Resource allocation method and device based on exponential function clustering and multi-objective optimization
Through exponential function clustering and multi-objective optimization resource allocation methods, the coverage and data synchronization problems in three-dimensional coordinated water rescue of unmanned systems are solved, accurate prediction of accident areas and efficient resource deployment are achieved, and rescue efficiency and resource utilization are improved.
Patent Information
- Application Number
- CN202510479281.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-16
- Publication Date
- 2025-08-01
AI Technical Summary
The existing three-dimensional collaborative water rescue technology of unmanned systems faces problems of large coverage, data synchronization and communication delay, resulting in low efficiency and low resource allocation efficiency, and difficult to cope with complex and changing water rescue scenarios.
The resource allocation method based on exponential function clustering and multi-objective optimization is adopted, and accident data is predicted through long-term and short-term memory models, combined with kernel density analysis and exponential function density clustering, resource scheduling is used using EGA algorithm, resource allocation strategies are dynamically adjusted, and rescue efficiency and resource utilization are optimized.
Accurate prediction of accident areas and efficient resource deployment have been achieved, rescue speed and collaborative operation efficiency have been improved, complex rescue scenarios have been adapted to complex rescue scenarios, and resource waste has been avoided.
Smart Images
Figure CN120410052A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of information management, and particularly to a resource allocation method and device based on exponential function clustering and multi-objective optimization. Background Art
[0002] In the field of modern three-dimensional cooperative water rescue of unmanned systems, various related technologies have been developed. For example, multi-platform cooperation technology, through the joint actions of unmanned aerial vehicles, unmanned boats and unmanned underwater vehicles, jointly completes water rescue tasks. These platforms can independently execute tasks such as target search, positioning and rescue, and through advanced communication and data fusion technologies, achieve information sharing and collaborative decision-making. In addition, target recognition and tracking technology based on visual perception and multi-sensor fusion, by introducing deep learning algorithms and multi-trajectory association mechanisms, improves the detection and tracking accuracy of distress targets. These technical designs focus on multiple aspects such as communication, data fusion, autonomous navigation, target recognition and tracking between unmanned systems.
[0003] However, the existing three-dimensional cooperative water rescue technology of unmanned systems faces many challenges in practical applications. First, the required coverage area of unmanned systems is large, as well as the data synchronization and communication delay problems between multiple platforms, which limit the efficiency and accuracy of cooperative operations, affect the search and rescue speed of unmanned systems, and result in low efficiency of water emergency resource allocation; second, in complex and changeable water rescue scenarios, traditional technologies are difficult to handle efficiently, and the network cooperation ability is insufficient, resulting in low utilization rate of rescue resources and high idle costs. Summary of the Invention
[0004] In view of this, the present invention proposes a resource allocation method and device based on exponential function clustering and multi-objective optimization.
[0005] The technical solution of the present invention is implemented as follows: In the first aspect of the present invention, a resource allocation method based on exponential function clustering and multi-objective optimization is provided, including:
[0006] Inputting historical accident data in a target area into a trained long short-term memory model to obtain accident prediction data in a preset future period of the target area; the accident prediction data includes accident locations and accident quantities;
[0007] Constructing point cloud data based on the accident prediction data, and performing spatial clustering analysis on the point cloud data by using a kernel density analysis algorithm and an exponential function density clustering method to obtain accident density distribution information and corresponding clustering center information in the target area;
[0008] Using the EGA algorithm to perform resource scheduling analysis on the accident density distribution information and the clustering center information to determine a target resource scheduling strategy.
[0009] Based on the above technical solutions, preferably, inputting the historical accident data in the target area into the trained long short-term memory model to obtain the accident prediction data of the target area in a preset future period includes:
[0010] Combining the spatial attention mechanism with the stacked LSTM structure to obtain the original long short-term memory model;
[0011] Using the mean square error as the loss function to train the original long short-term memory model to obtain the trained long short-term memory model;
[0012] Inputting the historical accident data in the target area into the trained long short-term memory model, and using a sliding window to obtain the accident prediction data of the target area in a preset future period.
[0013] Based on the above technical solutions, preferably, constructing point cloud data based on the accident prediction data, and using the kernel density analysis algorithm and the exponential function density clustering method to perform spatial clustering analysis on the point cloud data to obtain the accident density distribution information and the corresponding clustering center information in the target area, including:
[0014] Analyzing the point cloud data according to the K-nearest neighbor point search algorithm to generate an accident density heat map;
[0015] Obtaining the density change rate between adjacent points in the accident density heat map, and identifying the accident points with the density change rate greater than the first threshold as candidate cluster boundary points;
[0016] For each candidate cluster boundary point, obtaining the nearest neighbor inter-cluster points, and performing point cloud clustering on the candidate cluster boundary points and the inter-cluster points based on the exponential function density clustering to obtain the accident density distribution information and the corresponding clustering center information in the target area.
