CN117271099B - Automatic space data analysis scheduling system and method based on rule base - Google Patents

Abstract

An automatic scheduling system and method for spatial data analysis based on a rule base, including a spatial data input module, a data preprocessing module, a rule definition and matching module, a data processing automatic scheduling and optimization module, an error handling module, a user interface module, and a spatial data The input module is used to upload spatial data, the data preprocessing module is used to preprocess spatial data, the rule definition and matching module is used to define rule base rules and match spatial data and rules, and the data processing automatic scheduling and optimization module is used to schedule spatial data processing. tasks and optimizes the scheduling process, the error handling module is used to handle scheduling exceptions, and the user interface module is used to provide a user interface. The present invention proposes a distributed optimal matching algorithm that improves the two-way multiplier method to match spatial data and rules, and proposes an automatic scheduling algorithm that improves the deep Q network to automatically schedule spatial data processing tasks.

Description

An automatic scheduling system and method for spatial data analysis based on a rule base

Technical field

The invention relates to the fields of spatial data processing, optimal matching and automatic task scheduling, and specifically relates to an automatic scheduling system and method for spatial data analysis based on a rule base.

Background technique

Spatial data processing is a technical means to acquire, preprocess, analyze, store and visualize spatial data. These technologies are used to process different types of spatial data, such as geographic information system (GIS) data, remote sensing data, geolocation data, etc. , to perform various spatial data analysis tasks, obtain various types of spatial data, including satellite images, sensor data, map data and GPS trajectory data, and use various geographical information analysis and spatial data analysis methods, such as geospatial pattern recognition, Spatial buffer analysis and map algebra operations are used to derive useful information about spatial data, and predefined rule libraries and models are applied to spatial data to automate specific analysis tasks so that users can perform spatial data analysis more easily.

Optimal matching is a method used to determine the best matching rules. The purpose of this technique is to select appropriate rules from a rule base in order to perform spatial data analysis tasks and obtain the best results when considering the performance and constraints of different rules. , the suitability of each rule is usually evaluated using a set of evaluation criteria, including the rule's accuracy, precision, computational efficiency, and memory usage. The rule port model is ranked according to these evaluation criteria to determine which rule is most suitable for the current Analysis task, the choice of the optimal matching technique will depend on the specific needs of the problem, the nature of the rule base and the characteristics of the spatial data analysis task. The goal of these techniques is to ensure that the most suitable rules for the current task can be selected and executed to achieve the best analysis results.

Automatic task scheduling is a technology that can automatically identify, plan and arrange the execution of different spatial data analysis tasks without the need for manual intervention by the user. This technology is usually based on a set of pre-defined rules and conditions to perform the following tasks, based on conditions in the rule base Triggers are used to complete the data changes in the task description provided by the user. Plan the execution sequence and time of various analysis tasks, involving priority allocation, task dependency analysis and resource allocation, and allocate computing resources and storage resources according to task requirements. and data access rights in order to perform analysis tasks. Once the tasks are identified, planned and resource allocated, these tasks will be automatically executed. The goal of automatic scheduling technology is to reduce the user's operational burden and improve the efficiency and automation of spatial data analysis. Users It can set the rule base and task parameters when needed, and then automatically execute the task according to the automatic scheduling method. This is particularly useful for large-scale and complex spatial data analysis tasks, and can improve work efficiency and accuracy.

Contents of the invention

In response to the above problems, the present invention aims to provide an automatic scheduling method for spatial data analysis based on a rule base.

The purpose of the invention is achieved through the following technical solutions:

An automatic scheduling method for spatial data analysis based on a rule base, including a spatial data input module, a data preprocessing module, a rule definition and matching module, a data processing automatic scheduling and optimization module, an error handling module and a user interface module, a spatial data input module It is used to upload spatial data. The data preprocessing module is used to preprocess spatial data. The rule definition and matching module includes a rule base definition unit and a rule matching unit. The rule base definition unit is used to define rule base rules. The rule matching unit proposes improved two-way multiplication. The mathematical distributed optimal matching algorithm optimally matches spatial data with rule base rules. The data processing automatic scheduling and optimization module includes an automatic scheduling unit and a scheduling optimization unit. The automatic scheduling unit proposes an automatic scheduling algorithm that improves the deep Q network. Scheduling spatial data processing tasks, the scheduling optimization unit is used to optimize the automatic scheduling process, the error handling module is used to handle exceptions that occur during the scheduling process, and the user interface module is used to provide a user interface.

Furthermore, the spatial data input module uses satellite remote sensing, sensors, databases, and Internet data sources to obtain various types of spatial data, and uploads the spatial data to the rule base.

