A digital management system and method based on artificial intelligence

By introducing artificial intelligence technology into the digital management system, collecting and analyzing data, predicting demands and optimizing resource allocation, the problems of slow processing speed and lack of intelligence in the existing system are solved, and efficient and personalized management services are realized, which improves the The operational efficiency and resource utilization rate of the enterprise.

CN116663842BActive Publication Date: 2024-12-06BEIJING UNITED MEDIA TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202310709654.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-06-15
Publication Date
2024-12-06
Estimated Expiration
2043-06-15

AI Technical Summary

Technical Problem

The existing digital management system has slow processing speed, large storage space, high hardware requirements, lacks flexibility and intelligence, cannot adapt to changes in different enterprises and markets, and cannot provide personalized management services.

Method used

A digital management system based on artificial intelligence was designed, including data acquisition module, demand prediction module, neural network analysis module and resource management module. By collecting and analyzing data, predicting demands, optimizing resource configuration, the intelligent allocation and management of resources are realized. .

Benefits of technology

It realizes rapid analysis and processing of a variety of data, provides customized management solutions, improves the enterprise's operating management efficiency and resource utilization rate, and reduces waste and losses caused by changes in demand.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116663842B_ABST
    Figure CN116663842B_ABST
Patent Text Reader

Abstract

The present invention discloses a digital management system and method based on artificial intelligence, belonging to the field of digital management technology. The system includes a data acquisition module, a demand forecasting module, a neural network analysis module, a resource management module and a database; the data acquisition module is used to collect data related to demand; the demand forecasting module is used to predict the time node when future demand will appear; the neural network analysis module is used to allocate resources to demand when demand appears; the resource management module is used to use the forecasting module to find out the time period when demand does not appear in the time series and stop providing resources for the demand, while providing resources for new demand; the database is used to store the collected data related to demand. The present invention achieves flexible management of data by establishing a forecasting model to predict the time node when data appears, and using artificial intelligence to optimize the configuration of resources for these time nodes.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field, and in particular to a digital management system and method based on artificial intelligence. Background Art

[0002] With the rapid development of globalization and informatization, enterprises are facing increasingly fierce competition, and enterprise management has become more and more complex and challenging. In the prior art, digital management systems and methods mainly use traditional data warehouse and data mining technologies, which have slow processing speeds, require large storage space, and have high requirements for computer hardware equipment. Secondly, these systems and methods are mainly based on manually set rules and models for data analysis and processing, lacking flexibility and intelligence, unable to adapt to the changes and needs of different enterprises and markets, and unable to provide personalized management services. Compared with the prior art, artificial intelligence technology realizes the rapid analysis and mining of large amounts of data, can provide customized management solutions, and enable enterprises to achieve scientific, efficient, and intelligent management.

[0003] Therefore, what is needed now is an artificial intelligence-based digital management system and method that can simultaneously process multiple different types of data and customize corresponding management strategies. Summary of the invention

[0004] The purpose of the present invention is to provide a digital management system and method based on artificial intelligence to solve the problems raised in the above background technology.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: a digital management system based on artificial intelligence, the system includes a data acquisition module, a demand forecasting module, a neural network analysis module, a resource management module and a database;

[0006] The data acquisition module is used to collect data related to demand; the demand prediction module is used to predict the time nodes when future demand will appear; the neural network analysis module is used to allocate resources to demand when demand appears; the resource management module is used to use the prediction module to find out the time period when demand does not appear in the time series and stop providing resources for the demand, while providing resources for new demand; the database is used to store the collected data related to demand.

[0007] The data collection module includes a demand type collection unit and a demand counting unit; the demand type collection unit is used to perform statistics on a number of demands. For example, it can extract keywords from work orders, complaints or suggestions submitted by users, and automatically classify them into different demand categories, such as functional requirements, bug feedback, improvement suggestions, etc., and then count the number and proportion of each demand category; the demand counting unit is used to record the number of times a demand occurs and the duration of the demand in a time period, so as to understand the development trend of each demand category in different time periods, such as which demand types have received more attention in the most recent quarter, and the duration of each demand event. It is also convenient for us to count and predict the changing trends and outbreak points of demand in the next few time periods, so as to prepare countermeasures in advance.

[0008] The demand forecasting module includes a demand analysis unit and a forecast result output unit; the demand analysis unit is used to analyze historical demand data based on the information in the data acquisition module, including demand types, quantities, durations, and times of occurrence. The results of the analysis can be used to find trends and patterns in demand and understand the periodicity, alternation, rhythm, and other characteristics of demand; the forecast result output unit is used to output the time nodes of demand occurrence, and can predict future demand based on the patterns and trends in historical data, combined with machine learning, artificial intelligence, and other technologies. The forecast results can be specific demand quantities or the probability of demand occurrence. These forecast results can help us make better decisions and plans, thereby meeting user and market needs more efficiently.