[0017] Based on the above technical solutions, preferably, for each candidate cluster boundary point, obtaining the nearest neighbor inter-cluster points, and performing point cloud clustering on the candidate cluster boundary points and the inter-cluster points based on the exponential function density clustering to obtain the accident density distribution information and the corresponding clustering center information in the target area, including:
[0018] Determining the maximum distance from the inter-cluster points to the candidate cluster boundary points as the cut-off distance;
[0019] Within the cut-off distance, using the Euclidean distance as the distance between accident points, and constructing an initial local density model according to the exponential function to determine the local density and the constraint distance corresponding to the accident points in the target area;
[0020] Normalize the local density and the constraint distance, and determine the accident density distribution information and the cluster center information in the target area based on the index values obtained from the normalization.
[0021] Based on the above technical solutions, preferably, the initial local density model is:
[0022]
[0023] ρ i is the accident point density, d i is the distance between accident points, k i is the number of inter-cluster points of the nearest neighbor.
[0024] Based on the above technical solutions, preferably, using the EGA algorithm to perform resource scheduling analysis on the accident density distribution information and the cluster center information to determine the target resource scheduling strategy, including:
[0025] Use the EGA algorithm to perform resource scheduling analysis on the accident density distribution information and the cluster center information, and combine with the multi-criteria decision-making analysis algorithm to determine at least two resource allocation schemes;
[0026] Based on the NSGA-II algorithm, screen the resource allocation schemes to determine the target resource scheduling strategy.
[0027] Based on the above technical solutions, preferably, using the EGA algorithm to perform resource scheduling analysis on the accident density distribution information and the cluster center information, and combine with the multi-criteria decision-making analysis algorithm to determine at least two resource allocation schemes, including:
[0028] Use the EGA algorithm to perform resource scheduling analysis on the accident density distribution information and the cluster center information, and combine with the multi-criteria decision-making analysis algorithm to determine the total resource allocation duration and resource utilization rate corresponding to different resource allocation schemes, and obtain the Pareto optimal solution set.
[0029] Even more preferably, a second aspect of the present invention provides a resource allocation device based on exponential function clustering and multi-objective optimization, including: a data acquisition module, a clustering analysis module, and a scheduling analysis module; wherein,
[0030] The data acquisition module is configured to input the historical accident data in the target area into the trained long short-term memory model to obtain the accident prediction data in the target area during a preset future period; the accident prediction data includes the accident location and the number of accidents;
[0031] The clustering analysis module is configured to construct point cloud data based on the accident prediction data, and perform spatial clustering analysis on the point cloud data by using a kernel density analysis algorithm and an exponential function density clustering method to obtain accident density distribution information and corresponding clustering center information within the target area;
[0032] The scheduling analysis module is configured to perform resource scheduling analysis on the accident density distribution information and the clustering center information by using the EGA algorithm to determine a target resource scheduling strategy.
[0033] More preferably, a third aspect of the present invention provides an electronic device, including a processor and a memory; the memory stores a computer program, wherein the computer program, when executed by the processor, implements the resource allocation method based on exponential function clustering and multi-objective optimization described in the first aspect.
[0034] More preferably, a fourth aspect of the present invention provides a computer storage medium, on which a computer program is stored, wherein the computer program, when executed by a processor, implements the resource allocation method based on exponential function clustering and multi-objective optimization described in the first aspect.
[0035] The resource allocation method based on exponential function clustering and multi-objective optimization of the present invention has the following beneficial effects compared with the prior art:
[0036] 1. Through the learning and analysis of accident data by the long short-term memory model, the number and location of accidents can be accurately predicted, providing a scientific basis for the pre-allocation of rescue resources; secondly, by combining exponential density clustering to identify the accident black spot center, rescue resources can be targeted to be deployed in high-risk areas; finally, based on the multi-objective optimization of the EGA algorithm, the two key objectives of rescue efficiency and resource utilization rate can be balanced, avoiding the idleness and waste of resources.
[0037] 2. By combining the spatial attention mechanism with the stacked LSTM structure, on the basis of time series, dynamically learn the important feature areas of accident data in space, providing a spatially enhanced feature representation for the model; using multi-task learning enables the accident number prediction and accident location two tasks to share the underlying LSTM layer feature extraction part and perform predictions on different output layers respectively, enabling the model to optimize these two tasks simultaneously, predicting the number of accidents that may occur within a period of time in the surveyed area and the accident location, and improving the prediction efficiency.