Furthermore, the data preprocessing module preprocesses spatial data by cleaning and repairing errors, missing and inconsistent information in spatial data, while performing data calibration and data dimensionality reduction.

Furthermore, the rule base definition unit is used to define and manage rules. It uses conditions, operations and data processing step rules to guide the system on how to effectively process and analyze large-scale spatial data, and organizes the rules according to different tasks, analysis types and data. Types are classified to process spatial data according to applying appropriate rules.

Furthermore, the rule matching unit proposes to improve the distributed optimal matching algorithm of the two-way multiplier method to optimally match spatial data and rule base rules.

Further, the distributed optimal matching algorithm of the improved two-way multiplier method is as follows: First, the neighborhood radius is adaptively determined by calculating the distance between spatial data samples. Based on the neighborhood radius, the density of each type of data space sample is obtained. And increase the clustering center. At the same time, use the fuzzy clustering effectiveness index to judge the current clustering effect. Then, select the best cluster number and clustering center. Finally, optimize the clustering by minimizing the clustering objective function. Class results, as follows: Spatial dataset at turning points in, among,/> is the data set of turning points,/> is the density of the first turning point,/> is the density of the second turning point,/> For the first/> density of turning points,/> is the density of the nth turning point. In order to ensure the adaptability of the algorithm, the neighborhood radius is adaptively determined according to the equation, that is: , where M is the neighborhood radius,/> For the first/> density of turning points,/> For the first/> density of turning points,/> Yes/> and/> The Euclidean distance between/> When clustered into k, the cluster center for/> , where,/> ,/> is the selected group of cluster centers,/> for/> and collection/> The sum of the distances between the cluster centers in , the user fuzzy clustering effectiveness index is used to measure the clustering effect. The fuzzy clustering effectiveness index is:/> , where,/> For the first/> Class No./> data samples/> degree of membership,/> is the center of the m-th cluster,/> is the h-th cluster center,/> for/> and/> The distance between/> for/> and The distance between/> for/> and/> The greatest common divisor,/> for/> and/> The least common multiple of , and then search for the best turning area starting from the starting point;

Then solve the optimal matching rule. The matching objective function between spatial data and rules is , where R is the matching objective function between spatial data and rules, /> For continuous cross product operation,/> is a set of rules, A is a set of spatial data, a is the spatial data in A, B is a set of rules, b is the rule in B,/> is the utility function of spatial data a,/> is the utility function of rule b, is the optimal matching rule, and the two-way multiplier method is/> , and/> , where BM is the two-way multiplier objective function,/> is the utility function of spatial data a,/> is the utility function of rule b, X and Y are constant matrices, and c is a constant vector. In order to facilitate the solution of the matching objective function R, the matching objective function R is improved to/> through the two-way multiplier method. , that is/> , where,/> For the improved matching objective function, the Lagrange multiplier method is used to To solve, that is , where L is the Lagrangian functional formula,/> is the Lagrange multiplier. In order to make the solution process more accurate, adding the dual condition will/> Further improved to/> ,Right now , where,/> for/> The final matching objective function is to find the optimal matching rule as quickly as possible/> Only in this way can the spatial data and rule base rules be optimally matched. Therefore, the iterative process of the algorithm should be improved, that is, + ,/> + , where,/> is the number of iterations,/> For the first/> Continuous cross product operation under iterations,/> To find the smallest matching rule/> ,/> is a parameter that controls the convergence speed, For the first/> Matching rules under iterations,/> For the first/> Matching rules under iterations, for To solve, there are/> ;

Finally, according to the optimal matching rules obtained by solving Perform optimal matching between spatial data and rules, based on the optimal matching rules/> To match spatial data and rules, first, establish a feature steering area data set/> , find/> center point/> , then gives the distance between the turning areas of any feature, that is:/> , where,/> Turn region for feature/> and/> The Euclidean distance, the speed and route of the characteristic turning area are:/> , ,/> , where,/> is the number of turning points of a specific type in the characteristic turning area; calculate the first turning area/> and next turning area/> The difference between them is expressed by the total distance of the turning area, that is: ,in, ,/> for/> and/> The steering difference between them, and finally, the optimal matching rule is converted into the total distance in the steering area/> The minimum value of , that is:/> , where,/> is the weight of route distance,/> In order to turn the distance of the region, the distributed optimal matching algorithm of the improved two-way multiplier method first classifies the data, and in the classification process, each clustering center is obtained by adaptively determining the neighborhood radius and gradually increasing the cluster center based on the sample density. The density of the data point area, and then the matching objective function is improved twice to conveniently and accurately solve the optimal matching rule. Finally, the optimal matching rule is converted into the total distance in the spatial data steering area to realize the integration of spatial data and rule base. rules for optimal matching.