[0009] The neural network analysis module includes a demand feature input module, a model training unit and an analysis result output unit; the demand feature input module is used to digitally input the relevant features of the demand so that the neural network model can understand and process them. In some cases, these features can be the number of predicted demands or the probability of occurrence, or other demand-related features, such as user attributes, time series data, market trends, etc. By digitizing these demand features, we can convert them into a processable format; the model training unit is used to train the goal of resource allocation. The unit uses a neural network algorithm to learn and train the demand features so that the neural network can learn some inherent laws from historical demand data to better predict future demand. During the training process, the model needs to be continuously adjusted and optimized to improve the accuracy and efficiency of the prediction; the analysis result output unit is used to output the analysis results of the neural network model, which will output the prediction results processed by the neural network model to other modules for further decision-making and planning. The predicted result may be the number of demands that need to be met, or the number of resources that need to be allocated, and may also include information such as resource allocation plans. These analysis results can optimize resource utilization and improve the efficiency of the enterprise under the premise of ensuring quality.

[0010] The resource management module includes a demand replacement unit and a resource configuration unit; the demand replacement unit is used to delete non-existent demands and add new demands according to the results of the demand prediction module; the resource configuration unit is used to perform resource configuration for the newly added demands. The demand replacement unit is usually closely connected with the demand prediction module. The demand prediction module predicts future demands, and the demand replacement unit determines whether the existing demands are still necessary based on the prediction results. If not, the demands are deleted and new demands are added to keep the demands up to date and accurate. The demand replacement unit can also provide the resource configuration unit with information on deleted and newly added demands so that the resource configuration unit can perform the next resource configuration. The resource configuration unit performs resource configuration for the newly added demands based on the information provided by the demand replacement unit. Typical resources include manpower, equipment, time, etc. The resource configuration unit can use various algorithms, such as greedy algorithms, dynamic programming, etc., to allocate the required resources. The goal of resource configuration is to ensure that each demand can be guaranteed with appropriate resources, thereby improving product quality and customer satisfaction.

[0011] The database includes a data storage unit and a data retrieval unit; the data storage unit is used to store demand data, resource data, forecast data and statistical data; the data retrieval unit is used to provide information retrieval for data analysis, and the data retrieval unit can extract corresponding data from the database according to different needs, and pass the data to different analysis modules. For example, when the demand forecasting module needs to be trained, the data retrieval unit can provide historical data to the training module for training. When the analysis result output unit needs to output the analysis result, the data retrieval unit can provide the required data to the result output unit. Therefore, the data retrieval unit plays a very important role in the entire data analysis process, and the effective implementation of its function can improve the speed and accuracy of data analysis and processing.

[0012] A digital management method based on artificial intelligence, characterized by:

[0013] S1. Collect data related to demand;

[0014] S2. Based on the collected data, a mathematical model is established to predict the distribution of future demand;

[0015] S3. Immediately delete resource configurations that do not have demand based on demand distribution;

[0016] S4. On the basis of deleting the resource configurations that do not have demand, if there is no new demand, no new resource configuration will be made. If there is a new demand, the neural network model will be used to configure appropriate resources for it.

[0017] In S1, analyzing the distribution of demand in time series is the premise of realizing digital management. First, collect data related to demand:

[0018]

[0019] in, Indicates the type of demand. Indicates the number of types of demand, express The number of requirements, Indicates the time when the demand occurs. Indicates the number of times the demand occurs. Indicates The moment of need, represents the duration of the demand, Indicates the duration of the demand. Indicates The duration of the demand segment;

[0020] In S2, the following steps are included:

[0021] S201. After collecting the demand-related data, analyze the distribution pattern of past demand in time series and understand the distribution pattern of demand in time series. The demand sequence that can be observed at the time, with the time span set to 1, represents the distribution of demand at a point in time. To express, , analyze the law of demand in time series, that is, analyze the trend of demand appearance and disappearance, so the following time series mathematical model is established:

[0022]

[0023] in, Indicates at time The value of represents a constant, represents the residual, represents the autoregressive term coefficient, represents the coefficient of the moving average term, represents the order of the autoregressive term, represents the order of the moving average term;

[0024] The model collects all demand data from the forecast time point forward. Then, the data is processed and analyzed using appropriate algorithms and methods to derive the patterns of demand appearance and disappearance. Finally, based on these patterns, the distribution of demand in the future time series is further predicted.

