder based on a converter to obtain a historical sales data semantic feature vector;
the space enhancement module is used for enabling the historical sales data semantic feature vectors to pass through the space attention module so as to obtain enhanced historical sales data semantic feature vectors; and
and the sales volume prediction module is used for carrying out regression decoding on the enhanced historical sales data semantic feature vector through a decoder to obtain a decoding value, wherein the decoding value is used for representing the quaternary sales volume predicted value.
In the production management system of the clothing, the historical sales data semantic understanding module comprises: a context semantic coding unit for inputting the plurality of data item embedded vectors into the converter-based context encoder to obtain a plurality of historical sales data feature vectors; and the cascading unit is used for cascading the plurality of historical sales data feature vectors to obtain the historical sales data semantic feature vectors.
In the production management system of the clothing, the context semantic coding unit comprises: the conversion subunit is used for arranging the plurality of data item embedding vectors into input vectors and respectively converting the input vectors into query vectors and key vectors through a learnable embedding matrix; a self-attention subunit, configured to calculate a product between the query vector and a transpose vector of the key vector to obtain a self-attention correlation matrix; a standardized self-attention subunit, configured to perform a standardization process on the self-attention association matrix to obtain a standardized self-attention association matrix; an attention calculating subunit, configured to activate the normalized self-attention association matrix input Softmax activation function to obtain a self-attention feature matrix; and an attention applying subunit, configured to multiply the self-attention feature matrix with each of the plurality of data item embedding vectors to obtain the plurality of historical sales data feature vectors.
In the above-mentioned production management system of clothing, the space enhancement module includes: the depth convolution coding unit is used for performing depth convolution coding on the historical sales data semantic feature vector by using a convolution coding part of the spatial attention module so as to obtain an initial convolution feature map; a spatial attention unit for inputting the initial convolution feature map into a spatial attention portion of the spatial attention module to obtain a spatial attention map; an activation unit, configured to activate the spatial attention map through a Softmax activation function to obtain a spatial attention profile; a calculation unit, configured to calculate a point-by-point multiplication of the spatial attention feature map and the initial convolution feature map to obtain the historical sales feature map; and the pooling unit is used for carrying out the mean pooling processing of each feature matrix along the channel dimension on the historical sales feature graph so as to obtain the semantic feature vector of the enhanced historical sales data.
In the production management system of the clothing, the sales volume prediction module is configured to: performing a decoding regression on the enhanced historical sales data semantic feature vector using the decoder in a decoding formula to obtain the decoded value; wherein, the decoding formula is:wherein X is the enhanced historical sales data semantic feature vector, Y is the decoded value, W is a weight matrix,>representing matrix multiplication.
In the above-described garment production management system, further comprising a training module for training the word embedding layer, the converter-based context encoder, the spatial attention module, and the decoder; wherein, training module includes: the training history data acquisition unit is used for acquiring training history sales data; the training data cleaning unit is used for cleaning the training historical sales data to obtain a plurality of training data items; the training attribute embedding coding unit is used for respectively passing the data attribute of each training data item in the plurality of training data items through the word embedding layer to obtain a plurality of training data item attribute word embedding vectors; the training data adding unit is used for respectively adding the data value of each training data item in the plurality of training data items to the tail of each training data item attribute word embedding vector so as to obtain a plurality of training data item embedding vectors; a training history sales data semantic understanding unit for inputting the plurality of training data item embedded vectors into the converter-based context encoder to obtain training history sales data semantic feature vectors; the training space enhancement unit is used for enabling the training historical sales data semantic feature vectors to pass through the space attention module so as to obtain training enhancement historical sales data semantic feature vectors; the feature vector iteration unit is used for carrying out random reinforcement on the feature expression based on disturbance on the training enhancement historical sales data semantic feature vector so as to obtain an iteration enhancement historical sales data semantic feature vector; the decoding loss function obtaining unit is used for enabling the iteration enhanced historical sales data semantic feature vector to pass through the decoder so as to obtain a decoding loss function value; and a model training unit for training the word embedding layer, the converter-based context encoder, the spatial attention module and the decoder based on the decoding loss function value and traveling through a direction of gradient descent.
In the above-mentioned clothing production management system, the feature vector iteration unit includes: a random disturbance value generation subunit, configured to define a random disturbance function, and generate a predetermined number of random disturbance values through the random disturbance function, where the predetermined number is the same as the scale of the semantic feature vector of the training enhancement historical sales data; a random disturbance value arrangement subunit configured to arrange the predetermined number of random disturbance values into a random disturbance input vector; an activation subunit, configured to activate the random disturbance input vector through a Softmax function to obtain a normalized random disturbance input vector, where a sum of eigenvalues of all positions in the normalized random disturbance input vector is 1; the random disturbance applying subunit is used for calculating the point-by-point multiplication between the normalized random disturbance input vector and the training enhancement historical sales data semantic feature vector to obtain a random action enhancement historical sales data semantic feature vector; and the cascading subunit is used for cascading the random action enhancement historical sales data semantic feature vector and the training enhancement historical sales data semantic feature vector to obtain an iteration enhancement historical sales data semantic feature vector.
According to another aspect of the present application, there is provided a production management method of clothing, comprising:
acquiring historical sales data, wherein the historical sales data comprises colors, styles and fabric materials;
performing data cleaning on the historical sales data to obtain a plurality of data items, wherein the plurality of data items comprise data attributes and data values;
respectively passing the data attribute of each data item in the plurality of data items through a word embedding layer to obtain a plurality of data item attribute word embedding vectors;
respectively adding the data value of each data item in the plurality of data items to the tail of each data item attribute word embedding vector to obtain a plurality of data item embedding vectors;
inputting the plurality of data item embedded vectors into a converter-based context encoder to obtain historical sales data semantic feature vectors;
passing the historical sales data semantic feature vector through a spatial attention module to obtain an enhanced historical sales data semantic feature vector; and
and carrying out regression decoding on the semantic feature vector of the enhanced historical sales data through a decoder to obtain a decoding value, wherein the decoding value is used for representing the predicted value of the quaternary sales.