Furthermore, the automatic scheduling unit proposes an automatic scheduling algorithm that improves the deep Q network to automatically schedule spatial data processing tasks.

Further, the improved automatic scheduling algorithm of the deep Q network is as follows: the reward function is , where,/> is the reward function,/> is the scheduling strategy,/> is the normalization factor, is the maximum completion time of scheduling spatial data,/> is the lower limit of the maximum completion time of scheduling space data, and the cumulative reward is/> , among which, AR cumulative reward,/> is the reward at time t,/> is the reward at time t+1,/> is the reward at time t+2,/> is the reward at time t+N,/> is the discount factor, n is/> The time integer between, the update process of Q value in deep Q network is/> , where,/> Take action in state s/> Q value,/> To control the learning rate of the update step of the Q value in each iteration,/> To take action in state s Reward obtained after,/> is the discount factor,/> To take action/> The new state obtained after,/> for in new status/> The optimal scheduling action in the scheduling strategy under , in order to solve the overestimation problem of deep Q network, // Improved to two independent Q-value functions, namely/> , , where,/> is the first independent Q-value function,/> is the second independent Q-value function, // Evaluate the best action for automatic scheduling, which will Used to update the Q value, through the interactive use of two independent Q value functions, to solve the overestimation problem of the deep Q network. In order to have better adaptability in the automatic scheduling process, the learning rate/> Improved by learning rate decay to/> To improve algorithm performance, i.e./> , where,/> is the improved learning rate, m is the attenuation factor,/> Update the number of steps for the iteration and add/> The factor controls the attenuation amplitude of the learning rate, that is/> , where,/> To join/> The learning rate after the factor, improves the automatic scheduling algorithm of the deep Q network. First, the original depth Q value function is decomposed into two independent depth Q value functions to solve the overestimation problem of the deep Q network, and then a pair of learning rate attenuation and attenuation amplitude control factors are proposed. The learning rate is improved to make the automatic scheduling process adaptive and better convergent, thereby realizing automatic scheduling of spatial data processing tasks.

Furthermore, the scheduling optimization unit adjusts the execution order of tasks by dynamically scheduling real-time spatial data, and establishes a monitoring system to track the performance and resource utilization of task execution, as well as changes in task execution time, to optimize the automatic scheduling process.

Furthermore, the error processing module monitors the entire data analysis and scheduling process, as well as the input data, to detect potential errors and abnormal situations. Once an error occurs, the error processing module will record the type, time and related information of the error, and perform fault management. Troubleshooting and problem analysis to handle and manage errors and anomalies that occur during spatial data analysis and automated scheduling.

Furthermore, the user interface module is used to provide users with a visual and interactive interface so that users can easily apply spatial data analysis automatic scheduling methods to manage spatial data, and at the same time allow users to input spatial data to be processed and rule base rules. , users can set specific requirements for analysis tasks through the interface. Users can view the results of analysis tasks through the user interface and present them in the form of graphics and charts. The visualization of the results helps users better understand the spatial data analysis results.

Beneficial effects created by the present invention: The innovative point of the present invention is to propose an automatic scheduling method for spatial data analysis based on a rule base, which is used for automatic scheduling of spatial data analysis. Through the spatial data input module, data preprocessing module, and rules The integration of the definition and matching module, data processing automatic scheduling and optimization module, error handling module and user interface module provides an automatic scheduling method for spatial data processing and analysis, and proposes a distributed optimal matching algorithm that improves the two-way multiplier method for spatial data processing. Data processing tasks are automatically scheduled. The innovation of the present invention is that the distributed optimal matching algorithm that improves the two-way multiplier method first classifies the data, and during the classification process, the neighborhood radius and sample density are adaptively determined. Gradually increase the clustering center to obtain the density of each data point area, then improve the matching objective function twice to conveniently and accurately solve the optimal matching rules, and finally convert the optimal matching rules into the total number of points in the spatial data steering area. distance to achieve optimal matching of spatial data and rule base rules, and an improved deep Q network automatic scheduling algorithm is proposed to automatically schedule spatial data processing tasks. The innovation of the present invention is that the improved deep Q network automatic scheduling algorithm first combines the original The deep Q-value function is decomposed into two independent deep Q-value functions to solve the overestimation problem of the deep Q-network, and then the learning rate attenuation and attenuation amplitude control factors are proposed to improve the learning rate to make the automatic scheduling process adaptive and more efficient. Good convergence enables automatic scheduling of spatial data processing tasks, effectively improves the working effect of an automatic scheduling method for spatial data analysis based on a rule base, and provides a more comprehensive and accurate method for an automatic scheduling method for spatial data analysis based on a rule base. Technical support is provided to provide better decision-making support for a safe, scientific and efficient spatial data analysis automatic scheduling method based on a rule base. At the same time, the present invention involves an optimal matching algorithm and a reinforcement learning algorithm to provide people with convenient and efficient An automatic scheduling method for spatial data analysis based on a rule base can also consolidate the foundation for the development of other application fields. In the era of flourishing development of spatial data processing, optimal matching and automatic task scheduling, spatial data processing, optimal matching and The integration of automatic task scheduling has laid a solid foundation for the development of multi-domain integration, and can be applied to multiple industries and fields in the market, providing a new development direction for the integration of spatial data processing, optimal matching and automatic task scheduling. , which has contributed important application value to the field of spatial data processing technology.