[0025] S202. Further forecast the demand by establishing a future demand forecasting model. The model is as follows:

[0026]

[0027] in, Represents the predicted future time The value on represents a constant, represents the residual, represents the autoregressive term coefficient, represents the coefficient of the moving average term, represents the order of the autoregressive term, represents the order of the moving average term;

[0028] In this model, represents a point in time in the future, and represents all past demand data, that is, the model is based on the demand distribution law analyzed by the time series mathematical model to obtain the distribution of future demand. Since the dimension of the demand distribution described by the future demand forecasting model is different from that of the time series mathematical model, the forecast result is further restored to the dimension of the original demand distribution to obtain the final forecast result. By establishing this forecasting model, enterprises can more accurately predict future demand trends, thereby allocating and planning resources more effectively. For example, when it is predicted that demand will rise sharply, enterprises can increase the corresponding resource allocation in advance to meet future demand. At the same time, the model can also reduce the waste and loss of resources caused by changes in demand, and improve the production efficiency and economic benefits of enterprises.

[0029] S203. The distribution of future demand can be obtained through the future demand forecasting model. However, since the focus is on predicting the trend of demand generation and disappearance, the distribution of demand will be inaccurate for some unpredictable emergencies. Considering that the accuracy of the forecast value plays a vital role in resource management and allocation, it is necessary to correct the predicted demand distribution and establish a forecast value error correction model:

[0030]

[0031] in, Represents the predicted value at a future time point, Represents data at a historical point in time. Represents the weighting coefficient, which can be adjusted as needed;

[0032] For some unpredictable emergencies, the accuracy of the forecast value will indeed be affected to a certain extent, which may cause errors in the predicted demand distribution. Therefore, enterprises need to establish a forecast value correction model to correct the forecast error. The process of establishing a forecast value correction model needs to rely on real-time demand data and forecast results, as well as past historical data. First, through real-time data and forecast results, determine the scope and trend of the forecast error, that is, the difference between the forecast value and the actual value. Secondly, use historical data to analyze the source and law of the forecast error in order to further optimize the forecast model. For example, for some emergencies, relevant forecast models can be established in advance by analyzing their characteristics and impacts. Finally, according to the analysis results, take appropriate measures to correct the forecast value. For example, you can adjust the resource allocation plan, postpone or advance product production, etc. to adapt to actual demand changes and reduce the impact of forecast errors.

[0033] By establishing a forecast value correction model, we can more accurately grasp the actual demand situation and more effectively manage and plan resources. At the same time, the model can help enterprises cope with the impact of emergencies, make up for the shortcomings of forecast analysis, and improve the production efficiency and innovation ability of enterprises.

[0034] S204: Use the prediction value error correction model to correct the output of the future demand prediction model, and construct the distribution status of demand at future time nodes.

[0035] In S3, the resource configuration without demand is deleted immediately according to the demand distribution. The demand distribution is analyzed by the future demand prediction model to establish the following resource discrimination model:

[0036]

[0037] in, Indicates time The sum of all the demand quantities that exist at Indicates the demand quantity, Indicates at time When the demand quantity The result of the assignment;

[0038] According to the resource discrimination model, a discrimination mechanism for resource allocation regarding the generation and disappearance of demand is generated: First, collect the time series The maximum value of When, with The process from 1 to the maximum value, for , that is, the demand quantity is 1, and the actual situation is analyzed. The set value matches the condition. If the actual required quantity is indeed 1, then ,otherwise ,along with The increase is calculated in this way The value of After the value, each The value corresponds to Add up all the values ​​of ,like , indicating that there is no demand at this moment, so the resource configuration can be deleted completely. If, Resources need to be allocated differently according to the different demand quantities.

[0039] This discrimination mechanism based on the resource discrimination model can help enterprises understand the actual demand more accurately so as to configure resources more accurately. By collecting the maximum value in the time series and analyzing the matching status between the actual demand quantity and the set value, the generation and disappearance of demand can be judged more accurately. After traversing all the values, if it is found that there is no demand at a certain moment, the resource configuration at that moment can be deleted to save resources. If there is demand, different resource allocations can be made according to the different quantities of demand to meet the actual needs.

