Decision tree-based dispatching and production scheduling method

By processing and dimensionality reduction of data based on decision tree methods, building a decision tree model solves the problem of lack of scientificity and systematicity of traditional labor dispatching and production scheduling methods, realizing the generation and optimization of real-time labor dispatching and production scheduling solutions, improving the Productivity and customer satisfaction.

CN118917631AActive Publication Date: 2024-11-08INSPUR YUNZHOU (SHANDONG) IND INTERNET CO LTD

Patent Information

Application Number
CN202411396968.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-09
Publication Date
2024-11-08
Estimated Expiration
2044-10-09

AI Technical Summary

Technical Problem

The traditional method of dispatching production scheduling relies on manual experience, lacks scientificity and systematicity, and is difficult to adapt to changes in the modern production environment. The existing data-driven method has high complexity, large calculation volume, and is difficult to adjust in real time.

Method used

Using a decision tree-based method, by processing and reducing the dimensions of the collected data, building a decision tree model, and generating a dispatched production plan in real time to adapt to changes in production environment and demand.

Benefits of technology

It improves data quality and consistency, realizes intelligent scheduling and production scheduling optimization of production tasks, reduces production costs, shortens lead times, and improves customer satisfaction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118917631A_ABST
    Figure CN118917631A_ABST
Patent Text Reader

Abstract

The invention provides a dispatching and production scheduling method based on a decision tree, and belongs to the technical field of data processing, and the method comprises the steps: collecting data related to dispatching and production scheduling to determine first data, and carrying out the first processing of the first data to determine second data; determining a first sub-feature, a second sub-feature, a third sub-feature and a fourth sub-feature based on the second data, and determining a first comprehensive feature and a second comprehensive feature; determining a first feature based on the second data and the first comprehensive feature, determining a second feature based on the second data and the second comprehensive feature, and determining a dimension reduction feature; and constructing a decision tree model based on the second data and the dimension reduction features, and determining and executing a dispatching and production scheduling scheme of the real-time order information based on the decision tree model. All data related to dispatching and production scheduling can be comprehensively analyzed, the data quality and consistency are improved, a dispatching and production scheduling scheme with high feasibility and high execution efficiency is generated in real time to adapt to changes of the production environment and requirements, the production cost is reduced, and the delivery cycle is shortened.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method for dispatching work and scheduling production based on a decision tree. Background Art

[0002] In the manufacturing industry, work scheduling is a vital link, which involves the rational allocation of production resources and the effective execution of production tasks. Traditional methods of work scheduling often rely on manual experience and intuition, lacking scientificity and systematicity. With the rapid development of the manufacturing industry and the continuous changes in production needs, traditional methods can no longer meet the requirements of modern production.

[0003] In recent years, with the continuous development of artificial intelligence and big data technology, data-driven scheduling methods have gradually become a research hotspot. These methods collect and analyze production data, use machine learning and other algorithms to build models, and achieve intelligent monitoring and optimization of the production process. However, existing methods still have some limitations, such as high model complexity, large amount of calculation, and difficulty in adapting to real-time changes.

[0004] Therefore, the present invention provides a method for dispatching work and scheduling production based on a decision tree. Summary of the invention

[0005] The present invention provides a method for dispatching and scheduling work based on a decision tree, which is used for determining second data by performing a first processing on collected first data, analyzing the second data to determine a first comprehensive feature, a second comprehensive feature, a first feature and a second feature, and determining a dimensionality reduction feature, building a decision tree model according to the second data and the dimensionality reduction feature, determining and executing a dispatching and scheduling plan for real-time order information based on the decision tree model, and can comprehensively analyze all data related to dispatching and scheduling work, improve data quality and consistency, generate a dispatching and scheduling plan in real time to adapt to changes in production environment and demand, simply and efficiently realize intelligent scheduling and scheduling optimization of production tasks, improve the feasibility and execution efficiency of the dispatching and scheduling plan, reduce production costs, shorten delivery cycles, and improve customer satisfaction.

[0006] The present invention provides a method for dispatching work and scheduling production based on a decision tree, comprising: 101: collecting data related to work dispatching and production scheduling to determine first data, and performing first processing on the first data to determine second data; 102: Determine a first sub-feature, a second sub-feature, a third sub-feature, and a fourth sub-feature based on the second data, and determine a first comprehensive feature and a second comprehensive feature; 103: determining a first feature based on the second data and the first comprehensive feature, and determining a second feature based on the second data and the second comprehensive feature, and determining a dimensionality reduction feature; 104: Build a decision tree model based on the second data and the dimension reduction features, and determine and execute a work dispatching and production scheduling plan based on the real-time order information based on the decision tree model.

[0007] According to a decision tree-based work dispatching and production scheduling method provided by the present invention, the first data includes a plurality of historical order information; The first processing includes data cleaning, missing value processing and data normalization on the first data.

[0008] According to a decision tree-based work dispatching and production scheduling method provided by the present invention, a first sub-feature, a second sub-feature, a third sub-feature, and a fourth sub-feature are determined based on second data, including: Extracting data related to the value in the second data to determine first sub-data, wherein the first sub-data includes a plurality of value categories related to the value in the second data; Performing statistical analysis on each numerical category in the first sub-data respectively, and determining the first sub-feature based on the statistical analysis results of all numerical categories; Extracting time-related data from the second data to determine second sub-data, wherein the second sub-data includes a plurality of time-related time series categories from the second data; Performing time series analysis on each time series category in the second sub-data respectively, and determining the second sub-feature based on the time series analysis results of all time series categories; Extract all the remaining data except the first sub-data and the second sub-data from the second data to determine the third sub-data, wherein the third sub-data includes a plurality of classification categories; determining a third sub-feature based on each category in the third sub-data; Encode all identifiers under each classification category in the third sub-data respectively, interactively determine worker sub-features for all identifier codes under all classification categories related to workers, interactively determine equipment sub-features for all identifier codes under all classification categories related to equipment, and interactively determine environment sub-features for all identifier codes under all classification categories related to the environment, wherein the worker sub-features include multiple worker vectors, the equipment sub-features include multiple equipment vectors, and the environment sub-features include multiple environment vectors; All identification codes under the worker sub-feature, the equipment sub-feature, and the environment sub-feature are arranged and combined to determine a fourth sub-feature, wherein the fourth sub-feature includes a plurality of category vectors.