Description of the drawings

The invention and creation are further explained using the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the invention and creation. For those of ordinary skill in the art, without exerting creative work, they can also obtain the results according to the following drawings. Other drawings.

Figure 1 is a schematic structural diagram of the present invention.

Detailed ways

The present invention will be further described in conjunction with the following examples.

An automatic scheduling method for spatial data analysis based on a rule base, including a spatial data input module, a data preprocessing module, a rule definition and matching module, a data processing automatic scheduling and optimization module, an error handling module and a user interface module, a spatial data input module It is used to upload spatial data. The data preprocessing module is used to preprocess spatial data. The rule definition and matching module includes a rule base definition unit and a rule matching unit. The rule base definition unit is used to define rule base rules. The rule matching unit proposes improved two-way multiplication. The mathematical distributed optimal matching algorithm optimally matches spatial data with rule base rules. The data processing automatic scheduling and optimization module includes an automatic scheduling unit and a scheduling optimization unit. The automatic scheduling unit proposes an automatic scheduling algorithm that improves the deep Q network. Scheduling spatial data processing tasks, the scheduling optimization unit is used to optimize the automatic scheduling process, the error handling module is used to handle exceptions that occur during the scheduling process, and the user interface module is used to provide a user interface.

Preferably, the spatial data input module uses satellite remote sensing, sensors, databases, and Internet data sources to obtain various types of spatial data, such as geographic information system (GIS) data, meteorological data, terrain data, and population data, and uploads the spatial data to the rule base.

Preferably, the data preprocessing module preprocesses spatial data by cleaning and repairing errors, missing and inconsistent information in spatial data, while performing data calibration and data dimensionality reduction, reducing computational complexity and improving analysis efficiency. .

Preferably, the rule base definition unit is used to define and manage rules, guide the system how to effectively process and analyze large-scale spatial data through conditions, operations and data processing step rules, and apply the rules according to different tasks, analysis types and data Types are classified to process spatial data according to the application of appropriate rules, improving the efficiency of spatial data analysis while ensuring consistency and repeatability.

Preferably, the rule matching unit is proposed to improve the distributed optimal matching algorithm of the two-way multiplier method to optimally match the spatial data and the rules of the rule base.

Specifically, the distributed optimal matching algorithm of the improved two-way multiplier method is as follows: First, the neighborhood radius is adaptively determined by calculating the distance between spatial data samples. According to the neighborhood radius, the density of each type of data space sample is obtained. And increase the clustering center. At the same time, use the fuzzy clustering effectiveness index to judge the current clustering effect. Then, select the best cluster number and clustering center. Finally, optimize the clustering by minimizing the clustering objective function. Class results, as follows: Spatial dataset at turning points , where KP is the data set of turning points, /> is the density of the first turning point,/> is the density of the second turning point,/> For the first/> density of turning points,/> is the density of the nth turning point. Since the density of the turning point/> is its neighborhood radius/> The number of adjacent turning points within, so the turning point density is related to the domain radius. In order to ensure the adaptability of the algorithm, the neighborhood radius is adaptively determined according to the equation, that is:/> , where M is the neighborhood radius,/> For the first/> density of turning points,/> For the first/> density of turning points,/> Yes/> and/> Euclidean distance between, when KP is clustered into/> When, the clustering center/> for/> ,in, is the selected group of cluster centers,/> for/> and collection/> The sum of the distances between the clustering centers in , the user fuzzy clustering effectiveness index is used to measure the clustering effect. The fuzzy clustering effectiveness index is:/> , where,/> for the first Class No./> data samples/> degree of membership,/> is the center of the m-th cluster,/> is the h-th cluster center,/> for/> and/> The distance between/> for and/> The distance between/> for/> and/> The greatest common divisor,/> for/> and/> The least common multiple of , and then search for the best turning area starting from the starting point;