[0040] In S4, the following steps are included:

[0041] S401. When new demands appear, they cannot be generalized due to differences between them, and characteristic information of the demands needs to be collected;

[0042] S402: Digitize the characteristics of the demand as the input vector of the neural network model. The neural network is used to customize resource allocation for the emerging demand. Therefore, the following neural network model is established:

[0043]

[0044] in, is represented as a feature vector, is the output value of the hidden layer, is the weight, is the bias, is the output of the previous layer, is the output value of the output layer, is the weight of the output layer, is the bias of the output layer, is the output of the last hidden layer, is the loss function, is the number of training samples;

[0045] After the neural network model is established, the neural network model is trained according to the restrictions of the input features, and the output value in the neural network is calculated for each input feature vector. , further, after obtaining the output value in the neural network, calculate the loss function value Gradient ; and for each hidden layer Node, calculate the gradient of the current node , and calculate the gradient of its input weights and biases by the chain rule and ;

[0046] S404, after calculating the data in the neural network model, the data is applied. First, the output value of the neural network is calculated by providing an input vector; further, the output value is compared with the true value to calculate the value of the loss function; next, the gradient of the loss function is used to adjust the network parameters to minimize the loss function; specifically, the gradient is calculated starting from the output layer in a reverse manner, and the error signal is propagated from the output layer to all nodes of the hidden layer and the input layer to optimize all parameters;

[0047] S405. By executing the process of adjusting the network parameters, the parameters are further updated. According to the calculated gradient, each parameter is updated. Through continuous iteration, the parameters of the network are optimized so that the effect of the demand resource configuration reaches the best state.

[0048] This process is actually a model training process. By constantly adjusting parameters and trying different parameter combinations, the model can more accurately predict the demand change trend and distribution of demand resources. The trained and optimized neural network model can maximize the use of limited resources while meeting demand, thereby improving resource utilization and efficiency. For example, arranging the quantity and time period of resources on the production line, adjusting the resource allocation of the supply chain, and reasonably arranging the working hours of personnel can all be optimized and decided through the trained model. In addition, through reasonable demand resource allocation and optimization, the production efficiency and economic benefits of the enterprise can be improved, production costs can be reduced, and product quality and market competitiveness can be improved. Therefore, in the process of enterprise operation and management, using neural network models to predict and configure demand resources is a very effective means and strategy.

[0049] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0050] 1. Based on factors such as the time node and duration of demand, a mathematical model is established, and the mathematical model is used to predict the distribution of future demand. This is conducive to taking measures in advance when responding to demand, alleviating the pressure of digital management to a certain extent, and helping managers to formulate corresponding management strategies.

[0051] 2. According to the demand forecasting model, old demands disappear and new demands arise at different time points. Therefore, a discrimination mechanism is established for this situation. The resource allocation of old demands is deleted using this discrimination mechanism, and artificial intelligence technology is used to customize appropriate resource allocation for new demands. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0053] Figure 1 It is a schematic diagram of the module structure of a digital management system and method based on artificial intelligence of the present invention;

[0054] Figure 2 It is a method flow structure diagram of a digital management system and method based on artificial intelligence of the present invention. DETAILED DESCRIPTION

[0055] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] See also Figure 1-2 ,The present invention provides a technical solution: a digital management system based on artificial intelligence, the system includes a data acquisition module, a demand forecasting module, a neural network analysis module, a resource management module and a database;

[0057] The data collection module is used to collect data related to demand; the demand forecasting module is used to predict the time nodes when future demand will appear; the neural network analysis module is used to allocate resources to demand when it appears; the resource management module is used to use the forecasting module to find out the time period when demand does not appear in the time series and stop providing resources for the demand, while providing resources for new demand; the database is used to store the collected data related to demand.

[0058] The data collection module includes a demand type collection unit and a demand counting unit; the demand type collection unit is used to collect statistics for a number of demands. For example, it can extract keywords from work orders, complaints or suggestions submitted by users, and automatically classify them into different demand categories, such as functional requirements, bug feedback, improvement suggestions, etc., and then count the number and proportion of each demand category; the demand counting unit is used to record the number of times a demand occurs and the duration of the demand in a time period. It can understand the development trend of each demand category in different time periods, such as which demand types have received more attention in the most recent quarter, and the duration of each demand event. It also makes it easier for us to count and predict the changing trends and outbreak points of demand in the next few time periods, so that we can prepare countermeasures in advance.

[0059] The demand forecasting module includes a demand analysis unit and a forecast result output unit; the demand analysis unit is used to analyze historical demand data based on the information in the data acquisition module, including demand types, quantities, durations, and times of occurrence. The results of the analysis can be used to find trends and patterns in demand and understand the periodicity, alternation, and rhythm of demand; the forecast result output unit is used to output the time nodes of demand occurrence, and can predict future demand based on the patterns and trends in historical data combined with machine learning, artificial intelligence and other technologies. The forecast results can be specific demand quantities or the probability of demand occurrence. These forecast results can help us make better decisions and plans, thereby meeting user and market needs more efficiently.