[0009] According to a decision tree-based work dispatching and production scheduling method provided by the present invention, determining the first comprehensive feature and the second comprehensive feature includes: Performing standardization processing on the first sub-feature, the second sub-feature, and the third sub-feature, and determining a first comprehensive feature based on the first sub-feature, the second sub-feature, and the third sub-feature after the standardization processing; The first sub-feature, the second sub-feature, and the fourth sub-feature are standardized, and a second comprehensive feature is determined based on the standardized first sub-feature, the second sub-feature, and the fourth sub-feature, wherein the second comprehensive feature includes a plurality of feature vectors.

[0010] According to a decision tree-based work dispatching and production scheduling method provided by the present invention, the first feature is determined based on the second data and the first comprehensive feature, including: Determine a corresponding first order vector for each piece of historical order information in the second data according to the first comprehensive feature; Determine an order matrix based on the first order vector of all historical order information; Wherein, N1 represents the number of historical order information in the second data, and N2 represents the number of sub-features in the first comprehensive feature. , , represent the order value of the first sub-feature of the first historical order information in the second data, the feature value of the j-th sub-feature, and the feature value of the N2-th sub-feature in the first comprehensive feature, respectively. , , They respectively represent the order value of the 1st sub-feature, the feature value of the jth sub-feature, and the feature value of the N2th sub-feature of the i-th historical order information in the second data in the first comprehensive feature, , , Respectively represent the feature value of the 1st sub-feature, the feature value of the jth sub-feature, and the order value of the N2th sub-feature of the N1th historical order information in the second data in the first comprehensive feature; determining a first eigenvalue and a first eigenvector based on the order matrix; Where C is the covariance matrix, is the transposed matrix of the order matrix, is the kth first eigenvalue, The kth first eigenvector represented by ; Sort the first eigenvalues ​​from large to small, and calculate the importance rates of the first Nu first eigenvalues; in, , represents the importance rate of the first Nu eigenvalues ​​after sorting, represents the adjustment parameters, represents the pth first eigenvalue of the order matrix after sorting, represents the pth first eigenvalue of the order matrix after sorting based on the adjustment parameter, represents the cumulative importance of the first Nu first eigenvalues ​​after the order matrix is ​​sorted, represents the cumulative importance of all first eigenvalues ​​after the order matrix is ​​sorted, It represents the first importance rate of the first Nu first eigenvalues ​​after the order matrix is ​​sorted, represents the first weight of the first importance rate, represents the second importance rate of the first Nu first eigenvalues ​​after the order matrix is ​​sorted, a second weight representing a second importance rate; Extract all Nu values ​​with importance greater than 95%, and select the top values ​​after sorting. The first eigenvalues ​​construct the first projection matrix; The second data is projected into the first projection matrix to determine the first feature.

[0011] According to a decision tree-based work dispatching and production scheduling method provided by the present invention, a second feature is determined based on the second data and the second comprehensive feature, and a dimension reduction feature is determined, including: Match all historical order information in the second data based on the second comprehensive feature, determine an order set for each feature vector, and a second order vector for each historical order information in the order set; Determining a second eigenvalue and a second eigenvector based on the order set of all eigenvectors; in, Represents the deviation value of the corresponding feature vector of all order sets, represents the deviation value between all eigenvectors, represents the bth second order vector in the order set of the ath feature vector, represents the second order fitting vector in the order set of the a-th eigenvector, The difference vector between the bth second order vector and the second order fitting vector in the order set of the ath feature vector, represents the transposed row vector of the difference vector between the bth second order vector and the mean of the second order vector in the order set of the ath eigenvector, represents the number of the second order vector in the order set of the a-th feature vector, represents the number of eigenvectors in the second comprehensive feature, The total fitted value of the second order vector representing the set of orders for all feature vectors in the second data, The difference vector between the second order fitting vector of the order set of the a-th feature vector and the total fitting value of the second order vector, The transposed row vector of the difference vector between the second order fitting vector and the total fitting value of the second order vector for the order set of the a-th eigenvector, is the cth second eigenvalue of, is the cth second eigenvector, ; Select the first The second eigenvalues ​​construct the second projection matrix; Projecting the second data into a second projection matrix to determine a second feature; Features that appear simultaneously in the first feature and the second feature are extracted to determine the dimension reduction features.

[0012] According to a method for dispatching and scheduling work based on a decision tree provided by the present invention, a decision tree model is constructed based on the second data and the dimension reduction feature, and a dispatching and scheduling plan for real-time order information is determined and executed based on the decision tree model, including: Divide the second data into training set and test set, Build a decision tree model based on the dimensionality reduction features, historical order requirements of all historical order information in the training set, and historical execution plans; The historical order requirements of all historical order information in the test set are used as the input of the trained decision tree model, and the decision tree model outputs the decision execution plan corresponding to each historical order information; The model performance is evaluated based on the historical execution plan and decision execution plan of each historical order information, and the decision tree model is first optimized based on the evaluation results.

[0013] According to a method for dispatching and scheduling work based on a decision tree provided by the present invention, a decision tree model is constructed based on the second data and the dimension reduction feature, and a dispatching and scheduling plan for real-time order information is determined and executed based on the decision tree model, and further includes: Input the real-time order information into the optimized decision tree model to generate a work dispatching and production scheduling plan for the real-time order information; Execute the work dispatching and production scheduling plan, evaluate the execution result of the work dispatching and production scheduling plan, and perform a second optimization on the decision tree model based on the execution result.