Then solve the optimal matching rule. The matching objective function between spatial data and rules is , where R is the matching objective function between spatial data and rules, /> For continuous cross product operation,/> is a set of rules, A is a set of spatial data,/> is the spatial data in A, B is the set of rules, b is the rule in B,/> For spatial data/> The utility function of /> is the utility function of rule b,/> is the optimal matching rule, and the two-way multiplier method is/> ,and , where BM is the two-way multiplier objective function,/> For spatial data/> The utility function of /> is the utility function of rule b, X and Y are constant matrices, and c is a constant vector. In order to facilitate the solution of the matching objective function R, the matching objective function R is improved to/> through the two-way multiplier method. ,Right now ,in, For the improved matching objective function, the Lagrange multiplier method is used to To solve, that is , where L is the Lagrangian functional formula,/> is the Lagrange multiplier. In order to make the solution process more accurate, adding the dual condition will/> Further improved to/> ,Right now , where,/> for/> The final matching objective function is to find the optimal matching rule as quickly as possible/> Only in this way can the spatial data and rule base rules be optimally matched. Therefore, the iterative process of the algorithm should be improved, that is, + , , where,/> is the number of iterations,/> For the first/> Continuous cross product operation under iterations,/> To find the smallest matching rule/> ,/> is a parameter that controls the convergence speed,/> For the first/> Matching rules under iterations, For the first/> Matching rules under iterations, pair/> To solve, there is ;

Finally, according to the optimal matching rules obtained by solving Perform optimal matching between spatial data and rules, based on the optimal matching rules/> To match spatial data and rules, first, establish a feature steering area data set/> , find/> center point/> , then gives the distance between the turning areas of any feature, that is:/> , where,/> Turn region for feature/> and/> The Euclidean distance, the speed and route of the characteristic turning area are:/> , ,/> , where,/> is the number of turning points of a specific type in the characteristic turning area; calculate the first turning area/> and next turning area/> The difference between them is expressed by the total distance of the turning area, that is: ,in, ,/> for/> and/> The steering difference between them, and finally, the optimal matching rule is converted into the total distance in the steering area/> The minimum value of , that is: , where,/> is the weight of route distance,/> In order to turn the distance of the region, the distributed optimal matching algorithm of the improved two-way multiplier method first classifies the data, and in the classification process, each clustering center is obtained by adaptively determining the neighborhood radius and gradually increasing the cluster center based on the sample density. The density of the data point area, and then the matching objective function is improved twice to conveniently and accurately solve the optimal matching rule. Finally, the optimal matching rule is converted into the total distance in the spatial data steering area to realize the integration of spatial data and rule base. rules for optimal matching.

Preferably, the automatic scheduling unit proposes an automatic scheduling algorithm that improves the deep Q network to automatically schedule spatial data processing tasks.

Specifically, the automatic scheduling algorithm of the improved deep Q network is as follows: the reward function is , where,/> is the reward function,/> is the scheduling strategy,/> is the normalization factor, is the maximum completion time of scheduling spatial data,/> is the lower limit of the maximum completion time of scheduling spatial data. This reward function is more suitable for the scheduling problem of spatial data. The cumulative reward is Among them, AR is the cumulative reward,/> is the reward at time t,/> is the reward at time t+1,/> is the reward at time t+2,/> is the reward at time t+N,/> is the discount factor, n is/> time integer between, the update process of Q value in deep Q network is , where,/> To take action in state s/> Q value,/> To control the learning rate of the update step of the Q value in each iteration,/> To take action in state s Reward obtained after,/> is the discount factor,/> To take action/> The new state obtained after,/> for in new status/> The optimal scheduling action in the scheduling strategy under the Q value is updated through continuous iteration to better estimate the long-term cumulative reward of taking each action in each state. In order to solve the overestimation problem of the deep Q network, // Improved to two independent Q-value functions, namely/> , , among which,/> is the first independent Q-value function,/> is the second independent Q-value function, // Evaluate the best action for automatic scheduling, which will Used to update the Q value, through the interactive use of two independent Q value functions, to solve the overestimation problem of the deep Q network. In order to have better adaptability in the automatic scheduling process, the learning rate/> Improved by learning rate decay to/> To improve algorithm performance, i.e./> , among which,/> is the improved learning rate, m is the attenuation factor,/> Update the number of steps for the iteration and add/> The factor controls the attenuation amplitude of the learning rate, that is/> , among which,/> To join/> The learning rate after the factor, improves the automatic scheduling algorithm of the deep Q network. First, the original depth Q value function is decomposed into two independent depth Q value functions to solve the overestimation problem of the deep Q network, and then a pair of learning rate attenuation and attenuation amplitude control factors are proposed. The learning rate is improved to make the automatic scheduling process adaptive and better convergent, thereby realizing automatic scheduling of spatial data processing tasks.