[0060] The neural network analysis module includes a demand feature input module, a model training unit, and an analysis result output unit; the demand feature input module is used to digitally input the relevant features of the demand so that the neural network model can understand and process them. In some cases, these features can be the number of predicted demands or the probability of occurrence, or other demand-related features, such as user attributes, time series data, market trends, etc. By digitizing these demand features, we can convert them into a processable format; the model training unit is used to train the goal of resource allocation. The unit uses a neural network algorithm to learn and train demand features so that the neural network can learn some inherent laws from historical demand data to better predict future demand. During the training process, the model needs to be continuously adjusted and optimized to improve the accuracy and efficiency of the prediction; the analysis result output unit is used to output the analysis results of the neural network model. It will output the prediction results processed by the neural network model to other modules for further decision-making and planning. The prediction results may be the number of demands that need to be met, or the number of resources that need to be allocated, and may also include information such as resource allocation plans. These analysis results can optimize resource utilization and improve the efficiency of the enterprise under the premise of ensuring quality.

[0061] The resource management module includes a demand replacement unit and a resource configuration unit; the demand replacement unit is used to delete non-existent demands and add new demands according to the results of the demand forecasting module; the resource configuration unit is used to configure resources for the newly added demands. The demand replacement unit is usually closely connected with the demand forecasting module. The demand forecasting module predicts future demands, and the demand replacement unit determines whether the existing demands are still necessary based on the forecast results. If not, the demand is deleted and a new demand is added to keep the demand up to date and accurate. The demand replacement unit can also provide the resource configuration unit with information on deleted and newly added demands so that the resource configuration unit can perform the next resource configuration. The resource configuration unit configures resources for the newly added demands based on the information provided by the demand replacement unit. Typical resources include manpower, equipment, time, etc. The resource configuration unit can use various algorithms, such as greedy algorithms, dynamic programming, etc., to allocate the required resources. The goal of resource configuration is to ensure that each demand can be guaranteed with an appropriate amount of resources, thereby improving product quality and customer satisfaction.

[0062] The database includes a data storage unit and a data retrieval unit; the data storage unit is used to store demand data, resource data, forecast data and statistical data; the data retrieval unit is used to provide information retrieval for data analysis. The data retrieval unit can extract corresponding data from the database according to different needs and pass the data to different analysis modules. For example, when the demand forecasting module needs to be trained, the data retrieval unit can provide historical data to the training module for training. When the analysis result output unit needs to output the analysis result, the data retrieval unit can provide the required data to the result output unit. Therefore, the data retrieval unit plays a very important role in the entire data analysis process, and the effective implementation of its functions can improve the speed and accuracy of data analysis and processing.

[0063] A digital management method based on artificial intelligence, characterized by:

[0064] S1. Collect data related to demand;

[0065] S2. Based on the collected data, a mathematical model is established to predict the distribution of future demand;

[0066] S3. Immediately delete resource configurations that do not have demand based on demand distribution;

[0067] S4. On the basis of deleting the resource configurations that do not have demand, if there is no new demand, no new resource configuration will be made. If there is a new demand, the neural network model will be used to configure appropriate resources for it.

[0068] In S1, analyzing the distribution of demand in time series is the premise of realizing digital management. First, collect data related to demand:

[0069]

[0070] in, Indicates the type of demand. Indicates the number of types of demand, express The number of requirements, Indicates the time when the demand occurs. Indicates the number of times the demand occurs. Indicates The moment of need, represents the duration of the demand, Indicates the duration of the demand. Indicates The duration of the demand segment;

[0071] In S2, the following steps are included:

[0072] S201. After collecting the demand-related data, analyze the distribution pattern of past demand in time series and understand the distribution pattern of demand in time series. The demand sequence that can be observed at the time, with the time span set to 1, represents the distribution of demand at a point in time. To express, , analyze the law of demand in time series, that is, analyze the trend of demand appearance and disappearance, so the following time series mathematical model is established:

[0073]

[0074] in, Indicates at time The value of represents a constant, represents the residual, represents the autoregressive term coefficient, represents the coefficient of the moving average term, represents the order of the autoregressive term, represents the order of the moving average term;

[0075] The model collects all demand data from the forecast time point forward. Then, the data is processed and analyzed using appropriate algorithms and methods to derive the patterns of demand appearance and disappearance. Finally, based on these patterns, the distribution of demand in the future time series is further predicted.