[0014] Compared with the prior art, the present invention has the following beneficial effects: The second data is determined by performing a first processing on the collected first data, the first comprehensive feature, the second comprehensive feature, the first feature and the second feature are determined by analyzing the second data, and the dimensionality reduction feature is determined, a decision tree model is constructed according to the second data and the dimensionality reduction feature, and a work dispatching and production scheduling plan for the real-time order information is determined and executed based on the decision tree model. All data related to work dispatching and production scheduling can be comprehensively analyzed to improve data quality and consistency, and work dispatching and production scheduling plans can be generated in real time to adapt to changes in production environment and demand, so as to simply and efficiently realize intelligent scheduling and production scheduling optimization of production tasks, improve the feasibility and execution efficiency of work dispatching and production scheduling plans, reduce production costs, shorten delivery cycles, and improve customer satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0016] Figure 1 It is a flowchart of a decision tree-based work dispatching and production scheduling method provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] Embodiment 1: The embodiment of the present invention provides a method for dispatching work and scheduling production based on a decision tree, such as Figure 1 As shown, including: 101: collecting data related to work dispatching and production scheduling to determine first data, and performing first processing on the first data to determine second data; 102: Determine a first sub-feature, a second sub-feature, a third sub-feature, and a fourth sub-feature based on the second data, and determine a first comprehensive feature and a second comprehensive feature; 103: determining a first feature based on the second data and the first comprehensive feature, and determining a second feature based on the second data and the second comprehensive feature, and determining a dimensionality reduction feature; 104: Build a decision tree model based on the second data and the dimension reduction features, and determine and execute a work dispatching and production scheduling plan based on the real-time order information based on the decision tree model.

[0019] In this embodiment, the second data represents data after the collected first data is cleaned, missing value filled, normalized, and transformed.

[0020] In this embodiment, the first sub-feature represents a feature determined after statistical analysis is performed on all data related to the numerical value in the second data.

[0021] In this embodiment, the second sub-feature represents a feature determined after time series analysis is performed on all time-related data in the second data.

[0022] In this embodiment, the third sub-feature represents a feature determined for all categories included in all the remaining data in the second data except the first sub-data and the second sub-data.

[0023] In this embodiment, the fourth sub-feature is a plurality of category vectors obtained by arranging and combining all identification codes under the worker sub-feature, the equipment sub-feature, and the environment sub-feature.

[0024] In this embodiment, the first comprehensive feature is determined according to the first sub-feature, the second sub-feature and the third sub-feature, and the second comprehensive feature is determined according to the first sub-feature, the second sub-feature and the fourth sub-feature.

[0025] In this embodiment, the dimension reduction features are all features that appear in both the first feature and the second feature.

[0026] The beneficial effects of the above technical solution are as follows: by performing a first processing on the collected first data to determine the second data, analyzing the second data to determine the first comprehensive feature, the second comprehensive feature, the first feature and the second feature, and determining the dimensionality reduction feature, a decision tree model is constructed according to the second data and the dimensionality reduction feature, and the dispatching and scheduling plan of the real-time order information is determined and executed based on the decision tree model. All data related to dispatching and scheduling can be comprehensively analyzed to improve data quality and consistency, and dispatching and scheduling plans can be generated in real time to adapt to changes in the production environment and demand. The intelligent scheduling and scheduling optimization of production tasks can be realized simply and efficiently, and the feasibility and execution efficiency of the dispatching and scheduling plans can be improved, so as to reduce production costs, shorten delivery cycles, and improve customer satisfaction.

[0027] Embodiment 2: An embodiment of the present invention provides a method for dispatching and scheduling work based on a decision tree, wherein the first data includes a plurality of historical order information; The first processing includes data cleaning, missing value processing and data normalization on the first data.

[0028] In this embodiment, the historical order information includes worker working hours, dispatch time, execution time, equipment usage time, order completion time, worker's specialty, worker skill level, worker's years of work experience, worker shift, equipment type, equipment status, working temperature, humidity and light, etc.; In this embodiment, data cleaning means deleting duplicate records, processing abnormal values, and correcting data entry errors on the collected first data related to work dispatch and production scheduling.

[0029] In this embodiment, the missing value processing means deleting missing values ​​or filling missing values ​​in the first data after data cleaning.

[0030] In this embodiment, data normalization means converting the first data after the missing value processing into a unified scale.

[0031] The beneficial effects of the above technical solution are as follows: by determining the first data and the first processing, a data basis can be provided for determining the first sub-feature, the second sub-feature, the third sub-feature and the fourth sub-feature, and all data related to work dispatching and production scheduling can be comprehensively analyzed to improve data quality and consistency.

[0032] Embodiment 3: An embodiment of the present invention provides a method for dispatching and scheduling work based on a decision tree, which determines a first sub-feature, a second sub-feature, a third sub-feature, and a fourth sub-feature based on second data, including: Extracting data related to the value in the second data to determine first sub-data, wherein the first sub-data includes a plurality of value categories related to the value in the second data; Performing statistical analysis on each numerical category in the first sub-data respectively, and determining the first sub-feature based on the statistical analysis results of all numerical categories; Extracting time-related data from the second data to determine second sub-data, wherein the second sub-data includes a plurality of time-related time series categories from the second data; Performing time series analysis on each time series category in the second sub-data respectively, and determining the second sub-feature based on the time series analysis results of all time series categories; Extract all the remaining data except the first sub-data and the second sub-data from the second data to determine the third sub-data, wherein the third sub-data includes a plurality of classification categories; determining a third sub-feature based on each category in the third sub-data; Encode all identifiers under each classification category in the third sub-data respectively, interactively determine worker sub-features for all identifier codes under all classification categories related to workers, interactively determine equipment sub-features for all identifier codes under all classification categories related to equipment, and interactively determine environment sub-features for all identifier codes under all classification categories related to the environment, wherein the worker sub-features include multiple worker vectors, the equipment sub-features include multiple equipment vectors, and the environment sub-features include multiple environment vectors; All identification codes under the worker sub-feature, the equipment sub-feature, and the environment sub-feature are arranged and combined to determine a fourth sub-feature, wherein the fourth sub-feature includes a plurality of category vectors.