In the above-described method for production management of garments, inputting the plurality of data item embedded vectors into a converter-based context encoder to obtain historical sales data semantic feature vectors, comprising: inputting the plurality of data item embedded vectors into the converter-based context encoder to obtain a plurality of historical sales data feature vectors; and cascading the plurality of historical sales data feature vectors to obtain the historical sales data semantic feature vector.
In the above-described method for managing clothing production, inputting the plurality of data item embedded vectors into the converter-based context encoder to obtain a plurality of historical sales data feature vectors, comprising: arranging the plurality of data item embedding vectors into input vectors, and respectively converting the input vectors into query vectors and key vectors through a learning embedding matrix; calculating the product between the query vector and the transpose vector of the key vector to obtain a self-attention correlation matrix; carrying out standardization processing on the self-attention association matrix to obtain a standardized self-attention association matrix; inputting the standardized self-attention association matrix into a Softmax activation function to activate so as to obtain a self-attention feature matrix; and multiplying the self-attention feature matrix with each data item embedding vector in the plurality of data item embedding vectors to obtain the plurality of historical sales data feature vectors.
Compared with the prior art, the clothing production management system and the clothing production management method provided by the application predict sales volume in future seasons by carrying out semantic modeling and predictive analysis on historical sales data so as to facilitate the establishment of reasonable inventory plans and purchasing plans. According to the scheme, through accurate sales prediction, the inventory can be better managed, excessive stock backlog is avoided or the situation that market demands cannot be met is avoided, and the stability and economic benefit of a supply chain are improved.
Drawings
The above and other objects, features and advantages of the present application will become more apparent by describing embodiments of the present application in more detail with reference to the attached drawings. The accompanying drawings are included to provide a further understanding of embodiments of the application and are incorporated in and constitute a part of this specification, illustrate the application and together with the embodiments of the application, and not constitute a limitation to the application. In the drawings, like reference numerals generally refer to like parts or steps.
Fig. 1 is a block diagram of a production management system for garments according to an embodiment of the application.
Fig. 2 is a schematic architecture diagram of a production management system for garments according to an embodiment of the application.
FIG. 3 is a block diagram of a historical sales data semantic understanding module in a production management system for apparel in accordance with an embodiment of the present application.
Fig. 4 is a block diagram of a context semantic coding unit in a production management system of a garment according to an embodiment of the present application.
Fig. 5 is a block diagram of a space enhancement module in a garment production management system according to an embodiment of the application.
Fig. 6 is a block diagram of a training module in a garment production management system according to an embodiment of the application.
Fig. 7 is a flowchart of a method of production management of garments according to an embodiment of the application.
Detailed Description
Hereinafter, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. It should be apparent that the described embodiments are only some embodiments of the present application and not all embodiments of the present application, and it should be understood that the present application is not limited by the example embodiments described herein.
Summary of the application
In view of the above problems, the technical idea of the present application is to predict sales in future seasons by performing semantic modeling and predictive analysis on historical sales data so as to make reasonable inventory plans and purchase plans. According to the scheme, through accurate sales prediction, the inventory can be better managed, excessive stock backlog is avoided or the situation that market demands cannot be met is avoided, and the stability and economic benefit of a supply chain are improved.
The method comprises the steps of obtaining historical sales data, wherein the historical sales data comprise information such as colors, styles, fabric materials and the like, and the premise is that clothing sales prediction is carried out. These data items are all important factors considered by consumers in purchasing, can reflect the preference and purchasing habit of the consumers for clothing products, and are helpful for analyzing and predicting future market demands and trends. Therefore, in the present scheme, first, the history sales data that acquires information including color, style, fabric material, and the like is acquired.
Next, to improve data quality, facilitate data processing and analysis, historical sales data is data cleaned to sort the data into a format that is convenient for computer processing and analysis. It is contemplated that there may be some non-canonical, repetitive or erroneous data in the historical sales data that can seriously affect the accuracy of the subsequent predictive model. By data cleansing the historical sales data, such invalid data can be removed, thereby improving data quality. The historical sales data after data cleaning can be classified according to attributes and values, and relevant statistical analysis is performed to find out the relation and characteristics among the data, so that a basis is provided for predictive modeling.
In order to convert the data attribute information into a vector form which can be processed by a computer, when the historical sales data is subjected to predictive analysis, the data attributes of all data items in the historical sales data are respectively passed through a word embedding layer to obtain a plurality of data item attribute word embedding vectors. For data attributes in historical sales data, which are typically discrete variables, they cannot be directly input into the neural network model. By using word embedding techniques, discrete variables can be converted into a continuous vector representation, thereby facilitating processing and computation in the neural network. Because a large number of data attributes exist in the historical sales data, the historical sales data can be compressed into a low-dimensional vector space through a word embedding technology, and the model calculation complexity and the storage space are reduced. In particular, word embedding techniques are able to map similar words to a similar vector space and preserve the semantic similarity of the words. In the historical sales data, some data attributes may have similar or related meanings, and through word embedding, the related attributes can be mapped to a similar vector space, so that the learning capacity of a prediction model is improved.
Then, in order to fuse the data of different characteristics into one vector representation to improve the accuracy of the prediction model, the data value of each data item in the historical sales data is respectively added to the tail end of each data item attribute word embedding vector. There may be an association or interaction between individual data items in the historical sales data, and by fusing data values with the attribute word embedding vectors, the associated information of these data may be better utilized. And, a plurality of data items in the historical sales data are fused into a vector space, so that the data characteristics can be conveniently expressed into a global vector form, and the subsequent model can be conveniently used.
It is contemplated that there may be complex semantic associations between individual items of data in the historical sales data, such as relationships between different attributes of color, style, and fabric. By means of the context encoder based on the converter, semantic associations between different data items can be taken into account and converted into a global semantic feature vector. Moreover, the data items and attributes in the historical sales data can be quite large, so that the data volume is huge, and by using a context encoder based on a converter, the semantic information of the historical sales data can be compressed into a vector space with lower dimensionality, so that the data volume and the model calculation complexity are reduced.