Preferably, the scheduling optimization unit adjusts the execution order of tasks by dynamically scheduling real-time spatial data, and establishes a monitoring system to track the performance and resource utilization of task execution, as well as changes in task execution time, to optimize the automatic scheduling process.

Preferably, the error processing module monitors the entire data analysis and scheduling process, as well as the input data, to detect potential errors and abnormal situations. Once an error occurs, the error processing module will record the type, time and related information of the error, and perform fault management. Troubleshooting and problem analysis to handle and manage errors and exceptions that occur during spatial data analysis and automatic scheduling. For some known errors and exceptions, the error handling module attempts to automatically handle them. For example, if some If the data is missing, it will be filled in according to the predefined rules. If there is a problem with the rule matching, the matching rules will be adjusted.

Preferably, the user interface module is used to provide users with a visual and interactive interface so that users can easily apply spatial data analysis automatic scheduling methods to manage spatial data, and at the same time allow users to input spatial data to be processed and rule base rules. , such as spatial data sets, rule bases, analysis task parameters, etc. Users can set specific requirements for analysis tasks through the interface, including required rules, analysis methods and expected results. Users can view the results of analysis tasks through the user interface and use Presented in the form of graphs and charts, the visualization of results helps users better understand spatial data analysis results.

An automatic scheduling method for spatial data analysis based on a rule base is proposed for automatic scheduling of spatial data analysis. Through the spatial data input module, data preprocessing module, rule definition and matching module, data processing automatic scheduling and optimization module, error The integration of the processing module and the user interface module provides an automatic scheduling method for spatial data processing and analysis, and proposes a distributed optimal matching algorithm that improves the two-way multiplier method to automatically schedule spatial data processing tasks. The innovation of the present invention is that , the distributed optimal matching algorithm that improves the two-way multiplier method first classifies the data, and during the classification process, the density of each data point area is obtained by adaptively determining the neighborhood radius and gradually increasing the cluster center based on the sample density. , then improve the matching objective function twice to conveniently and accurately solve the optimal matching rules, and finally convert the optimal matching rules into the total distance in the spatial data steering area to achieve optimal matching of spatial data and rule base rules. , an improved automatic scheduling algorithm for deep Q networks is proposed to automatically schedule spatial data processing tasks. The innovation of the present invention is that the improved automatic scheduling algorithm for deep Q networks first decomposes the original depth Q-value function into two independent depth Q-value functions. Solve the overestimation problem of the deep Q network, and then propose learning rate attenuation and attenuation amplitude control factors to improve the learning rate so that the automatic scheduling process is adaptive and can converge better, realizing automatic scheduling of spatial data processing tasks, and effectively Improve the working effect of an automatic scheduling method for spatial data analysis based on a rule base, provide more comprehensive and accurate technical support for an automatic scheduling method for spatial data analysis based on a rule base, and provide a safe, scientific and efficient method based on The spatial data analysis automatic scheduling method of the rule base provides better decision support. At the same time, the present invention relates to the optimal matching algorithm and the reinforcement learning algorithm, providing people with a convenient and efficient spatial data analysis automatic scheduling method based on the rule base. It can also solidify the foundation for the development of other application fields. In an era when spatial data processing, optimal matching and automatic task scheduling are flourishing, the integration of spatial data processing, optimal matching and automatic task scheduling has laid a solid foundation for the development of multi-field integration. It provides a new development direction for the integration of spatial data processing, optimal matching and automatic task scheduling, and contributes important application value to the field of spatial data processing technology.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the scope of the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art will understand that , the technical solution of the present invention can be modified without departing from the essence and scope of the technical solution of the present invention.

Claims (7)