[0033] In this embodiment, the numerical categories included in the first sub-data may be the number of historical orders of the second data, the equipment operating time of each historical order information, and the working hours of workers, etc.

[0034] In this embodiment, the statistical analysis may include performing trend analysis, dispersion analysis, peak and valley value determination, and extreme value determination on the data values ​​of all historical order information under each numerical value category.

[0035] In this embodiment, the first sub-feature includes basic numerical data features and composite features. The basic numerical data features may be order quantity features and average working hours of workers, and the composite features may be worker efficiency and equipment capacity efficiency, etc.

[0036] In this embodiment, the time series categories included in the second sub-data may be the dispatch time, execution time, equipment usage time, order completion time, etc. of all historical order information.

[0037] In this embodiment, the time series analysis may include performing moving average, periodicity decomposition, autocorrelation coefficient calculation, and difference on all time values ​​under each time series category.

[0038] In this embodiment, the second sub-feature may be dispatch time volatility (dispatching time is different in different time periods), execution time trend (execution time is shortened due to improved production efficiency), and equipment usage time periodicity.

[0039] In this embodiment, the classification categories included in the third sub-data may be the type of work that the worker is good at, the worker's skill level, the worker's years of work, the worker's shift, equipment type, equipment status, working temperature, suitability, and lighting, etc.

[0040] In this embodiment, the third sub-feature includes all classification categories. If the third sub-data has several classification categories, the third sub-feature has several features. For example, if the third sub-data includes 6 classification categories, namely worker skill level, worker working experience, worker shift, equipment type, equipment status and environmental area, the third sub-feature includes 6 features, namely worker skill level feature, worker working experience feature, worker shift feature, equipment type feature, equipment status feature and environmental area feature.

[0041] In this embodiment, all identification codes under all classification categories related to the worker are interactively determined to determine the worker's sub-features, for example, classification categories: worker skill level (such as: primary, intermediate, senior), worker working years (such as: 1-3 years, 3-5 years, more than 5 years), worker's proficiency in work (such as: welder, electrician, machine operator), worker shift (such as: morning shift, evening shift, night shift), identification code: skill level code: S1 (primary), S2 (intermediate), S3 (senior), working years code: Y1 ( The worker vectors are: S2, Y1, T1, B1 (representing a mid-level welder with 1-3 years of experience, responsible for the morning shift), S3, Y3, T2, B3 (representing a senior electrician with more than 5 years of experience, responsible for the night shift).

[0042] In this embodiment, the worker sub-feature can be used to identify workers with specific skill levels, experience, job types, and shifts.

[0043] In this embodiment, all identification codes under all classification categories related to the equipment are interactively determined to determine the equipment sub-features, for example, classification categories: equipment type (such as mechanical equipment, electronic equipment, sensor equipment), equipment status (such as normal, under repair, faulty), equipment location (such as production line A, production line B, warehouse), identification code: equipment type code: M1 (mechanical equipment), M2 (electronic equipment), M3 (sensor equipment), equipment status code: N1 (normal), N2 (under repair), N3 (faulty), equipment location code: L1 (production line A), L2 (production line B), L3 (warehouse), equipment sub-feature vector example: equipment vector 1: M1, N1, L1 (indicating a normally working mechanical equipment, located on production line A), equipment vector 2: M3, N2, L3 (indicating a sensor equipment under repair, located in a warehouse).

[0044] In this embodiment, the device sub-features may be used to identify a device having a specific device type, device status, device location, and the like.

[0045] In this embodiment, all identification codes under all classification categories related to the environment are interactively determined to determine environmental sub-features, for example, classification categories: temperature range (such as low temperature, medium temperature, high temperature), humidity range (such as dry, moderate, humid), lighting conditions (such as weak light, normal light, strong light), identification codes: temperature range codes: p1 (low temperature), p2 (medium temperature), p3 (high temperature), humidity range codes: H1 (dry), H2 (moderate), H3 (humid), lighting condition codes: Q1 (weak light), Q2 (normal light), Q3 (strong light), environment sub-feature vector examples: environment vector 1: p2, H2, Q2 (indicating an environment of medium temperature, moderate humidity, and normal light), environment vector 2: p3, H3, Q1 (indicating an environment of high temperature, humidity, and weak light).

[0046] In this embodiment, the environment sub-feature may be used to identify an environment having specific temperature, humidity, and lighting conditions.

[0047] In this embodiment, different category vectors of the fourth sub-feature are determined by different identification code permutations and combinations of all categories included in the worker sub-feature, the equipment sub-feature, and the environment sub-feature. For example, category vector 1: S2-Y1-T1-B1-M1-N1-L1-p2-H2-Q2, represents an intermediate welder with 1-3 years of work experience operating a mechanical equipment in a normal state in the morning shift. The equipment is located in production line A, and the environment is medium temperature, moderate humidity and normal light. Category vector 2: S3-Y3-T2-B2-M2-N2-L2-p3-H1-Q3, represents a senior electrician with more than 5 years of work experience operating an electronic equipment under maintenance in the evening shift. The equipment is located in production line B, and the environment is high temperature, dryness and strong light.

[0048] In this embodiment, worker, equipment, and environmental sub-features are combined to identify solutions that can achieve the best coordination under specific production conditions. For example, certain specific workers may show better efficiency or product quality when operating certain equipment in a specific environment. By quickly screening combinations of workers, equipment, and environments that meet specific conditions, workers with relevant skills and experience, available equipment, and a suitable production environment are automatically matched.