In order to better capture the relation and importance among all data items in the historical sales data, a spatial attention module is adopted to carry out weighted fusion on semantic feature vectors of the historical sales data, so that the expression of key features is enhanced. The data items and the attributes in the historical sales data are quite many, some features are more important to predictive analysis, and the semantic feature vectors of the historical sales data can be weighted and fused through the spatial attention module, so that the influence of the important features is highlighted, and the accuracy of subsequent predictive analysis is improved.
And finally, carrying out regression decoding on the semantic feature vector of the enhanced historical sales data through a decoder so as to convert semantic information of the historical sales data into a quaternary sales volume predicted value and finish the sales volume predicted task. In the previous processing process, the historical sales data is subjected to multiple embedding and encoding operations, finally, the semantic feature vector of the enhanced historical sales data is obtained, regression decoding is carried out through a decoder, and the processing processes can be restored to obtain the quaternary sales quantity predicted value based on the original data. By decoding the semantic feature vector of the enhanced historical sales data into the predicted value of the quaternary sales volume, the future sales volume can be predicted, and important references are provided for the aspects of enterprise production, purchasing, inventory and the like.
In particular, in the solution of the present application, it is considered that there is a slight variation or disturbance of the input data or model at the time of actual inference. These minor variations or disturbances may be due to data noise, data imperfections, variations in data distribution, etc., and in particular, the input data may contain noise, such as data acquisition errors, which may have an impact on the training and inference of the model. The input data may lack certain features or attributes, or have missing values. Such imperfections can lead to uncertainty in the model in processing the data. Models are usually modeled based on a certain data distribution during training, but the data distribution in practical application may change, and the change of the distribution also causes disturbance of the model. Thus, if some tiny random perturbations can be added to the enhanced historical sales data semantic feature vector, so that the deep neural network model and classifier can adapt to these perturbations, it is clear that the resistance to the challenge sample can be improved, thereby improving the robustness of the model.
Based on the above, in the technical scheme of the present application, performing disturbance-based feature expression random reinforcement on the training enhancement historical sales data semantic feature vector to obtain an iteration enhancement historical sales data semantic feature vector, including: defining a random disturbance function; generating a predetermined number of random disturbance values through the random disturbance function, wherein the predetermined number is the same as the scale of the training enhanced historical sales data semantic feature vector; arranging the predetermined number of random disturbance values into a random disturbance input vector; activating the random disturbance input vector through a Softmax function to obtain a normalized random disturbance input vector, wherein the sum of characteristic values of all positions in the normalized random disturbance input vector is 1; calculating the position-based point multiplication between the normalized random disturbance input vector and the training enhancement historical sales data semantic feature vector to obtain a random action enhancement historical sales data semantic feature vector; and cascading the random action enhanced historical sales data semantic feature vector and the training enhanced historical sales data semantic feature vector to obtain an iterative enhanced historical sales data semantic feature vector.
In each iteration cycle of the training phase, disturbance-based feature expression random reinforcement is carried out on the enhanced historical sales data semantic feature vector so as to add some tiny random disturbance in the enhanced historical sales data semantic feature vector in each iteration cycle, so that a deep neural network model and a decoder can adapt to the disturbance, and the resistance to an antagonistic sample is improved. By the method, complexity of the model can be reduced, overfitting is avoided, interpretation of the model is improved, resistance of the model to noise and abnormal values can be enhanced, robustness of the model is improved, convergence speed and convergence of the model can be improved, optimization efficiency of the model is improved, sensitivity and sparsity of the model can be adjusted, and feature selection capability of the model is improved.
Having described the basic principles of the present application, various non-limiting embodiments of the present application will now be described in detail with reference to the accompanying drawings.
Exemplary System
Fig. 1 is a block diagram of a production management system for garments according to an embodiment of the application. As shown in fig. 1, a production management system 100 of clothing according to an embodiment of the present application includes: a historical data collection module 110, configured to obtain historical sales data, where the historical sales data includes color, style, and fabric material; a data cleansing module 120, configured to perform data cleansing on the historical sales data to obtain a plurality of data items, where the plurality of data items include data attributes and data values; the attribute embedding encoding module 130 is configured to pass the data attributes of each of the plurality of data items through the word embedding layer to obtain a plurality of data item attribute word embedding vectors; a data adding module 140, configured to add data values of each data item in the plurality of data items to ends of the attribute word embedding vectors of each data item, so as to obtain a plurality of data item embedding vectors; a historical sales data semantic understanding module 150 for inputting the plurality of data item embedded vectors into a converter-based context encoder to obtain historical sales data semantic feature vectors; a space enhancement module 160, configured to pass the historical sales data semantic feature vector through a space attention module to obtain an enhanced historical sales data semantic feature vector; and a sales volume prediction module 170, configured to perform regression decoding on the enhanced historical sales data semantic feature vector through a decoder to obtain a decoded value, where the decoded value is used to represent a quaternary sales volume predicted value.
Fig. 2 is a schematic architecture diagram of a production management system for garments according to an embodiment of the application. As shown in fig. 2, first, historical sales data including color, style, and fabric material is obtained. And then, carrying out data cleaning on the historical sales data to obtain a plurality of data items, wherein the plurality of data items comprise data attributes and data values. And then, respectively passing the data attributes of each data item in the plurality of data items through a word embedding layer to obtain a plurality of data item attribute word embedding vectors. And secondly, respectively adding the data value of each data item in the plurality of data items to the tail of each data item attribute word embedding vector to obtain a plurality of data item embedding vectors. The plurality of data item embedded vectors are then input to a converter-based context encoder to derive historical sales data semantic feature vectors. And then, passing the historical sales data semantic feature vector through a spatial attention module to obtain an enhanced historical sales data semantic feature vector. And finally, carrying out regression decoding on the semantic feature vector of the enhanced historical sales data through a decoder to obtain a decoding value, wherein the decoding value is used for representing the predicted value of the quaternary sales.