1. An automatic scheduling system for spatial data analysis based on a rule base, which is characterized by including a spatial data input module, a data preprocessing module, a rule definition and matching module, a data processing automatic scheduling and optimization module, an error handling module and a user interface module, the spatial data input module is used to upload spatial data, the data preprocessing module is used to preprocess spatial data, the rule definition and matching module includes a rule base definition unit and a rule matching unit, the rule base definition unit is used to define rule base rules, rules The matching unit uses a distributed optimal matching algorithm with an improved two-way multiplier method to optimally match spatial data with rule base rules. The data processing automatic scheduling and optimization module includes an automatic scheduling unit and a scheduling optimization unit. The automatic scheduling unit proposes an improved depth Q The network's automatic scheduling algorithm automatically schedules spatial data processing tasks, the scheduling optimization unit is used to optimize the automatic scheduling process, the error handling module is used to handle exceptions that occur during the scheduling process, and the user interface module is used to provide a user interface; The distributed optimal matching algorithm of the improved two-way multiplier method is specifically as follows: first, the neighborhood radius is adaptively determined by calculating the distance between spatial data samples, and based on the neighborhood radius, the density of each type of data space sample is obtained, and Increase the clustering center, and use the fuzzy clustering effectiveness index to judge the current clustering effect. Then, select the optimal number of clusters and clustering centers. Finally, optimize the clustering by minimizing the clustering objective function. The results are as follows: in the spatial data set of turning points KP = {kp ₁ , kp ₂ ,...,kp _i ,...,kp _n }, where KP is the data set of turning points, and kp ₁ is the first turning point The density of points, kp ₂ is the density of the second turning point, kp _i is the density of the i-th turning point, kp _n is the density of the n-th turning point, in order to ensure the adaptability of the algorithm, the neighborhood radius is based on the equation Adaptive determination, that is: Among them, M is the neighborhood radius, kp _i is the density of the i-th turning point, kp _j is the density of the j-th turning point, d(kp _i , kp _j ) is the Euclidean distance between kp _i and kp _j , when KP is clustered into k, the cluster center O _k is O _k ={kp _i |i＝argmax(B _i ×|M _ε (kp)|)}, where,/> O＝{O ₁ ,O ₂ ,…,O _k-1 } is a selected group of cluster centers,/> It is the sum of the distances between tp _i and the cluster center in set O. The user fuzzy clustering effectiveness index is used to measure the clustering effect. The fuzzy clustering effectiveness index is:/> Among them, U _m,i is the membership degree of the i-th (i=1,2,...,n) data sample kp _i in the m-th (m=1,2,...,k) class, and O _m is the m-th Cluster center, O _h is the h-th cluster center, d(O _m ,O _h ) is the distance between O _m and O _h , d(kp _i ,O _m ) is the distance between kp _i and O _m Distance, maxd(O _m ,O _h ) is the greatest common divisor of O _m and O _h , mind(O _m , O _h ) is the least common multiple of O _m and O _h , and then start from the starting point to find the best turn area; Then solve the optimal matching rule. The matching objective function between spatial data and rules is Among them, R is the matching objective function between spatial data and rules, ΠΓ is the continuous cross product operation, Γ is the set of rules, A is the set of spatial data, a is the spatial data in A, B is the set of rules, and b is For the rules in B, f _a is the utility function of spatial data a, g _b is the utility function of rule b, π _ab is the optimal matching rule, the two-way multiplier method is BM=min(f _a +g _b ), and Xa +Yb=c, where BM is a two-way multiplier objective function, f _a is the utility function of spatial data a, g _b is the utility function of rule b, X and Y are constant matrices, and c is a constant vector to facilitate pair matching The objective function is R to solve, and the matching objective function R is improved to R' through the two-way multiplier method, that is Among them, R' is the improved matching objective function, which is solved by the Lagrange multiplier method, that is, L=R'+∑ _a∈A ∑ _b∈B λ(f _a (π _ab )-g _a (π _a )), where L is the Lagrange functional formula and λ is the Lagrange multiplier. In order to make the solution process more accurate, the dual condition is added to further improve R' to R", that is, /> Among them, R" is the matching objective function after R'. Only by finding the optimal matching rule π _ab as quickly as possible can the spatial data and rule base rules be optimally matched. Therefore, the iterative process of the algorithm is improved, that is/ > Among them, k is the number of iterations, ΠΓ(k+1) is the continuous cross product operation under the (k+1)th iteration, In order to find the smallest matching rule π _ab , η is the parameter that controls the convergence speed, π _ab (k) is the matching rule under the k-th iteration, π _ab (k+1) is the matching rule under the k+1-th iteration , solving for π _ab (k+1), we have Finally, the optimal matching between spatial data and rules is performed according to the optimal matching rule π _ab (k) obtained from the solution, and the matching between spatial data and rules is performed according to the optimal matching rule π _ab (k). First, the characteristics are established Turning to the regional dataset KR _typei , find center point/> Then the distance between the turning regions of any feature is given, i.e.: in, is the Euclidean distance between the characteristic turning area i and i+k, and the speed and route of the characteristic turning area are:/> Among them, x is the number of specific types of turning points in the characteristic turning area; the difference between the first