[0049] The beneficial effects of the above technical solution are as follows: determining the first sub-feature, the second sub-feature, the third sub-feature and the fourth sub-feature based on the second data can provide a data basis for determining the first comprehensive feature and the second comprehensive feature, comprehensively analyze all data related to work dispatching and production scheduling, and improve the feature expression capability.

[0050] Embodiment 4: The embodiment of the present invention provides a method for dispatching and scheduling work based on a decision tree, which determines a first comprehensive feature and a second comprehensive feature, including: Performing standardization processing on the first sub-feature, the second sub-feature, and the third sub-feature, and determining a first comprehensive feature based on the first sub-feature, the second sub-feature, and the third sub-feature after the standardization processing; The first sub-feature, the second sub-feature, and the fourth sub-feature are standardized, and a second comprehensive feature is determined based on the standardized first sub-feature, the second sub-feature, and the fourth sub-feature, wherein the second comprehensive feature includes a plurality of feature vectors.

[0051] In this embodiment, all features included in the first sub-feature, the second sub-feature, and the third sub-feature are standardized, all features included in the first sub-feature, the second sub-feature, and the third sub-feature are converted to the same scale or distribution, and all the standardized features are combined into a first comprehensive feature. For example, the first sub-feature includes 2 numerical data features and 2 composite features, the second sub-feature includes 3 features, and the third sub-feature includes 4 features. The features in the first sub-feature can be Z-score standardized, Min-Max standardized, etc., the features in the second sub-feature can be time-difference standardized, periodicity processed, etc., and the classification category of the third sub-feature can be One-Hot category encoded, Label category encoded, etc. The first comprehensive feature is determined based on the processed 4 features in the first sub-feature, 3 features in the second sub-feature, and 4 features in the third sub-feature, and the first comprehensive feature includes 11 standardized features.

[0052] In this embodiment, all features included in the first sub-feature and the second sub-feature are standardized, all features included in the first sub-feature and the second sub-feature are converted to the same scale or distribution, all features after the standardized processing are combined, and the combination results are respectively combined again with the fourth sub-feature including each category vector, and the second comprehensive feature is determined according to all the re-combined results. For example, the first sub-feature includes 3 numerical data features and 2 composite features, and the second sub-feature includes 3 features. The features in the first sub-feature can be Z-score standardized, Min-Max standardized, etc., and the features in the second sub-feature can be time-difference standardized, periodic processed, etc., and 7 standardized features are determined by combining 4 features in the first sub-feature and 3 features in the second sub-feature after processing, and the 7 standardized features and each category vector in the fourth sub-feature are combined again. Each combination result includes 7 standardized features and 1 category vector in the fourth sub-feature. The fourth sub-feature includes several category vectors, and the second comprehensive feature includes several feature vectors.

[0053] The beneficial effects of the above technical solution are as follows: determining the first comprehensive feature and the second comprehensive feature according to the first sub-feature, the second sub-feature, the third sub-feature and the fourth sub-feature can provide a data basis for determining the first feature and the second feature, and determine the correlation and mutual influence between different features, thereby improving the decision tree model's ability to understand and classify data.

[0054] Embodiment 5: An embodiment of the present invention provides a method for dispatching and scheduling work based on a decision tree, which determines a first feature based on second data and a first comprehensive feature, including: Determine a corresponding first order vector for each piece of historical order information in the second data according to the first comprehensive feature; Determine an order matrix based on the first order vector of all historical order information; Wherein, N1 represents the number of historical order information in the second data, and N2 represents the number of sub-features in the first comprehensive feature. , , represent the order value of the first sub-feature of the first historical order information in the second data, the feature value of the j-th sub-feature, and the feature value of the N2-th sub-feature in the first comprehensive feature, respectively. , , They respectively represent the order value of the 1st sub-feature, the feature value of the jth sub-feature, and the feature value of the N2th sub-feature of the i-th historical order information in the second data in the first comprehensive feature, , , Respectively represent the feature value of the 1st sub-feature, the feature value of the jth sub-feature, and the order value of the N2th sub-feature of the N1th historical order information in the second data in the first comprehensive feature; determining a first eigenvalue and a first eigenvector based on the order matrix; Where C is the covariance matrix, is the transposed matrix of the order matrix, is the kth first eigenvalue, The kth first eigenvector represented by ; Sort the first eigenvalues ​​from large to small, and calculate the importance rates of the first Nu first eigenvalues; in, , represents the importance rate of the first Nu eigenvalues ​​after sorting, represents the adjustment parameters, represents the pth first eigenvalue of the order matrix after sorting, represents the pth first eigenvalue of the order matrix after sorting based on the adjustment parameter, represents the cumulative importance of the first Nu first eigenvalues ​​after the order matrix is ​​sorted, represents the cumulative importance of all first eigenvalues ​​after the order matrix is ​​sorted, It represents the first importance rate of the first Nu first eigenvalues ​​after the order matrix is ​​sorted, represents the first weight of the first importance rate, represents the second importance rate of the first Nu first eigenvalues ​​after the order matrix is ​​sorted, a second weight representing a second importance rate; Extract all Nu values ​​with importance greater than 95%, and select the top values ​​after sorting. The first eigenvalues ​​construct the first projection matrix; The second data is projected into the first projection matrix to determine the first feature.

[0055] In this embodiment, each piece of historical order information in the second data is matched with the first comprehensive feature to determine the corresponding first order vector.

[0056] In this embodiment, the feature value of the jth sub-feature of the i-th historical order information in the first comprehensive feature in the second data represents the matching result of the i-th historical order information and the j-th sub-feature in the first comprehensive feature. If the feature expression of the j-th sub-feature is a numerical value, the feature value of the j-th sub-feature of the i-th historical order information in the first comprehensive feature is determined according to the numerical value corresponding to the j-th sub-feature and the i-th historical order information; if the feature expression of the j-th sub-feature is a classification, the feature value of the j-th sub-feature of the i-th historical order information in the first comprehensive feature is determined according to the identification code corresponding to the j-th sub-feature and the i-th historical order information.