In the production management system 100 for clothing, the historical data collection module 110 is configured to obtain historical sales data, where the historical sales data includes colors, styles and fabric materials. The method comprises the steps of obtaining historical sales data, wherein the historical sales data comprise information such as colors, styles, fabric materials and the like, and the premise is that clothing sales prediction is carried out. These data items are all important factors considered by consumers in purchasing, can reflect the preference and purchasing habit of the consumers for clothing products, and are helpful for analyzing and predicting future market demands and trends. Therefore, the technical idea of the application is to predict sales in future seasons by semantic modeling and predictive analysis of historical sales data so as to facilitate the establishment of reasonable inventory plans and purchase plans. Therefore, in the technical scheme of the application, the historical sales data of information including color, style, fabric material and the like is firstly obtained.
In the production management system 100 of the garment, the data cleansing module 120 is configured to perform data cleansing on the historical sales data to obtain a plurality of data items, where the plurality of data items include data attributes and data values. In order to improve data quality, facilitate data processing and analysis, historical sales data is data cleaned to sort the data into a format that is convenient for computer processing and analysis. It is contemplated that there may be some non-canonical, repetitive or erroneous data in the historical sales data that can seriously affect the accuracy of the subsequent predictive model. By data cleansing the historical sales data, such invalid data can be removed, thereby improving data quality. The historical sales data after data cleaning can be classified according to attributes and values, and relevant statistical analysis is performed to find out the relation and characteristics among the data, so that a basis is provided for predictive modeling.
In the production management system 100 of the garment, the attribute embedding encoding module 130 is configured to pass the data attributes of each of the plurality of data items through the word embedding layer to obtain a plurality of data item attribute word embedding vectors. In order to convert the data attribute information into a vector form which can be processed by a computer, when the historical sales data is subjected to predictive analysis, the data attributes of all data items in the historical sales data are respectively passed through a word embedding layer to obtain a plurality of data item attribute word embedding vectors. For data attributes in historical sales data, which are typically discrete variables, they cannot be directly input into the neural network model. By using word embedding techniques, discrete variables can be converted into a continuous vector representation, thereby facilitating processing and computation in the neural network. Because a large number of data attributes exist in the historical sales data, the historical sales data can be compressed into a low-dimensional vector space through a word embedding technology, and the model calculation complexity and the storage space are reduced. In particular, word embedding techniques are able to map similar words to a similar vector space and preserve the semantic similarity of the words. In the historical sales data, some data attributes may have similar or related meanings, and through word embedding, the related attributes can be mapped to a similar vector space, so that the learning capacity of a prediction model is improved.
In the production management system 100 of the above-mentioned garment, the data adding module 140 is configured to add the data value of each of the plurality of data items to the end of the attribute word embedding vector of each of the plurality of data items to obtain a plurality of data item embedding vectors. In order to fuse the data of different characteristics into one vector representation to improve the accuracy of the prediction model, the data values of each data item in the historical sales data are respectively added to the tail of the embedded vector of each data item attribute word. There may be an association or interaction between individual data items in the historical sales data, and by fusing data values with the attribute word embedding vectors, the associated information of these data may be better utilized. And, a plurality of data items in the historical sales data are fused into a vector space, so that the data characteristics can be conveniently expressed into a global vector form, and the subsequent model can be conveniently used.
In the production management system 100 of the garment described above, the historical sales data semantic understanding module 150 is configured to input the plurality of data item embedded vectors into a context encoder based on a converter to obtain historical sales data semantic feature vectors. It is contemplated that there may be complex semantic associations between individual items of data in the historical sales data, such as relationships between different attributes of color, style, and fabric. By means of the context encoder based on the converter, semantic associations between different data items can be taken into account and converted into a global semantic feature vector. Moreover, the data items and attributes in the historical sales data can be quite large, so that the data volume is huge, and by using a context encoder based on a converter, the semantic information of the historical sales data can be compressed into a vector space with lower dimensionality, so that the data volume and the model calculation complexity are reduced.
FIG. 3 is a block diagram of a historical sales data semantic understanding module in a production management system for apparel in accordance with an embodiment of the present application. As shown in fig. 3, the historical sales data semantic understanding module 150 includes: a context semantic coding unit 151 for inputting the plurality of data item embedded vectors into the converter-based context encoder to obtain a plurality of historical sales data feature vectors; and a concatenation unit 152, configured to concatenate the plurality of historical sales data feature vectors to obtain the historical sales data semantic feature vector.
Fig. 4 is a block diagram of a context semantic coding unit in a production management system of a garment according to an embodiment of the present application. As shown in fig. 4, the context semantic coding unit 151 includes: a converter subunit 1511, configured to arrange the plurality of data item embedding vectors into input vectors, and then convert the input vectors into query vectors and key vectors respectively through a learnable embedding matrix; a self-attention subunit 1512 for calculating a product between the query vector and a transpose vector of the key vector to obtain a self-attention correlation matrix; a normalized self-attention subunit 1513, configured to perform normalization processing on the self-attention correlation matrix to obtain a normalized self-attention correlation matrix; an attention computation subunit 1514 configured to activate the normalized self-attention correlation matrix input Softmax activation function to obtain a self-attention feature matrix; and an attention applying subunit 1515 configured to multiply the self-attention feature matrix with each of the plurality of data item embedding vectors to obtain the plurality of historical sales data feature vectors.
In the production management system 100 of the above-mentioned clothing, the space enhancement module 160 is configured to pass the historical sales data semantic feature vector through a space attention module to obtain an enhanced historical sales data semantic feature vector. In order to better capture the relation and importance among all data items in the historical sales data, a spatial attention module is adopted to carry out weighted fusion on semantic feature vectors of the historical sales data, so that the expression of key features is enhanced. The data items and the attributes in the historical sales data are quite many, some features are more important to predictive analysis, and the semantic feature vectors of the historical sales data can be weighted and fused through the spatial attention module, so that the influence of the important features is highlighted, and the accuracy of subsequent predictive analysis is improved.