[0038] 3. The multi-criteria decision analysis algorithm is adopted, taking multiple different factors as multiple criteria to construct a multi-criteria decision analysis model. By setting the weights of each criterion and comprehensively evaluating the rescue capabilities of resource allocation plans, it can adapt to complex and changeable water rescue scenarios, flexibly respond to different accident distributions and rescue requirements. By dynamically adjusting the resource allocation strategy, the system can quickly respond to environmental changes and improve the efficiency and accuracy of collaborative operations. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.
[0040] Figure 1 It is a schematic flowchart of a resource allocation method based on exponential function clustering and multi-objective optimization provided by an embodiment of the present invention;
[0041] Figure 2 It is a schematic diagram of the training process of the long short-term memory model provided by an embodiment of the present invention;
[0042] Figure 3 It is a schematic flowchart of forming accident black spots by using exponential function density clustering provided by an embodiment of the present invention;
[0043] Figure 4 It is a schematic flowchart of realizing multi-objective optimization based on the EGA algorithm provided by an embodiment of the present invention;
[0044] Figure 5 It is a schematic diagram of the optimized allocation of rescue resources based on the EGA algorithm provided by an embodiment of the present invention;
[0045] Figure 6 It is a schematic diagram of the principle of resource allocation based on exponential function clustering and multi-objective optimization provided by an embodiment of the present invention;
[0046] Figure 7 It is a schematic structural diagram of a resource allocation device based on exponential function clustering and multi-objective optimization provided by an embodiment of the present invention;
[0047] Figure 8 It is a schematic structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0048] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in combination with the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.
[0049] In some embodiments, as Figure 1 shown, Figure 1 is a schematic flowchart of a resource allocation method based on exponential function clustering and multi-objective optimization provided by an embodiment of the present invention; a resource allocation method based on exponential function clustering and multi-objective optimization provided by the present invention includes:
[0050] S110, input the historical accident data in the target area into the trained long short-term memory model to obtain the accident prediction data in the target area in a preset future period; the accident prediction data includes the accident location and the number of accidents.
[0051] Here, the target area can be part or all of the expected investigation area, and the historical accident data includes accident data for many years. The accident location can be represented by longitude and latitude, and multiple accidents can correspond to the same accident location.
[0052] In some embodiments, S110, input the historical accident data in the target area into the trained long short-term memory model to obtain the accident prediction data in the target area in a preset future period, including:
[0053] Combine the spatial attention mechanism with the stacked LSTM structure to obtain the original long short-term memory model;
[0054] Use the mean square error as the loss function to train the original long short-term memory model to obtain the trained long short-term memory model;
[0055] Input the historical accident data in the target area into the trained long short-term memory model, and use a sliding window to obtain the accident prediction data in the target area in a preset future period.
[0056] In this embodiment, for the specific training process of the long short-term memory model, please refer to Figure 2The spatial attention mechanism is combined with the stacked LSTM structure to obtain the original long short-term memory model. The original long short-term memory model includes an LSTM layer, an Attention layer, a Dropout layer, an LSTM layer, an LSTM layer, a Dropout layer, and a Dense layer arranged in sequence. For the normalized accident parameter sequence, the long short-term memory model not only focuses on the information input each time but also grasps the overall change trend. A spatial attention mechanism is added after the first LSTM layer, enabling the model to automatically learn the important feature regions of accident data in space, better capture the spatial patterns of accident occurrence, and avoid excessive computational burden. On this basis, a multi-task learning algorithm is introduced, and accident quantity prediction and accident location determination are learned simultaneously as two related tasks. By sharing the underlying LSTM layer feature extraction part and then making predictions in different output layers respectively, the model can optimize the performance of these two tasks simultaneously, improving the accuracy of accident quantity prediction and the precision of accident location. The number of iterations is set to 400, and according to the UAT method, two fully connected layers are used to fit the training model results. The mean squared error (MSE) is used as the loss function, and the error between the predicted value and the true value of accident occurrence can be calculated. The specific formula for MSE is:
[0057]
[0058] where n represents the total number of accidents, y i represents the true accident quantity or location of the i-th sample, represents the predicted accident quantity or location of the i-th sample.