[0057] In this embodiment, the order matrix M is an N2×N1 matrix, and the transposed matrix of the order matrix is is a N1×N2 matrix, and the covariance matrix is ​​a N2×N2 matrix.

[0058] In this embodiment, when α>1, the first importance ratio places more emphasis on large eigenvalues, when 0<α<1, the first importance ratio places more emphasis on small eigenvalues, and when α=1, the first importance ratio treats all eigenvalues ​​equally.

[0059] In this embodiment, each first eigenvalue corresponds to a first eigenvector.

[0060] In this embodiment, the importance rate of the first Nu eigenvalues ​​represents the importance of the first Nu first eigenvalues ​​after sorting among all the first eigenvalues.

[0061] In this embodiment, It represents the minimum Nu value with an importance rate greater than 95% among the first eigenvalues ​​after sorting, and the average number of features of the first comprehensive feature.

[0062] In this embodiment, the first projection matrix represents a matrix that projects the order matrix from the original feature space to the feature space after dimensionality reduction.

[0063] The beneficial effects of the above technical solution are as follows: determining the first feature based on the second data and the first comprehensive feature can provide a data basis for determining the dimensionality reduction feature, ensuring that the decision tree model can capture the key information most relevant to work dispatching and production scheduling, so as to generate a highly feasible work dispatching and production scheduling plan in real time.

[0064] Embodiment 6: The embodiment of the present invention provides a method for dispatching and scheduling work based on a decision tree, which determines a second feature based on second data and a second comprehensive feature, and determines a dimensionality reduction feature, including: Match all historical order information in the second data based on the second comprehensive feature, determine an order set for each feature vector, and a second order vector for each historical order information in the order set; Determining a second eigenvalue and a second eigenvector based on the order set of all eigenvectors; in, Represents the deviation value of the corresponding feature vector of all order sets, represents the deviation value between all eigenvectors, represents the bth second order vector in the order set of the ath feature vector, represents the second order fitting vector in the order set of the a-th eigenvector, The difference vector between the bth second order vector and the second order fitting vector in the order set of the ath feature vector, represents the transposed row vector of the difference vector between the bth second order vector and the mean of the second order vector in the order set of the ath eigenvector, represents the number of the second order vector in the order set of the a-th feature vector, represents the number of eigenvectors in the second comprehensive feature, The total fitted value of the second order vector representing the set of orders for all feature vectors in the second data, The difference vector between the second order fitting vector of the order set of the a-th feature vector and the total fitting value of the second order vector, The transposed row vector of the difference vector between the second order fitting vector and the total fitting value of the second order vector for the order set of the a-th eigenvector, is the cth second eigenvalue of, is the cth second eigenvector, ; Select the first The second eigenvalues ​​construct the second projection matrix; Projecting the second data into a second projection matrix to determine a second feature; Features that appear simultaneously in the first feature and the second feature are extracted to determine the dimension reduction features.

[0065] In this embodiment, each historical order information in the second data is matched with each feature vector in the second comprehensive feature, and the second order vector of each historical order information is assigned to the most matching feature vector. The order set of each feature vector is determined based on the second order vectors of all historical order information assigned to each feature vector. For example, each historical order information in the second data is matched with each feature vector in the second comprehensive feature, and the similarity or consistency between each historical order information and each feature vector in the second comprehensive feature can be calculated, and each historical order information is assigned to the feature vector with the highest similarity or consistency. If 5 historical order information are assigned to a feature vector in the second comprehensive feature, the order set of the feature vector includes these 5 historical order information.

[0066] In this embodiment, Represents the total product of the deviations of all second order vectors in the order set of the a-th eigenvector from the a-th eigenvector.

[0067] In this embodiment, the second order fitting vector may represent all order sets of the corresponding feature vector.

[0068] In this embodiment, the second order vector total fitting value may represent all order sets of all feature vectors.

[0069] In this embodiment, Represents the total product of the eigenvectors corresponding to the deviations of all second order vectors in the order set of all eigenvectors.

[0070] In this embodiment, Represents the total product of the second order fitted value of each eigenvector and the deviation from the total fitted value of the second order vector.

[0071] In this embodiment, the second projection matrix represents a matrix that projects all order sets from the original feature space to the feature space after dimensionality reduction.

[0072] In this embodiment, there are several features that appear in both the first feature and the second feature, and there are several features included in the dimensionality reduction feature.

[0073] The beneficial effects of the above technical solution are as follows: determining the second feature based on the second data and the second comprehensive feature, and determining the dimensionality reduction feature, can provide a data basis for building a decision tree model, reduce the complexity of the model, and improve the generalization ability and efficiency of the model, so as to realize the intelligent scheduling and production scheduling optimization of production tasks in a simple and efficient manner.

[0074] Embodiment 7: The embodiment of the present invention provides a method for dispatching and scheduling work based on a decision tree, which constructs a decision tree model based on second data and dimensionality reduction features, and determines and executes a dispatching and scheduling plan for real-time order information based on the decision tree model, including: Divide the second data into training set and test set, Build a decision tree model based on the dimensionality reduction features, historical order requirements of all historical order information in the training set, and historical execution plans; The historical order requirements of all historical order information in the test set are used as the input of the trained decision tree model, and the decision tree model outputs the decision execution plan corresponding to each historical order information; The model performance is evaluated based on the historical execution plan and decision execution plan of each historical order information, and the decision tree model is first optimized based on the evaluation results.

[0075] In this embodiment, the second data is divided into 80 / 20 or 70 / 30 ratios, 80% or 70% of the data is used for training, and the remaining 20% ​​or 30% is used for testing.

[0076] In this embodiment, based on the dimensionality reduction features, the historical order information, historical order requirements and historical execution plans in the training set are used as inputs of the decision tree model, and the decision tree model is trained using the training set data.