[0076] S202. Further forecast the demand by establishing a future demand forecasting model. The model is as follows:

[0077]

[0078] in, Represents the predicted future time The value on represents a constant, represents the residual, represents the autoregressive term coefficient, represents the coefficient of the moving average term, represents the order of the autoregressive term, represents the order of the moving average term;

[0079] In this model, represents a point in time in the future, and represents all past demand data, that is, the model is based on the demand distribution law analyzed by the time series mathematical model to obtain the distribution of future demand. Since the dimension of the demand distribution described by the future demand forecasting model is different from that of the time series mathematical model, the forecast result is further restored to the dimension of the original demand distribution to obtain the final forecast result. By establishing this forecasting model, enterprises can more accurately predict future demand trends, thereby allocating and planning resources more effectively. For example, when it is predicted that demand will rise sharply, enterprises can increase the corresponding resource allocation in advance to meet future demand. At the same time, the model can also reduce the waste and loss of resources caused by changes in demand, and improve the production efficiency and economic benefits of enterprises.

[0080] S203. The distribution of future demand can be obtained through the future demand forecasting model. However, since the focus is on predicting the trend of demand generation and disappearance, the distribution of demand will be inaccurate for some unpredictable emergencies. Considering that the accuracy of the forecast value plays a vital role in resource management and allocation, it is necessary to correct the predicted demand distribution and establish a forecast value error correction model:

[0081]

[0082] in, Represents the predicted value at a future time point, Represents data at a historical point in time. Represents the weighting coefficient, which can be adjusted as needed;

[0083] For some unpredictable emergencies, the accuracy of the forecast value will indeed be affected to a certain extent, which may cause errors in the predicted demand distribution. Therefore, enterprises need to establish a forecast value correction model to correct the forecast error. The process of establishing a forecast value correction model needs to rely on real-time demand data and forecast results, as well as past historical data. First, through real-time data and forecast results, determine the scope and trend of the forecast error, that is, the difference between the forecast value and the actual value. Secondly, use historical data to analyze the source and law of the forecast error in order to further optimize the forecast model. For example, for some emergencies, relevant forecast models can be established in advance by analyzing their characteristics and impacts. Finally, according to the analysis results, take appropriate measures to correct the forecast value. For example, you can adjust the resource allocation plan, postpone or advance product production, etc. to adapt to actual demand changes and reduce the impact of forecast errors.

[0084] By establishing a forecast value correction model, we can more accurately grasp the actual demand situation and more effectively manage and plan resources. At the same time, the model can help enterprises cope with the impact of emergencies, make up for the shortcomings of forecast analysis, and improve the production efficiency and innovation ability of enterprises.

[0085] S204: Use the prediction value error correction model to correct the output of the future demand prediction model, and construct the distribution status of demand at future time nodes.

[0086] In S3, the resource configuration without demand is deleted immediately according to the demand distribution. The demand distribution is analyzed by the future demand prediction model to establish the following resource discrimination model:

[0087]

[0088] in, Indicates time The sum of all the demand quantities that exist at Indicates the demand quantity, Indicates at time When the demand quantity The result of the assignment;

[0089] According to the resource discrimination model, a discrimination mechanism for resource allocation regarding the generation and disappearance of demand is generated: First, collect the time series The maximum value of When, with The process from 1 to the maximum value, for , that is, the demand quantity is 1, and the actual situation is analyzed. The set value matches the condition. If the actual required quantity is indeed 1, then ,otherwise ,along with The increase is calculated in this way The value of After the value, each The value corresponds to Add up all the values ​​of ,like , indicating that there is no demand at this moment, so the resource configuration can be deleted completely. If, Resources need to be allocated differently according to the different demand quantities.

[0090] This discrimination mechanism based on the resource discrimination model can help enterprises understand the actual demand more accurately so as to configure resources more accurately. By collecting the maximum value in the time series and analyzing the matching status between the actual demand quantity and the set value, the generation and disappearance of demand can be judged more accurately. After traversing all the values, if it is found that there is no demand at a certain moment, the resource configuration at that moment can be deleted to save resources. If there is demand, different resource allocations can be made according to the different quantities of demand to meet the actual needs.