[0059] After the model training is completed, an iterative prediction method is adopted to predict the accident data in the survey area within the next year. Use [X t-k+1 ,…,X t-1 ,X t as the input of the model, and the output result uses the data X t+1 at the t+1 moment. At the same time, only a sliding window is used to predict the data at the next time point during the prediction process. Thus, after predicting the data X t+1 , use [X t-k+1 ,…,X t-1 ,X t ,X t+1 as the new input again to obtain X t+2 , and so on, iteratively obtaining the prediction results of 12 groups of accident data within the next year. It should be noted that the above is only an exemplary illustration, and the prediction time can also be one month, one quarter, etc.
[0060] S120. Construct point cloud data based on accident prediction data, and perform spatial clustering analysis on the point cloud data by using the kernel density analysis algorithm and the exponential function density clustering method to obtain the accident density distribution information and the corresponding clustering center information within the target area.
[0061] In this embodiment, each accident prediction point can be regarded as a point (longitude, latitude, frequency) in three-dimensional space. If the data is sparse, supplementary points can be generated by interpolation or simulation. Through the kernel density analysis algorithm, the discrete point cloud data can be converted into a continuous density surface to reflect the accident spatial aggregation. Combining the density threshold and the spatial neighborhood relationship, the density correlation between points is quantified through an exponential decay function, so as to obtain the accident density distribution information and the corresponding clustering center information within the target area.
[0062] In some embodiments, S120. Construct point cloud data based on accident prediction data, and perform spatial clustering analysis on the point cloud data by using the kernel density analysis algorithm and the exponential function density clustering method to obtain the accident density distribution information and the corresponding clustering center information within the target area, including:
[0063] Analyze the point cloud data according to the K-nearest neighbor search algorithm to generate an accident density heat map;
[0064] Obtain the density change rate between adjacent points in the accident density heat map, and confirm the accident points with the density change rate greater than the first threshold as candidate cluster boundary points;
[0065] For each candidate cluster boundary point, obtain the nearest inter-cluster point, and perform point cloud clustering on the candidate cluster boundary point and the inter-cluster point based on the exponential function density clustering to obtain the accident density distribution information and the corresponding clustering center information within the target area.
[0066] In some embodiments, for each candidate cluster boundary point, obtain the nearest inter-cluster point, and perform point cloud clustering on the candidate cluster boundary point and the inter-cluster point based on the exponential function density clustering to obtain the accident density distribution information and the corresponding clustering center information within the target area, including:
[0067] Determine the farthest distance from the inter-cluster point to the candidate cluster boundary point as the cut-off distance;
[0068] Within the cut-off distance, use the Euclidean distance as the distance between accident points, and construct an initial local density model according to the exponential function to determine the local density and the constraint distance corresponding to the accident points within the target area;
[0069] Use normalization to standardize the local density and the constraint distance, and determine the accident density distribution information and the clustering center information within the target area based on the index values obtained by the standardization.
[0070] In this embodiment, please refer toFigure 3 , Figure 3 is a schematic flow chart of forming accident black spots by using exponential function density clustering provided by an embodiment of the present invention; First, by using the kernel density analysis method, by analyzing the data point density and distribution pattern on the accident density heat map, select the area with concentrated rescue resources and high accident prediction frequency as the sample point O, and according to the formula
[0071]
[0072] obtain the farthest distance from the surrounding points of adjacent areas within the investigation range to point O and use this as the cut-off distance. Where ɑ is the azimuth resolution, k is the number of accident points adjacent to point O, and D is the scanning distance.
[0073] Within the cut-off distance, use the Euclidean distance as the distance d between accident points in the cut-off area i , and construct an initial local density model according to the exponential function to ensure capturing the non-linear change of the predicted accident point density and making it sensitively reflect the subtle differences in accident point density. To accurately determine the clustering center, use the included angle between vectors to determine the positive and negative of the distance, and use the minimum distance σ i as the constraint factor. Use normalization to standardize the local density and constraint distance of each accident prediction point to the range of [0, 1] to ensure the comparability of accident prediction point data, so as to better observe and analyze the index values of prediction points. According to the clustering center identification and analysis, by observing the standardized index values, identify the accident points with significant advantages in local density and constraint distance, and this accident point is the required clustering center. Use the nearest neighbor analysis to further screen the clustering center. If the nearest neighbor distance is less than the cut-off distance, the two adjacent clusters belong to the same class and merge them together. According to the index values of the prediction points, generate an accident density map to show the accident density distribution and clustering center in different regions.
[0074] In some embodiments, the initial local density model is:
[0075]
[0076] ρ i is the accident point density, d i is the distance between accident points, k i is the number of inter-cluster points of the nearest neighbor.
[0077] S130. Use the EGA algorithm to perform resource scheduling analysis on the accident density distribution information and clustering center information to determine the target resource scheduling strategy.