[0077] In this embodiment, all historical order information and historical order demands in the test set are input into a trained decision tree model, and the decision tree model outputs a prediction decision execution plan corresponding to each historical order information.

[0078] In this embodiment, the decision tree model automatically learns the causal relationship between features and historical execution plans through a training process. The model will construct splitting rules based on how each feature in the training set affects the final decision. For example, if a certain type of order demand is associated with a specific resource allocation plan (historical execution plan), the model will learn this association and refer to these features in future decisions to make similar allocation decisions.

[0079] In this embodiment, the decision execution plan predicted by the model is compared with the actual historical execution plan in the test set to evaluate the model performance. Methods such as accuracy, precision, recall, and F1-score can be used.

[0080] In this embodiment, according to the evaluation result, the decision tree model is first optimized. The first optimization may be adjusting model parameters, pruning processing, etc.

[0081] The beneficial effects of the above technical solution are as follows: by constructing a decision tree model based on the second data and dimensionality reduction features, a work dispatching and production scheduling plan can be generated in real time to adapt to changes in the production environment and demand, and intelligent scheduling and production scheduling optimization of production tasks can be realized simply and efficiently, thereby improving the decision-making quality of work dispatching and production scheduling, and improving the feasibility and execution efficiency of the work dispatching and production scheduling plan.

[0082] Embodiment 8: The embodiment of the present invention provides a method for dispatching and scheduling work based on a decision tree, which constructs a decision tree model based on the second data and the dimensionality reduction feature, determines and executes a dispatching and scheduling plan for real-time order information based on the decision tree model, and further includes: Input the real-time order information into the optimized decision tree model to generate a work dispatching and production scheduling plan for the real-time order information; Execute the work dispatching and production scheduling plan, evaluate the execution result of the work dispatching and production scheduling plan, and perform a second optimization on the decision tree model based on the execution result.

[0083] In this embodiment, new real-time order information is input into the optimized decision tree model, and the model generates a corresponding work dispatching and production scheduling plan based on the input real-time order information.

[0084] In this embodiment, the generated work dispatching and production scheduling plan is actually executed, and the execution effect of the work dispatching and production scheduling plan is evaluated, which may include the comparison between the actual completion status and the expected target, the execution time, the resource utilization efficiency and other aspects.

[0085] In this embodiment, based on the execution results, the decision tree model is further optimized, which may include retraining the model, improving the model structure, or continuously adjusting the model through a feedback mechanism to improve its adaptability to real-time orders and prediction accuracy.

[0086] The beneficial effects of the above technical solution are: determining and executing the dispatching and scheduling plan based on real-time order information according to the decision tree model, comprehensively analyzing all data related to dispatching and scheduling, improving data quality and consistency, generating dispatching and scheduling plans in real time to adapt to changes in the production environment and demand, simply and efficiently realizing intelligent scheduling and scheduling optimization of production tasks, improving the feasibility and execution efficiency of dispatching and scheduling plans, reducing production costs, shortening delivery cycles, and improving customer satisfaction.

[0087] The method embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative work.

[0088] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0089] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for dispatching work and scheduling production based on a decision tree, characterized in that: include: 101: collecting data related to work dispatching and production scheduling to determine first data, and performing first processing on the first data to determine second data; 102: Determine a first sub-feature, a second sub-feature, a third sub-feature, and a fourth sub-feature based on the second data, and determine a first comprehensive feature and a second comprehensive feature; 103: determining a first feature based on the second data and the first comprehensive feature, and determining a second feature based on the second data and the second comprehensive feature, and determining a dimensionality reduction feature; 104: Build a decision tree model based on the second data and the dimension reduction features, and determine and execute a work dispatching and production scheduling plan based on the real-time order information based on the decision tree model.

2. The method for dispatching work and scheduling production based on a decision tree according to claim 1, characterized in that: The first data includes a plurality of historical order information; The first processing includes data cleaning, missing value processing and data normalization on the first data.

3. The method for dispatching work and scheduling production based on a decision tree according to claim 1, characterized in that: Determining the first sub-feature, the second sub-feature, the third sub-feature, and the fourth sub-feature based on the second data includes: Extracting data related to the value in the second data to determine first sub-data, wherein the first sub-data includes a plurality of value categories related to the value in the second data; Performing statistical analysis on each numerical category in the first sub-data respectively, and determining the first sub-feature based on the statistical analysis results of all numerical categories; Extracting time-related data from the second data to determine second sub-data, wherein the second sub-data includes a plurality of time-related time series categories from the second data; Performing time series analysis on each time series category in the second sub-data respectively, and determining the second sub-feature based on the time series analysis results of all time series categories; Extract all the remaining data except the first sub-data and the second sub-data from the second data to determine the third sub-data, wherein the third sub-data includes a plurality of classification categories; determining a third sub-feature based on each category in the third sub-data; Encode all identifiers under each classification category in the third sub-data respectively, interactively determine worker sub-features for all identifier codes under all classification categories related to workers, interactively determine equipment sub-features for all identifier codes under all classification categories related to equipment, and interactively determine environment sub-features for all identifier codes under all classification categories related to the environment, wherein the worker sub-features include multiple worker vectors, the equipment sub-features include multiple equipment vectors, and the environment sub-features include multiple environment vectors; All identification codes under the worker sub-feature, the equipment sub-feature, and the environment sub-feature are arranged and combined to determine a fourth sub-feature, wherein the fourth sub-feature includes a plurality of category vectors.

4. The method for dispatching work and scheduling production based on a decision tree according to claim 3, characterized in that: Determining the first comprehensive feature and the second comprehensive feature includes: Performing standardization processing on the first sub-feature, the second sub-feature, and the third sub-feature, and determining a first comprehensive feature based on the first sub-feature, the second sub-feature, and the third sub-feature after the standardization processing; The first sub-feature, the second sub-feature, and the fourth sub-feature are standardized, and a second comprehensive feature is determined based on the standardized first sub-feature, the second sub-feature, and the fourth sub-feature, wherein the second comprehensive feature includes a plurality of feature vectors.