Fig. 5 is a block diagram of a space enhancement module in a garment production management system according to an embodiment of the application. As shown in fig. 5, the spatial enhancement module 160 includes: a depth convolution encoding unit 161, configured to perform depth convolution encoding on the historical sales data semantic feature vector by using a convolution encoding portion of the spatial attention module to obtain an initial convolution feature map; a spatial attention unit 162 for inputting the initial convolution feature map into a spatial attention portion of the spatial attention module to obtain a spatial attention map; an activation unit 163 for activating the spatial attention map by Softmax activation function to obtain a spatial attention profile; a calculation unit 164 for calculating a point-by-point multiplication of the spatial attention profile and the initial convolution profile to obtain the historical sales profile; and a pooling unit 165, configured to perform an average pooling process on each feature matrix along the channel dimension on the historical sales feature map to obtain the semantic feature vector of the enhanced historical sales data.
In the production management system 100 for clothing, the sales volume prediction module 170 is configured to perform regression decoding on the semantic feature vector of the enhanced historical sales data by using a decoder to obtain a decoded value, where the decoded value is used to represent a predicted value of the sales volume in the quarter. And carrying out regression decoding on the semantic feature vector of the enhanced historical sales data through a decoder so as to convert semantic information of the historical sales data into a quaternary sales volume predicted value and complete sales volume prediction tasks. In the previous processing process, the historical sales data is subjected to multiple embedding and encoding operations, finally, the semantic feature vector of the enhanced historical sales data is obtained, regression decoding is carried out through a decoder, and the processing processes can be restored to obtain the quaternary sales quantity predicted value based on the original data. By decoding the semantic feature vector of the enhanced historical sales data into the predicted value of the quaternary sales volume, the future sales volume can be predicted, and important references are provided for the aspects of enterprise production, purchasing, inventory and the like.
Accordingly, in one specific example, the sales prediction module 170 is configured to: performing a decoding regression on the enhanced historical sales data semantic feature vector using the decoder in a decoding formula to obtain the decoded value; wherein, the decoding formula is: Wherein X is the enhanced historical sales data semantic feature vector, Y is the decoded value, W is a weight matrix,>representing matrix multiplication.
It should be appreciated that training of the word embedding layer, the converter-based context encoder, the spatial attention module and the decoder is required before utilizing the neural network model described above. That is, in the production management system of the clothing of the present application, a training module for training the word embedding layer, the converter-based context encoder, the spatial attention module, and the decoder is further included.
Fig. 6 is a block diagram of a training module in a garment production management system according to an embodiment of the application. As shown in fig. 6, the training module 200 includes: a training history data acquisition unit 210 for acquiring training history sales data; a training data cleansing unit 220, configured to perform data cleansing on the training historical sales data to obtain a plurality of training data items; a training attribute embedding encoding unit 230, configured to pass the data attributes of each training data item in the plurality of training data items through the word embedding layer to obtain a plurality of training data item attribute word embedding vectors; a training data adding unit 240, configured to add data values of each training data item in the plurality of training data items to ends of the attribute word embedding vectors of each training data item, respectively, so as to obtain a plurality of training data item embedding vectors; a training history sales data semantic understanding unit 250 for inputting the plurality of training data item embedding vectors into the converter-based context encoder to obtain training history sales data semantic feature vectors; a training space enhancement unit 260, configured to pass the training historical sales data semantic feature vector through the spatial attention module to obtain a training enhancement historical sales data semantic feature vector; the feature vector iteration unit 270 is configured to perform random reinforcement on the training enhancement historical sales data semantic feature vector based on disturbance feature expression to obtain an iteration enhancement historical sales data semantic feature vector; a decoding loss function obtaining unit 280, configured to pass the iteration enhanced historical sales data semantic feature vector through the decoder to obtain a decoding loss function value; and a model training unit 290 for training the word embedding layer, the converter-based context encoder, the spatial attention module, and the decoder based on the decoding loss function value and traveling through a direction of gradient descent.
In particular, in the solution of the present application, it is considered that there is a slight variation or disturbance of the input data or model at the time of actual inference. These minor variations or disturbances may be due to data noise, data imperfections, variations in data distribution, etc., and in particular, the input data may contain noise, such as data acquisition errors, which may have an impact on the training and inference of the model. The input data may lack certain features or attributes, or have missing values. Such imperfections can lead to uncertainty in the model in processing the data. Models are usually modeled based on a certain data distribution during training, but the data distribution in practical application may change, and the change of the distribution also causes disturbance of the model. Thus, if some tiny random perturbations can be added to the enhanced historical sales data semantic feature vector, so that the deep neural network model and classifier can adapt to these perturbations, it is clear that the resistance to the challenge sample can be improved, thereby improving the robustness of the model.
Therefore, in the technical scheme of the application, the random reinforcement of the disturbance-based feature expression is carried out on the training enhancement historical sales data semantic feature vector to obtain the iterative enhancement historical sales data semantic feature vector, which comprises the following steps: defining a random disturbance function; generating a predetermined number of random disturbance values through the random disturbance function, wherein the predetermined number is the same as the scale of the training enhanced historical sales data semantic feature vector; arranging the predetermined number of random disturbance values into a random disturbance input vector; activating the random disturbance input vector through a Softmax function to obtain a normalized random disturbance input vector, wherein the sum of characteristic values of all positions in the normalized random disturbance input vector is 1; calculating the position-based point multiplication between the normalized random disturbance input vector and the training enhancement historical sales data semantic feature vector to obtain a random action enhancement historical sales data semantic feature vector; and cascading the random action enhanced historical sales data semantic feature vector and the training enhanced historical sales data semantic feature vector to obtain an iterative enhanced historical sales data semantic feature vector.