turning area KR _i and the next turning area KR _i+1 is calculated, expressed by the total distance of the turning area, that is:/> Among them,/> θ is the steering difference between KR _i and KR _i+1 . Finally, the optimal matching rule is converted into the minimum value of the total distance D _total in the steering area, that is: D _total =D _course ω _C +D _distance ω _D , Among them, ω _C is the weight of the route distance, and ω _D is the distance of the turning area. The distributed optimal matching algorithm that improves the two-way multiplier method first classifies the data, and adaptively determines the neighborhood radius and Based on the sample density, the cluster center is gradually increased to obtain the density of each data point area, and then the matching objective function is improved twice to conveniently and accurately solve the optimal matching rule, and finally the optimal matching rule is converted into a spatial data steering area The total distance in , achieving optimal matching of spatial data and rule base rules; The automatic scheduling unit proposes an automatic scheduling algorithm that improves the deep Q network to automatically schedule spatial data processing tasks; the automatic scheduling algorithm of the improved deep Q network is specifically as follows: the reward function is Among them, r(π) is the reward function, π is the scheduling strategy, μ is the normalization factor, and T _max is the maximum completion time of the scheduling space data,/> is the lower limit of the maximum completion time of scheduling spatial data, and the cumulative reward is Among them, AR is the cumulative reward, r _t is the reward at time t, r _t+1 is the reward at time t+1, r _t+2 is the reward at time t+2, r _{t+N is the reward} at time t+ The reward at N time, γ is the discount factor, n is the time integer between [0, N], the update process of Q value in the deep Q network is Q(s,a)=(1-α)Q(s,a) )+α[r+γmax _a' Q(s',a')], where Q(s,a) is the Q value of taking action a in state s, and α controls the update of the Q value in each iteration The learning rate of the step length, r is the reward obtained after taking action a in state s, γ is the discount factor, s' is the new state obtained after taking action a, a' is the scheduling policy under new state s' _The optimal scheduling action _of ,a)+α[r+γmax _a' Q ₁ (s′,a′)], Q ₂ (s,a)＝(1-α)Q ₂ (s,a)+α[r+γmax _a' Q ₁ (s',a')], where Q ₁ (s, a) is the first independent Q-value function, Q ₂ (s, a) is the second independent Q-value function, Q ₁ (s,a) evaluates the best action of automatic scheduling, uses Q ₂ (s,a) to update the Q value, and uses two independent Q value functions interactively to solve the overestimation problem of deep Q networks. In order to make automatic scheduling The process has better adaptability, so the learning rate α is improved to α' through learning rate attenuation to improve algorithm performance, that is, α' = α·e ^-m·epoch , where α' is the improved learning rate , m is the attenuation factor, epoch is the number of iterative update steps, and the factor factor is added to control the attenuation amplitude of the learning rate, that is, α”=α’·factor, where α” is the learning rate after adding the factor factor, improving the deep Q network The automatic scheduling algorithm first decomposes the original depth Q-value function into two independent depth Q-value functions to solve the overestimation problem of the deep Q-network, and then proposes learning rate attenuation and attenuation amplitude control factors to improve the learning rate so that the automatic scheduling process has It is adaptive and can achieve better convergence and realize automatic scheduling of spatial data processing tasks. 2. An automatic scheduling system for spatial data analysis based on a rule base according to claim 1, characterized in that the spatial data input module uses satellite remote sensing, sensors, databases, and Internet data sources to obtain various types of spatial data. data and upload spatial data to the rule base. 3. An automatic scheduling system for spatial data analysis based on a rule base according to claim 1, characterized in that the data preprocessing module simultaneously cleans and repairs errors, missing and inconsistent information in the spatial data. Data calibration and data dimensionality reduction to preprocess spatial data. 4. A spatial data analysis automatic scheduling system based on a rule base according to claim 1, characterized in that the rule base definition unit is used to define and manage rules, guided by conditions, operations and data processing step rules. How the system can effectively process and analyze large-scale spatial data and classify rules according to different tasks, analysis types and data types to process spatial data according to the application of appropriate rules. 5. An automatic scheduling system for spatial data analysis based on a rule base according to claim 1, characterized in that the rule matching unit proposes to improve the distributed optimal matching algorithm of the two-way multiplier method for spatial data and rule base rules. Make the best match. 6. An automatic scheduling system for spatial data analysis based on a rule base according to claim 1, characterized in that the scheduling optimization unit adjusts the execution order of tasks by dynamically scheduling real-time spatial data and establishes a monitoring system to track tasks. The execution performance and resource utilization, as well as changes in task execution time, realize the optimization of the automatic scheduling process; the error handling module monitors the entire data analysis and scheduling process, as well as the input data, to detect potential errors and abnormal situations. , Once an error occurs, the error handling module will record the type, time and related information of the error, and perform troubleshooting and problem analysis to handle and manage errors and exceptions that occur during spatial data analysis and automatic scheduling. 7. A method for automatic scheduling of spatial data analysis using the system according to any one of claims 1 to 6.