5. The method for dispatching work and scheduling production based on a decision tree according to claim 1, characterized in that: Determining the first feature based on the second data and the first comprehensive feature includes: Determine a corresponding first order vector for each piece of historical order information in the second data according to the first comprehensive feature; Determine an order matrix based on the first order vector of all historical order information; Wherein, N1 represents the number of historical order information in the second data, and N2 represents the number of sub-features in the first comprehensive feature. , , represent the order value of the first sub-feature of the first historical order information in the second data, the feature value of the j-th sub-feature, and the feature value of the N2-th sub-feature in the first comprehensive feature, respectively. , , They respectively represent the order value of the 1st sub-feature, the feature value of the jth sub-feature, and the feature value of the N2th sub-feature of the i-th historical order information in the second data in the first comprehensive feature, , , Respectively represent the feature value of the 1st sub-feature, the feature value of the jth sub-feature, and the order value of the N2th sub-feature of the N1th historical order information in the second data in the first comprehensive feature; determining a first eigenvalue and a first eigenvector based on the order matrix; Where C is the covariance matrix, is the transposed matrix of the order matrix, is the kth first eigenvalue, The kth first eigenvector represented by ; Sort the first eigenvalues ​​from large to small, and calculate the importance rates of the first Nu first eigenvalues; in, , represents the importance rate of the first Nu eigenvalues ​​after sorting, represents the adjustment parameters, represents the pth first eigenvalue of the order matrix after sorting, represents the pth first eigenvalue of the order matrix after sorting based on the adjustment parameter, represents the cumulative importance of the first Nu first eigenvalues ​​after the order matrix is ​​sorted, represents the cumulative importance of all first eigenvalues ​​after the order matrix is ​​sorted, It represents the first importance rate of the first Nu first eigenvalues ​​after the order matrix is ​​sorted, represents the first weight of the first importance rate, represents the second importance rate of the first Nu first eigenvalues ​​after the order matrix is ​​sorted, a second weight representing a second importance rate; Extract all Nu values ​​with importance greater than 95%, and select the top values ​​after sorting. The first eigenvalues ​​construct the first projection matrix; The second data is projected into the first projection matrix to determine the first feature.

6. The method for dispatching work and scheduling production based on a decision tree according to claim 4, characterized in that: Determining the second feature based on the second data and the second comprehensive feature, and determining the dimension reduction feature, including: Match all historical order information in the second data based on the second comprehensive feature, determine an order set for each feature vector, and a second order vector for each historical order information in the order set; Determining a second eigenvalue and a second eigenvector based on the order set of all eigenvectors; in, Represents the deviation value of the corresponding feature vector of all order sets, represents the deviation value between all eigenvectors, represents the bth second order vector in the order set of the ath feature vector, represents the second order fitting vector in the order set of the a-th eigenvector, The difference vector between the bth second order vector and the second order fitting vector in the order set of the ath feature vector, represents the transposed row vector of the difference vector between the bth second order vector and the mean of the second order vector in the order set of the ath eigenvector, represents the number of the second order vector in the order set of the a-th feature vector, represents the number of eigenvectors in the second comprehensive feature, The total fitted value of the second order vector representing the set of orders for all feature vectors in the second data, The difference vector between the second order fitting vector of the order set of the a-th feature vector and the total fitting value of the second order vector, The transposed row vector of the difference vector between the second order fitting vector and the total fitting value of the second order vector for the order set of the a-th eigenvector, is the cth second eigenvalue of, is the cth second eigenvector, ; Select the first The second eigenvalues ​​construct the second projection matrix; Projecting the second data into a second projection matrix to determine a second feature; Features that appear simultaneously in the first feature and the second feature are extracted to determine the dimension reduction features.

7. The method for dispatching work and scheduling production based on a decision tree according to claim 1, characterized in that: A decision tree model is constructed based on the second data and the dimension reduction features, and a dispatching and production scheduling plan for the real-time order information is determined and executed based on the decision tree model, including: Divide the second data into training set and test set, Build a decision tree model based on the dimensionality reduction features, historical order requirements of all historical order information in the training set, and historical execution plans; The historical order requirements of all historical order information in the test set are used as the input of the trained decision tree model, and the decision tree model outputs the decision execution plan corresponding to each historical order information; The model performance is evaluated based on the historical execution plan and decision execution plan of each historical order information, and the decision tree model is first optimized based on the evaluation results.

8. The method for dispatching work and scheduling production based on a decision tree according to claim 7, characterized in that: Building a decision tree model based on the second data and the dimension reduction features, determining and executing a dispatching and production scheduling plan for the real-time order information based on the decision tree model, and further comprising: Input the real-time order information into the optimized decision tree model to generate a work dispatching and production scheduling plan for the real-time order information; Execute the work dispatching and production scheduling plan, evaluate the execution result of the work dispatching and production scheduling plan, and perform a second optimization on the decision tree model based on the execution result.

Citation Information

Patent Citations

  • Improved decision tree classification algorithm for achieving search engine optimization technology

    CN106909626A

  • Recommendation method adopting deep characteristic matching

    CN106933996A

  • Production scheduling decision-making method and system

    CN113159979A

  • Trace value ratio linear discriminant analysis dimensionality reduction method and device for attribute graph adjustment and medium

    CN115358317A

  • Order scheduling optimization method based on improved crow search algorithm

    CN115629584A

Cited By

  • AI-driven cubilose precise nutrition formula design and high-value processing method and system

    CN120183618A

  • AI-driven bird's nest precise nutritional formula design and high-value processing methods and systems

    CN120183618B

  • Emergency dispatching and production scheduling method and device for refinery plant production

    CN121882495A

  • Emergency scheduling and production scheduling method and device for refinery production

    CN121882495B