In each iteration cycle of the training phase, disturbance-based feature expression random reinforcement is carried out on the enhanced historical sales data semantic feature vector so as to add some tiny random disturbance in the enhanced historical sales data semantic feature vector in each iteration cycle, so that a deep neural network model and a decoder can adapt to the disturbance, and the resistance to an antagonistic sample is improved. By the method, complexity of the model can be reduced, overfitting is avoided, interpretation of the model is improved, resistance of the model to noise and abnormal values can be enhanced, robustness of the model is improved, convergence speed and convergence of the model can be improved, optimization efficiency of the model is improved, sensitivity and sparsity of the model can be adjusted, and feature selection capability of the model is improved.
In summary, a production management system for apparel in accordance with an embodiment of the present application is illustrated that predicts sales in future seasons by semantic modeling and predictive analysis of historical sales data to facilitate the formulation of rational inventory plans and procurement plans. According to the scheme, through accurate sales prediction, the inventory can be better managed, excessive stock backlog is avoided or the situation that market demands cannot be met is avoided, and the stability and economic benefit of a supply chain are improved.
Exemplary method
Fig. 7 is a flowchart of a method of production management of garments according to an embodiment of the application. As shown in fig. 7, the production management method of the garment according to the embodiment of the present application includes the steps of: s110, acquiring historical sales data, wherein the historical sales data comprises colors, styles and fabric materials; s120, data cleaning is carried out on the historical sales data to obtain a plurality of data items, wherein the plurality of data items comprise data attributes and data values; s130, respectively passing the data attribute of each data item in the plurality of data items through a word embedding layer to obtain a plurality of data item attribute word embedding vectors; s140, respectively adding the data value of each data item in the plurality of data items to the tail of each data item attribute word embedding vector to obtain a plurality of data item embedding vectors; s150, inputting the plurality of data item embedded vectors into a context encoder based on a converter to obtain historical sales data semantic feature vectors; s160, passing the historical sales data semantic feature vector through a spatial attention module to obtain an enhanced historical sales data semantic feature vector; and S170, carrying out regression decoding on the semantic feature vector of the enhanced historical sales data through a decoder to obtain a decoding value, wherein the decoding value is used for representing the predicted value of the quarter sales.
In a specific example, in the method for managing clothing production described above, the step S150 of inputting the plurality of data item embedded vectors into a context encoder based on a converter to obtain historical sales data semantic feature vectors includes: inputting the plurality of data item embedded vectors into the converter-based context encoder to obtain a plurality of historical sales data feature vectors; and cascading the plurality of historical sales data feature vectors to obtain the historical sales data semantic feature vector.
In a specific example, in the method for managing production of clothing described above, inputting the plurality of data item embedding vectors into the converter-based context encoder to obtain a plurality of historical sales data feature vectors includes: arranging the plurality of data item embedding vectors into input vectors, and respectively converting the input vectors into query vectors and key vectors through a learning embedding matrix; calculating the product between the query vector and the transpose vector of the key vector to obtain a self-attention correlation matrix; carrying out standardization processing on the self-attention association matrix to obtain a standardized self-attention association matrix; inputting the standardized self-attention association matrix into a Softmax activation function to activate so as to obtain a self-attention feature matrix; and multiplying the self-attention feature matrix with each data item embedding vector in the plurality of data item embedding vectors to obtain the plurality of historical sales data feature vectors.
In a specific example, in the method for managing clothing production, the step S160, passing the historical sales data semantic feature vector through a spatial attention module to obtain an enhanced historical sales data semantic feature vector includes: performing depth convolution encoding on the historical sales data semantic feature vector by using a convolution encoding part of the spatial attention module to obtain an initial convolution feature map; inputting the initial convolution feature map into a spatial attention portion of the spatial attention module to obtain a spatial attention map; -passing said spatial attention map through a Softmax activation function to obtain a spatial attention profile; calculating the position-wise point multiplication of the space attention characteristic diagram and the initial convolution characteristic diagram to obtain the historical sales characteristic diagram; and carrying out the mean value pooling processing of each feature matrix along the channel dimension on the historical sales feature map to obtain the semantic feature vector of the enhanced historical sales data.
In a specific example, in the method for managing clothing production, the step S170 is performed to perform regression decoding on the enhanced historical sales data semantic feature vector by a decoder to obtain a decoded value, where the decoded value is used to represent a predicted value of a quarter sales volume, and the method includes: performing a decoding regression on the enhanced historical sales data semantic feature vector using the decoder in a decoding formula to obtain the decoded value; wherein, the decoding formula is: Wherein X is the enhanced historical sales data semantic feature vector, Y is the decoded value, W is a weight matrix,>representing matrix multiplication.
In a specific example, in the method for managing production of clothing described above, training the word embedding layer, the converter-based context encoder, the spatial attention module, and the decoder is further included; wherein training the word embedding layer, the converter-based context encoder, the spatial attention module, and the decoder comprises: acquiring training historical sales data; data cleaning is carried out on the training historical sales data to obtain a plurality of training data items; respectively passing the data attribute of each training data item in the plurality of training data items through the word embedding layer to obtain a plurality of training data item attribute word embedding vectors; respectively adding the data value of each training data item in the plurality of training data items to the tail of each training data item attribute word embedding vector to obtain a plurality of training data item embedding vectors; inputting the plurality of training data item embedded vectors into the converter-based context encoder to obtain training history sales data semantic feature vectors; passing the training historical sales data semantic feature vector through the spatial attention module to obtain a training enhancement historical sales data semantic feature vector; randomly strengthening the disturbance-based feature expression of the training enhancement historical sales data semantic feature vector to obtain an iteration enhancement historical sales data semantic feature vector; passing the iterative enhanced historical sales data semantic feature vector through the decoder to obtain a decoding loss function value; and training the word embedding layer, the converter-based context encoder, the spatial attention module, and the decoder based on the decoding loss function value and traveling through a direction of gradient descent.