[0091] In S4, the following steps are included:

[0092] S401. When new demands appear, they cannot be generalized due to differences between them, and characteristic information of the demands needs to be collected;

[0093] S402: Digitize the characteristics of the demand as the input vector of the neural network model. The neural network is used to customize resource allocation for the emerging demand. Therefore, the following neural network model is established:

[0094]

[0095] in, is represented as a feature vector, is the output value of the hidden layer, is the weight, is the bias, is the output of the previous layer, is the output value of the output layer, is the weight of the output layer, is the bias of the output layer, The output of the last hidden layer, is the loss function, is the number of training samples;

[0096] After the neural network model is established, the neural network model is trained according to the restrictions of the input features, and the output value in the neural network is calculated for each input feature vector. , further, after obtaining the output value in the neural network, calculate the loss function value Gradient ; and for each hidden layer Node, calculate the gradient of the current node , and calculate the gradient of its input weights and biases by the chain rule and ;

[0097] S404, after calculating the data in the neural network model, the data is applied. First, the output value of the neural network is calculated by providing an input vector; further, the output value is compared with the true value to calculate the value of the loss function; next, the gradient of the loss function is used to adjust the network parameters to minimize the loss function; specifically, the gradient is calculated starting from the output layer in a reverse manner, and the error signal is propagated from the output layer to all nodes of the hidden layer and the input layer to optimize all parameters;

[0098] S405, by executing the process of adjusting network parameters, further updating parameters, updating each parameter according to the calculated gradient, optimizing the parameters of the network through continuous iteration, so that the effect of demand resource configuration reaches the best state. Such a process is actually a model training process. By constantly adjusting parameters and constantly trying different parameter combinations, the model can more accurately predict the demand change trend and distribution status of demand resources. The trained and optimized neural network model can maximize the use of limited resources on the premise of meeting demand, thereby improving resource utilization and efficiency. For example, arranging the number and time period of resources on the production line, adjusting the resource allocation of the supply chain, and reasonably arranging the working hours of personnel can all be optimized and decided through the trained model. In addition, through reasonable demand resource configuration and optimization, the production efficiency and economic benefits of the enterprise can also be improved, production costs can be reduced, and product quality and market competitiveness can be improved. Therefore, in the process of enterprise operation and management, using neural network models to predict and configure demand resources is a very effective means and strategy.

[0099] Example 1: Assume we have a period of historical time series data in, , , , , , . Use the forecasting model to predict the demand quantity at three future time points, respectively , , .

[0100] List the distribution of demand quantity at each time node:

[0101]

[0102]

[0103]

[0104]

[0105]

[0106]

[0107]

[0108]

[0109]

[0110] Based on the above distribution, we can calculate the number of demands at each time node t ,Right now:

[0111]

[0112]

[0113]

[0114]

[0115]