[0078] In some embodiments, use the EGA algorithm to perform resource scheduling analysis on the accident density distribution information and clustering center information to determine the target resource scheduling strategy, including:
[0079] Use the EGA algorithm to perform resource scheduling analysis on accident density distribution information and clustering center information, and combine with the multi-criteria decision analysis algorithm to determine at least two resource allocation schemes;
[0080] Based on the NSGA-II algorithm, screen the resource allocation schemes to determine the target resource scheduling strategy.
[0081] In some embodiments, use the EGA algorithm to perform resource scheduling analysis on accident density distribution information and clustering center information, and combine with the multi-criteria decision analysis algorithm to determine at least two resource allocation schemes, including:
[0082] Use the EGA algorithm to perform resource scheduling analysis on accident density distribution information and clustering center information, and combine with the multi-criteria decision analysis algorithm to determine the total resource allocation duration and resource utilization rate corresponding to different resource allocation schemes, and obtain the Pareto optimal solution set.
[0083] In this embodiment, improving the emergency rescue efficiency under a certain resource allocation is taken as the core of the optimal emergency resource allocation strategy. On this premise, the configuration algorithm aims to improve the water rescue efficiency and enhance the utilization rate of the allocated resources, and selects the mooring location of the emergency unmanned boat, the parking location and quantity of the unmanned aerial vehicle as the objective functions respectively. For the specific configuration process, please refer to Figure 4 。
[0084] Taking the mooring location of the emergency unmanned ship, the parking location and quantity of the unmanned aerial vehicle as decision variables, an initial population P0 is generated using real number coding. Each individual represents a resource allocation plan, including the mooring location of the unmanned ship, the parking location and quantity of the unmanned aerial vehicle. The coordinate ranges of the mooring location of the unmanned ship and the parking location of the unmanned aerial vehicle are determined according to the accident density distribution in the rescue area, and the quantity of the unmanned aerial vehicle is randomly generated under the condition of meeting the resource constraints. The individuals in the population are divided into multiple levels according to the non-dominance relationship. For different resource allocation plans in the population, arbitrarily select individuals a and b among them. If a is not inferior to b in terms of both rescue efficiency and resource utilization rate, and is superior to b in at least one objective, then a dominates b. Regarding the evaluation strategy of rescue efficiency and resource utilization rate, a multi-criteria decision analysis (MCDA) algorithm is adopted. Different factors such as the index coefficient of accident black spots, spatial distance attenuation, and the rescue index coefficients of the unmanned ship and the unmanned aerial vehicle are used as multiple criteria to construct a multi-criteria decision analysis model. By setting the weights of each criterion, the rescue capabilities of resource allocation plans are comprehensively evaluated. Through non-dominated sorting, the individuals are divided into different frontiers (Pareto frontiers). The first layer of the frontier contains all individuals that are not dominated by any other individual, the second layer of the frontier contains all individuals that are only dominated by the individuals in the first layer of the frontier, and so on. At the same time, within each frontier, the individuals are sorted according to each objective value, and the difference between the objective values of adjacent individuals is calculated. The sum of these differences is used as the crowding degree of this individual. The specific calculation formula of the crowding degree is as follows:
[0085]
[0086] Among them, CD(i) represents the crowding degree of the i-th individual, and are the objective values of the adjacent individual after and before the i-th individual on the m-th objective function respectively, and are the maximum and minimum values of the m-th objective function on the current frontier, and M represents the total number of objective functions.
[0087] Based on the NSGA-II algorithm, individuals on the lower-level front are preferentially selected to preliminarily screen out resource allocation schemes with poor performance in rescue efficiency and resource utilization rate, ensuring the convergence of the population. At the same time, within the same front, individuals with a lower crowding degree are preferentially selected to maintain the diversity of the population. Select pairs of individuals in the population and perform simulated binary crossover and polynomial mutation operations to generate new individuals. The crossover operation is carried out on the coded values of the moored locations of the unmanned boats, the parking positions and quantities of the unmanned aerial vehicles, and the exploration and exploitation capabilities are balanced by controlling the crossover distribution index. For the individuals after crossover, polynomial mutation operations are performed with a mutation probability Pm. The mutation operation is carried out on each decision variable of the individual, and the mutation step size is adjusted according to the mutation distribution index to ensure the generation of feasible mutated individuals and increase the diversity of the resource allocation schemes. According to the non-dominated sorting and crowding degree information, high-quality individuals are selected from the current population and the population after crossover and mutation to form the next-generation population Pt+1. Individuals on the lower-level front and with a high crowding degree are preferentially selected to ensure the diversity and convergence of the population. Then, the next-generation population Pt+1 is used as the current population Pt and enters the next round of iteration. Repeat the above steps until the maximum number of iterations T is reached. Please refer to Figure 5 , Figure 5 which shows the Pareto Optimal solution set in terms of resource allocation. After the iteration ends, the non-dominated front is extracted from the final population, which is the Pareto solution set for the emergency rescue resource allocation problem. This solution set contains multiple resource allocation schemes that achieve an optimal balance between rescue efficiency and resource utilization rate.