In a specific example, in the method for managing clothing production, performing disturbance-based feature expression random reinforcement on the training enhanced historical sales data semantic feature vector to obtain an iterative enhanced historical sales data semantic feature vector, including: defining a random disturbance function; generating a predetermined number of random disturbance values through the random disturbance function, wherein the predetermined number is the same as the scale of the training enhanced historical sales data semantic feature vector; arranging the predetermined number of random disturbance values into a random disturbance input vector; activating the random disturbance input vector through a Softmax function to obtain a normalized random disturbance input vector, wherein the sum of characteristic values of all positions in the normalized random disturbance input vector is 1; calculating the position-based point multiplication between the normalized random disturbance input vector and the training enhancement historical sales data semantic feature vector to obtain a random action enhancement historical sales data semantic feature vector; and cascading the random action enhanced historical sales data semantic feature vector and the training enhanced historical sales data semantic feature vector to obtain an iterative enhanced historical sales data semantic feature vector.
Here, it will be understood by those skilled in the art that the specific operations of the respective steps in the production management method of the above-described clothing have been described in detail in the above description of the production management system of the clothing with reference to fig. 1 to 6, and thus, repetitive descriptions thereof will be omitted.
Claims (10)
1. A garment production management system, comprising:
the historical data acquisition module is used for acquiring historical sales data, wherein the historical sales data comprises colors, styles and fabric materials;
the data cleaning module is used for carrying out data cleaning on the historical sales data to obtain a plurality of data items, wherein the plurality of data items comprise data attributes and data values;
the attribute embedding coding module is used for respectively passing the data attribute of each data item in the plurality of data items through the word embedding layer to obtain a plurality of data item attribute word embedding vectors;
the data adding module is used for respectively adding the data value of each data item in the plurality of data items to the tail of each data item attribute word embedding vector so as to obtain a plurality of data item embedding vectors;
a historical sales data semantic understanding module for inputting the plurality of data item embedded vectors into a context encoder based on a converter to obtain a historical sales data semantic feature vector;
The space enhancement module is used for enabling the historical sales data semantic feature vectors to pass through the space attention module so as to obtain enhanced historical sales data semantic feature vectors; and
and the sales volume prediction module is used for carrying out regression decoding on the enhanced historical sales data semantic feature vector through a decoder to obtain a decoding value, wherein the decoding value is used for representing the quaternary sales volume predicted value.
2. The garment production management system of claim 1, wherein the historical sales data semantic understanding module comprises:
a context semantic coding unit for inputting the plurality of data item embedded vectors into the converter-based context encoder to obtain a plurality of historical sales data feature vectors; and
and the cascading unit is used for cascading the plurality of historical sales data feature vectors to obtain the historical sales data semantic feature vectors.
3. The production management system of apparel according to claim 2, characterized in that the contextual semantic coding unit comprises:
the conversion subunit is used for arranging the plurality of data item embedding vectors into input vectors and respectively converting the input vectors into query vectors and key vectors through a learnable embedding matrix;
A self-attention subunit, configured to calculate a product between the query vector and a transpose vector of the key vector to obtain a self-attention correlation matrix;
a standardized self-attention subunit, configured to perform a standardization process on the self-attention association matrix to obtain a standardized self-attention association matrix;
an attention calculating subunit, configured to activate the normalized self-attention association matrix input Softmax activation function to obtain a self-attention feature matrix; and
and the attention applying subunit is used for multiplying the self-attention characteristic matrix with each data item embedding vector in the plurality of data item embedding vectors to obtain the plurality of historical sales data characteristic vectors.
4. A production management system for garments according to claim 3, wherein said spatial enhancement module comprises:
the depth convolution coding unit is used for performing depth convolution coding on the historical sales data semantic feature vector by using a convolution coding part of the spatial attention module so as to obtain an initial convolution feature map;
a spatial attention unit for inputting the initial convolution feature map into a spatial attention portion of the spatial attention module to obtain a spatial attention map;
An activation unit, configured to activate the spatial attention map through a Softmax activation function to obtain a spatial attention profile;
a calculation unit, configured to calculate a point-by-point multiplication of the spatial attention feature map and the initial convolution feature map to obtain the historical sales feature map; and
and the pooling unit is used for carrying out the mean pooling processing of each feature matrix along the channel dimension on the historical sales feature graph so as to obtain the semantic feature vector of the enhanced historical sales data.
5. The garment production management system of claim 4, wherein the sales prediction module is configured to:
performing a decoding regression on the enhanced historical sales data semantic feature vector using the decoder in a decoding formula to obtain the decoded value; wherein, the decoding formula is:wherein X is the enhanced historical sales data semantic feature vector, Y is the decoded value, W is a weight matrix,>representing matrix multiplication.
6. The garment production management system of claim 5, further comprising a training module for training the word embedding layer, the converter-based context encoder, the spatial attention module, and the decoder;
Wherein, training module includes:
the training history data acquisition unit is used for acquiring training history sales data;
the training data cleaning unit is used for cleaning the training historical sales data to obtain a plurality of training data items;
the training attribute embedding coding unit is used for respectively passing the data attribute of each training data item in the plurality of training data items through the word embedding layer to obtain a plurality of training data item attribute word embedding vectors;
the training data adding unit is used for respectively adding the data value of each training data item in the plurality of training data items to the tail of each training data item attribute word embedding vector so as to obtain a plurality of training data item embedding vectors;
a training history sales data semantic understanding unit for inputting the plurality of training data item embedded vectors into the converter-based context encoder to obtain training history sales data semantic feature vectors;
the training space enhancement unit is used for enabling the training historical sales data semantic feature vectors to pass through the space attention module so as to obtain training enhancement historical sales data semantic feature vectors;
the feature vector iteration unit is used for carrying out random reinforcement on the feature expression based on disturbance on the training enhancement historical sales data semantic feature vector so as to obtain an iteration enhancement historical sales data semantic feature vector;
The decoding loss function obtaining unit is used for enabling the iteration enhanced historical sales data semantic feature vector to pass through the decoder so as to obtain a decoding loss function value; and
a model training unit for training the word embedding layer, the converter-based context encoder, the spatial attention module and the decoder based on the decoding loss function value and traveling through a direction of gradient descent.