[0116] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0117] Finally, it should be noted that the above are only preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A digital management method based on artificial intelligence, characterized by: S1. Collect data related to demand; S2. Based on the collected data, a mathematical model is established to predict the distribution of future demand; S3. Immediately delete resource configurations that do not have demand based on demand distribution; S4. On the basis of deleting the resource configurations that do not have demand, if there is no new demand, no new resource configuration will be made. If there is a new demand, the neural network model will be used to configure appropriate resources for it; In S1, analyzing the distribution of demand in time series is the premise of realizing digital management. First, collect data related to demand: in, Indicates the type of demand. Indicates the number of types of demand, express The number of requirements, Indicates the time when the demand occurs. Indicates the number of times the demand occurs. Indicates The moment of need, represents the duration of the demand, Indicates the duration of the demand. Indicates The duration of the demand segment; In S2, the following steps are included: S201. After collecting the demand-related data, analyze the distribution pattern of past demand in the time series and understand the demand sequence that can be observed at time t. Set the time span to 1, and use the collection of demand distribution at the time point. To express, , analyze the law of demand in time series, that is, analyze the trend of demand appearance and disappearance, so the following time series mathematical model is established: in, Indicates at time The value of represents a constant, represents the residual, represents the autoregressive term coefficient, represents the coefficient of the moving average term, represents the order of the autoregressive term, represents the order of the moving average term; The establishment of this model describes the trend of all demands in the time series from the prediction time point forward, and derives the law of the appearance and disappearance of demands, which can be used to further predict the distribution of demands in the future time series. S202. Forecast demand by establishing a future demand forecasting model. The model is as follows: in, Represents the predicted future time The value on represents a constant, represents the residual, represents the autoregressive term coefficient, represents the coefficient of the moving average term, represents the order of the autoregressive term, represents the order of the moving average term; In this model, represents a point in time in the future, and It represents all past demand data, that is, the model is based on the demand distribution law analyzed by the time series mathematical model to obtain the distribution of future demand. Since the dimension of the demand distribution described by the future demand forecasting model is different from that of the time series mathematical model, the forecast result is further restored to the dimension of the original demand distribution to obtain the final forecast result. S203. The distribution of future demand can be obtained through the future demand forecasting model. However, since the focus is on predicting the trend of demand generation and disappearance, the distribution of demand will be inaccurate for some unpredictable emergencies. Considering that the accuracy of the forecast value plays a vital role in resource management and allocation, it is necessary to correct the predicted demand distribution and establish a forecast value error correction model: in, Represents the predicted value at a future time point, Represents data at a historical point in time. Represents the weighting coefficient, which can be adjusted as needed; S204, using the forecast value error correction model to correct the output of the future demand forecast model, and construct the distribution status of demand at future time nodes; In S3, the resource configuration without demand is deleted immediately according to the demand distribution. The demand distribution is analyzed by the future demand prediction model to establish the following resource discrimination model: in, Indicates time The sum of all the demand quantities that exist at Indicates the demand quantity, Indicates at time When the demand quantity The result of the assignment; According to the resource discrimination model, a discrimination mechanism for resource allocation regarding the generation and disappearance of demand is generated: First, collect the time series The maximum value of When, with The process from 1 to the maximum value, for , that is, the demand quantity is 1, and the actual situation is analyzed. The set value matches the condition. If the actual required quantity is indeed 1, then ,otherwise ,along with The increase is calculated in this way The value of After the value, each The value corresponds to Add up all the values ​​to get ,like , indicating that there is no demand at this moment, so the resource configuration can be deleted completely. If, Resources need to be allocated differently according to the different demand quantities; In S4, the following steps are included: S401. When new demands appear, they cannot be generalized due to differences between them, and characteristic information of the demands needs to be collected; S402: Digitize the characteristics of the demand as the input vector of the neural network model. The neural network is used to customize the resource allocation for the emerging demand. Therefore, the following neural network model is established: in, is represented as a feature vector, is the output value of the hidden layer, is the weight, is the bias, is the output of the previous layer, is the output value of the output layer, is the weight of the output layer, is the bias of the output layer, is the output of the last hidden layer, is the loss function, is the number of training samples; After the neural network model is established, the neural network model is trained according to the restrictions of the input features, and the output value in the neural network is calculated for each input feature vector. , further, after obtaining the output value in the neural network, calculate the loss function value Gradient ; and for each hidden layer Node, calculate the gradient of the current node , and calculate the gradient of its input weights and biases by the chain rule and ; S404, after calculating the data in the neural network model, the data is applied. First, the output value of the neural network is calculated by providing an input vector; further, the output value is compared with the true value to calculate the value of the loss function; next, the gradient of the loss function is used to adjust the network parameters to minimize the loss function; specifically, the gradient is calculated starting from the output layer in a reverse manner, and the error signal is propagated from the output layer to all nodes of the hidden layer and the input layer to optimize all parameters; S405. By executing the process of adjusting the network parameters, the parameters are further updated. According to the calculated gradient, each parameter is updated. Through continuous iteration, the parameters of the network are optimized so that the effect of the demand resource configuration reaches the best state.

2. A digital management system based on artificial intelligence, the system is applied to the digital management method based on artificial intelligence as claimed in claim 1, characterized in that: The system includes a data acquisition module, a demand forecasting module, a neural network analysis module, a resource management module and a database; The data acquisition module is used to collect data related to demand; the demand prediction module is used to predict the time node when future demand will appear; the neural network analysis module is used to allocate resources to demand when demand appears; the resource management module is used to use the prediction module to find out the time period when demand does not appear in the time series and stop providing resources for the demand, while providing resources for new demand; the database is used to store the collected data related to demand; The demand forecasting module includes a demand analysis unit and a forecast result output unit; The demand analysis unit is used to analyze the distribution of demand in the historical time series according to the information in the data acquisition module; the prediction result output unit is used to output the time node when the demand occurs; The neural network analysis module includes a demand feature input module, a model training unit and an analysis result output unit; The demand feature input module is used to digitally input the relevant features of the demand; the model training unit is used to train the goal of resource allocation; the analysis result output unit is used to output the analysis result of the neural network model; The resource management module includes a demand replacement unit and a resource configuration unit; The demand replacement unit is used to delete non-existent demands and add new demands according to the results of the demand prediction module; the resource configuration unit is used to configure resources for the newly added demands.

3. According to claim 2, a digital management system based on artificial intelligence is characterized in that: The data collection module includes a demand type collection unit and a demand counting unit; The demand type collection unit is used to collect statistics on a number of demands; the demand counting unit is used to record the number of times a demand occurs in a time period and the duration of the demand.

4. According to claim 3, a digital management system based on artificial intelligence is characterized in that: The database includes a data storage unit and a data retrieval unit; The data storage unit is used to store demand data, resource data, forecast data and statistical data; the data retrieval unit is used to retrieve information for data analysis.