[0088] In an alternative embodiment, please refer to Figure 6 , Figure 6 which is a schematic diagram of the principle of exponential function clustering and multi-objective optimization provided by the embodiment of the present invention. Divided by the functional layer, a virtual mapping layer, a clustering analysis layer, and a decision deduction layer can be obtained. Among them, the virtual mapping layer mainly predicts the accident points based on the LSTM algorithm to obtain an accident prediction distribution map. The clustering analysis layer mainly completes point cloud clustering based on the exponential function density algorithm. The decision deduction layer mainly performs iterative calculations through the EGA algorithm to obtain the Pareto optimal solution set. Through the learning and analysis of accident data by the LSTM algorithm, the number and location of accidents can be accurately predicted, providing a scientific basis for the pre-allocation of rescue resources. On this basis, combined with exponential density clustering to identify the accident black spot center, rescue resources can be targeted and deployed in high-risk areas, and then based on the multi-objective optimization of the EGA algorithm, the two key objectives of rescue efficiency and resource utilization rate are balanced, avoiding the idle and waste of resources. By dynamically adjusting the resource allocation strategy, it is possible to quickly respond to environmental changes and improve the efficiency and accuracy of collaborative operations.
[0089] In some embodiments, please refer to Figure 7 , Figure 7Schematic diagram of a resource allocation device based on exponential function clustering and multi-objective optimization provided by an embodiment of the present invention. The present invention provides a resource allocation device 700 based on exponential function clustering and multi-objective optimization, including: a data acquisition module 710, a clustering analysis module 720, and a scheduling analysis module 730; wherein,
[0090] The data acquisition module 710 is configured to input historical accident data in the target area into a trained long short-term memory model to obtain accident prediction data in the target area for a preset future period; the accident prediction data includes accident locations and the number of accidents;
[0091] The clustering analysis module 720 is configured to construct point cloud data based on the accident prediction data, and perform spatial clustering analysis on the point cloud data by using a kernel density analysis algorithm and an exponential function density clustering method to obtain accident density distribution information and corresponding clustering center information in the target area;
[0092] The scheduling analysis module 730 is configured to perform resource scheduling analysis on the accident density distribution information and the clustering center information by using the EGA algorithm to determine the target resource scheduling strategy.
[0093] In some embodiments, the data acquisition module 710 is specifically configured to:
[0094] Combine the spatial attention mechanism with the stacked LSTM structure to obtain the original long short-term memory model;
[0095] Use the mean square error as the loss function to train the original long short-term memory model to obtain a trained long short-term memory model;
[0096] Input the historical accident data in the target area into the trained long short-term memory model, and use a sliding window to obtain accident prediction data in the target area for a preset future period.
[0097] In some embodiments, the clustering analysis module 720 is specifically configured to:
[0098] Analyze the point cloud data according to the K-nearest neighbor point search algorithm to generate an accident density heat map;
[0099] Obtain the density change rate between adjacent points in the accident density heat map, and confirm the accident points with a density change rate greater than the first threshold as candidate cluster boundary points;
[0100] For each candidate cluster boundary point, obtain the nearest inter-cluster point, and perform point cloud clustering on the candidate cluster boundary point and the inter-cluster point based on exponential function density clustering to obtain accident density distribution information and corresponding clustering center information in the target area.
[0101] In some embodiments, the clustering analysis module 720 is specifically configured to:
[0102] Determine the farthest distance from the inter-cluster points to the candidate cluster boundary points as the cut-off distance;
[0103] Within the cut-off distance, use the Euclidean distance as the distance between accident points, and construct an initial local density model based on the exponential function to determine the local density and constraint distance corresponding to the accident points in the target area;
[0104] Normalize the local density and constraint distance, and determine the accident density distribution information and clustering center information in the target area based on the index values obtained from the normalization.
[0105] In some embodiments, the initial local density model is:
[0106]
[0107] ρ i is the accident point density, d i is the distance between accident points, k i is the number of inter-cluster points of the nearest neighbor.
[0108] In some embodiments, the scheduling analysis module 730 is specifically configured to:
[0109] Use the EGA algorithm to perform resource scheduling analysis on the accident density distribution information and clustering center information, and combine with the multi-criteria decision analysis algorithm to determine at least two resource allocation schemes;
[0110] Based on the NSGA-II algorithm, screen the resource allocation schemes to determine the target resource scheduling strategy.
[0111] In some embodiments, the scheduling analysis module 730 is specifically configured to:
[0112] Use the EGA algorithm to perform resource scheduling analysis on the accident density distribution information and clustering center information, and combine with the multi-criteria decision analysis algorithm to determine the total resource allocation duration and resource utilization rate corresponding to different resource allocation schemes, and obtain the Pareto optimal solution set.
[0113] It should be noted that the resource allocation device based on exponential function clustering and multi-objective optimization provided in the embodiments of the present application and the resource allocation method based on exponential function clustering and multi-objective optimization provided in the embodiments of the present application are based on the same inventive concept. Therefore, the specific implementation of this embodiment can refer to the implementation of the foregoing resource allocation method based on exponential function clustering and multi-objective optimization, and the repeated parts will not be described again.
[0114] In some embodiments, please refer to Figure 8 , Figure 8A schematic structural diagram of an electronic device provided by an embodiment of the present application. An electronic device 800 provided by an embodiment of the present application includes a processor 810 and a memory 820; the memory 820 stores a computer program, and when the computer program is executed by the processor, the above-mentioned resource allocation method based on exponential function clustering and multi-objective optimization is implemented.
[0115] Specifically, the processor 810 may include, for example, a general microprocessor, an instruction set processor, and / or a related chipset, and / or a dedicated microprocessor (such as an application-specific integrated circuit (ASIC)), and so on. The processor 810 may also include on-board memory for caching purposes. The processor 810 may be a single processing unit or multiple processing units for performing different actions of the method flow according to the embodiment of the present application.
[0116] The memory 820 may be, for example, any medium capable of containing, storing, transmitting, propagating, or transporting instructions. For example, the memory 820 may include, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, device, or propagation medium. Specific examples of the memory 820 include: a magnetic storage device, such as a magnetic tape or a hard disk drive (HDD); an optical storage device, such as a compact disc (CD-ROM); it may also be, for example, a random access memory (RAM) or a flash memory; and / or a wired / wireless communication link.
[0117] The present application also provides a computer-readable medium, on which a computer program is stored, and when the program is executed by the processor, the above-mentioned resource allocation method based on exponential function clustering and multi-objective optimization is implemented. The computer-readable medium may be included in the device / device / system described in the above embodiment; or it may exist separately and not be assembled into the device / device / system. The above computer-readable medium carries one or more programs, and when the one or more programs are executed, the method according to the embodiment of the present application is implemented.
[0118] According to an embodiment of the present application, a computer-readable medium may be a computer-readable signal medium, a computer-readable storage medium, or any combination of the two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples of the computer-readable storage medium may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, the computer-readable storage medium may be any tangible medium that contains or stores a program, which can be used by or in combination with an instruction execution system, apparatus, or device. In the present application, the computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries the computer-readable program code. Such a propagated data signal may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer-readable signal medium may also be any computer-readable medium other than the computer-readable storage medium, which can send, propagate, or transmit a program for use by or in combination with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium may be transmitted by any appropriate medium, including but not limited to: wireless, wired, optical cable, radio frequency signal, etc., or any suitable combination of the above.
[0119] Those skilled in the art can understand that the features recited in the various embodiments and / or claims of the present application can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly recited in the present application. In particular, without departing from the spirit and teachings of the present application, the features recited in the various embodiments and / or claims of the present application can be combined and / or combined in various ways. All such combinations and / or combinations fall within the scope of the present application. Therefore, the scope of the present application should not be limited to the above embodiments, but should be determined not only by the appended claims but also by the equivalents of the appended claims. Any modifications, equivalent replacements, improvements, etc., made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.
Claims
1. A resource allocation method based on exponential function clustering and multi-objective optimization, characterized in that Including: Inputting historical accident data within a target area into a trained long short-term memory model to obtain accident prediction data for the target area in a preset future period; The accident prediction data includes accident locations and the number of accidents; Constructing point cloud data based on the accident prediction data, and performing spatial clustering analysis on the point cloud data using a kernel density analysis algorithm and an exponential function density clustering method to obtain accident density distribution information and corresponding clustering center information within the target area; Performing resource scheduling analysis on the accident density distribution information and the clustering center information using an EGA algorithm to determine a target resource scheduling strategy.
2. The resource allocation method based on exponential function clustering and multi-objective optimization according to claim 1, wherein The step of inputting historical accident data within a target area into a trained long short-term memory model to obtain accident prediction data for the target area in a preset future period includes: Combining a spatial attention mechanism with a stacked LSTM structure to obtain an original long short-term memory model; Training the original long short-term memory model using mean squared error as a loss function to obtain a trained long short-term memory model; Inputting the historical accident data within the target area into the trained long short-term memory model, and using a sliding window to obtain accident prediction data for the target area in a preset future period.
3. The resource allocation method based on exponential function clustering and multi-objective optimization according to claim 1, wherein The step of constructing point cloud data based on the accident prediction data, and performing spatial clustering analysis on the point cloud data using a kernel density analysis algorithm and an exponential function density clustering method to obtain accident density distribution information and corresponding clustering center information within the target area includes: Analyzing the point cloud data according to the K-nearest neighbor point search algorithm to generate an accident density heat map; Obtaining the density change rate between adjacent points in the accident density heat map, and identifying accident points with a density change rate greater than a first threshold as candidate cluster boundary points; For each candidate cluster boundary point, obtaining the nearest inter-cluster points, and performing point cloud clustering on the candidate cluster boundary points and the inter-cluster points based on exponential function density clustering to obtain accident density distribution information and corresponding clustering center information within the target area.
4. The resource allocation method based on exponential function clustering and multi-objective optimization according to claim 3, wherein, The step of, for each candidate cluster boundary point, obtaining the nearest inter-cluster points, and performing point cloud clustering on the candidate cluster boundary points and the inter-cluster points based on exponential function density clustering to obtain accident density distribution information and corresponding clustering center information within the target area includes: Determining the farthest distance from the inter-cluster points to the candidate cluster boundary point as the cut-off distance; Within the cut-off distance, using the Euclidean distance as the distance between accident points, and constructing an initial local density model according to an exponential function to determine the local density and the constraint distance corresponding to accident points within the target area; Normalizing the local density and the constraint distance, and determining accident density distribution information and clustering center information within the target area based on the index values obtained from the normalization.
5. The resource allocation method based on exponential function clustering and multi-objective optimization according to claim 4, characterized in that The initial local density model is: ρ i is the accident point density, d i is the distance between accident points, k i is the number of inter-cluster points of the nearest neighbor.
6. The resource allocation method based on exponential function clustering and multi-objective optimization according to claim 1, wherein The step of performing resource scheduling analysis on the accident density distribution information and the clustering center information using an EGA algorithm to determine a target resource scheduling strategy includes: Use the EGA algorithm to perform resource scheduling analysis on the accident density distribution information and the clustering center information, and combine with the multi-criteria decision analysis algorithm to determine at least two resource allocation schemes; Based on the NSGA-II algorithm, screen the resource allocation schemes to determine the target resource scheduling strategy.
7. The resource allocation method based on exponential function clustering and multi-objective optimization according to claim 6, characterized in that, The use of the EGA algorithm to perform resource scheduling analysis on the accident density distribution information and the clustering center information, and combine with the multi-criteria decision analysis algorithm to determine at least two resource allocation schemes includes: Use the EGA algorithm to perform resource scheduling analysis on the accident density distribution information and the clustering center information, and combine with the multi-criteria decision analysis algorithm to determine the total resource allocation duration and resource utilization rate corresponding to different resource allocation schemes, and obtain the Pareto optimal solution set.
8. A resource allocation device based on exponential function clustering and multi-objective optimization, characterized in that Includes: A data acquisition module, a clustering analysis module, and a scheduling analysis module; where The data acquisition module is configured to input historical accident data in the target area into a trained long short-term memory model to obtain accident prediction data in the target area during a preset future period; the accident prediction data includes accident locations and the number of accidents; The clustering analysis module is configured to construct point cloud data based on the accident prediction data, and use the kernel density analysis algorithm and the exponential function density clustering method to perform spatial clustering analysis on the point cloud data to obtain the accident density distribution information and the corresponding clustering center information in the target area; The scheduling analysis module is configured to use the EGA algorithm to perform resource scheduling analysis on the accident density distribution information and the clustering center information to determine the target resource scheduling strategy.
9. An electronic device, comprising a processor and a memory; the memory stores a computer program, wherein, When the computer program is executed by the processor, it implements the resource allocation method based on exponential function clustering and multi-objective optimization according to any one of claims 1 to 7.
10. A computer storage medium, characterized in that, A computer program is stored thereon, and when the computer program is executed by the processor, it implements the resource allocation method based on exponential function clustering and multi-objective optimization according to any one of claims 1 to 7.