7. The garment production management system of claim 6, wherein the feature vector iteration unit comprises:
a random disturbance value generation subunit, configured to define a random disturbance function, and generate a predetermined number of random disturbance values through the random disturbance function, where the predetermined number is the same as the scale of the semantic feature vector of the training enhancement historical sales data;
a random disturbance value arrangement subunit configured to arrange the predetermined number of random disturbance values into a random disturbance input vector;
an activation subunit, configured to activate the random disturbance input vector through a Softmax function to obtain a normalized random disturbance input vector, where a sum of eigenvalues of all positions in the normalized random disturbance input vector is 1;
The random disturbance applying subunit is used for calculating the point-by-point multiplication between the normalized random disturbance input vector and the training enhancement historical sales data semantic feature vector to obtain a random action enhancement historical sales data semantic feature vector; and
and the cascading subunit is used for cascading the random action enhancement historical sales data semantic feature vector and the training enhancement historical sales data semantic feature vector to obtain an iteration enhancement historical sales data semantic feature vector.
8. A method of manufacturing management of apparel, comprising:
acquiring historical sales data, wherein the historical sales data comprises colors, styles and fabric materials;
performing data cleaning on the historical sales data to obtain a plurality of data items, wherein the plurality of data items comprise data attributes and data values;
respectively passing the data attribute of each data item in the plurality of data items through a word embedding layer to obtain a plurality of data item attribute word embedding vectors;
respectively adding the data value of each data item in the plurality of data items to the tail of each data item attribute word embedding vector to obtain a plurality of data item embedding vectors;
Inputting the plurality of data item embedded vectors into a converter-based context encoder to obtain historical sales data semantic feature vectors;
passing the historical sales data semantic feature vector through a spatial attention module to obtain an enhanced historical sales data semantic feature vector; and
and carrying out regression decoding on the semantic feature vector of the enhanced historical sales data through a decoder to obtain a decoding value, wherein the decoding value is used for representing the predicted value of the quaternary sales.
9. The method of claim 8, wherein inputting the plurality of data item embedded vectors into a transducer-based context encoder to obtain historical sales data semantic feature vectors comprises:
inputting the plurality of data item embedded vectors into the converter-based context encoder to obtain a plurality of historical sales data feature vectors; and
cascading the plurality of historical sales data feature vectors to obtain the historical sales data semantic feature vector.
10. The method of claim 9, wherein inputting the plurality of data item embedded vectors into the converter-based context encoder to obtain a plurality of historical sales data feature vectors comprises:
Arranging the plurality of data item embedding vectors into input vectors, and respectively converting the input vectors into query vectors and key vectors through a learning embedding matrix;
calculating the product between the query vector and the transpose vector of the key vector to obtain a self-attention correlation matrix;
carrying out standardization processing on the self-attention association matrix to obtain a standardized self-attention association matrix;
inputting the standardized self-attention association matrix into a Softmax activation function to activate so as to obtain a self-attention feature matrix; and
and multiplying the self-attention characteristic matrix with each data item embedded vector in the plurality of data item embedded vectors respectively to obtain the plurality of historical sales data characteristic vectors.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310836182.5A CN116993393A (en) | 2023-07-07 | 2023-07-07 | Clothing production order management system and method thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310836182.5A CN116993393A (en) | 2023-07-07 | 2023-07-07 | Clothing production order management system and method thereof |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116993393A true CN116993393A (en) | 2023-11-03 |
Family
ID=88520532
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310836182.5A Pending CN116993393A (en) | 2023-07-07 | 2023-07-07 | Clothing production order management system and method thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116993393A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117455059A (en) * | 2023-11-06 | 2024-01-26 | 深圳市乐思软件技术有限公司 | Industry trend evaluation system based on data acquisition |
-
2023
- 2023-07-07 CN CN202310836182.5A patent/CN116993393A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117455059A (en) * | 2023-11-06 | 2024-01-26 | 深圳市乐思软件技术有限公司 | Industry trend evaluation system based on data acquisition |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Guo et al. | Supplier selection based on hierarchical potential support vector machine | |
| US20240177075A1 (en) | Method and system for hierarchical forecasting | |
| WO2019214455A1 (en) | Data sequence prediction method and computing apparatus | |
| Shchetinin | Cluster-based energy consumption forecasting in smart grids | |
| CN116993393A (en) | Clothing production order management system and method thereof | |
| CN110727855A (en) | Personalized recommendation method based on improved factorization machine | |
| CN111695024A (en) | Object evaluation value prediction method and system, and recommendation method and system | |
| CN116629977A (en) | Electronic commerce decision method and system based on big data | |
| CN114066510A (en) | Method and device for predicting commodity sales, computer equipment and storage medium | |
| CN115099502B (en) | Short-term power load prediction method based on inter-user power consumption behavior similarity | |
| CN113298546A (en) | Sales prediction method and device, and commodity processing method and device | |
| CN114547307A (en) | Text vector model training method, text matching method, device and equipment | |
| Mohamed et al. | Price Prediction of Seasonal Items Using Machine Learning and Statistical Methods. | |
| CN108846692B (en) | Multi-factor recurrent neural network-based consumer consumption behavior prediction method | |
| Andra | E-commerce Recommender System Using PCA and K-Means Clustering | |
| CN111738812B (en) | Information pushing method and system based on user group micro-segmentation | |
| CN116777528A (en) | Commodity information recommendation method and device, computer equipment and storage medium | |
| CN116703523A (en) | Electronic commerce system based on big data and method thereof | |
| CN115130554A (en) | Object classification method and device, electronic equipment and storage medium | |
| CN111782802B (en) | Method and system for obtaining commodity corresponding to national economy manufacturing industry based on machine learning | |
| CN116485084A (en) | Intelligent power material demand decision method and system based on data space | |
| CN114092188A (en) | Recommendation system algorithm of lightweight B2B E-commerce platform | |
| CN114169914A (en) | Feature construction method, device and storage medium for item sales prediction | |
| Spelta et al. | Wasserstein barycenter for link prediction in temporal networks | |
| Li et al. | Analysis and research of retail customer consumption behavior based on support vector machine |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |