Automatic event allocation method and system based on historical disposal information and multi-target genetic algorithm

Abstract

The invention discloses an event automatic allocation method and system based on historical disposal information and a multi-target genetic algorithm, and relates to the technical field of information technology and smart city management. Comprising the steps that 1, event data samples which are processed are extracted from a historical event library, each data sample comprises event features, historical processing departments and multiple processing performance indicators, and a training data set is constructed according to the data samples; step 2, training an adaptation degree prediction model based on the data set, learning a mapping relation from the combination of the event characteristics and the disposal departments to each performance indicator through the adaptation degree prediction model, and predicting each performance result possibly generated by any combination of the event and the disposal departments; step 3, when a new event arrives, predicting performance results generated by multiple performance indicators of each candidate department by using the fitness prediction model, and establishing a multi-objective optimization model; 4, solving the multi-objective optimization model by adopting a multi-objective genetic algorithm to obtain a group of Pareto optimal allocation schemes; and 5, selecting a final allocation scheme from the Pareto optimal solution set according to a preset strategy, and executing the final allocation scheme.

Description

TECHNICAL FIELD

[0001] The application discloses an event automatic distribution method and system based on historical treatment information and a multi-target genetic algorithm, and relates to the technical field of information technology and smart city management. BACKGROUND

[0002] Modern city governance and enterprise customer service centers have a large number of events such as citizen complaints, facility repair, and emergencies that need to be handled every day. These events need to be quickly and accurately distributed to the corresponding handling departments. The existing distribution mode highly depends on manual dispatchers to make judgments based on event descriptions and personal experience, which has obvious drawbacks:

[0003] 1) Low efficiency: manual processing speed is difficult to cope with a large number of events, which is easy to cause backlog;

[0004] 2) Strong subjectivity: the distribution quality is unstable, and there is a lack of unified and objective standards;

[0005] 3) Difficult to optimize as a whole: it is difficult to simultaneously consider handling efficiency, cost, satisfaction, workload balance and other multiple targets;

[0006] 4) Difficulty in knowledge inheritance: the experience of excellent dispatchers is difficult to quantify, solidify and replicate.

[0007] Existing automatic distribution systems are mostly based on rule engines and match through pre-set "if-then" rules. Although the efficiency is improved, the rule system has poor flexibility and cannot handle complex and fuzzy situations. Moreover, the rule maintenance cost is high, and it lacks the ability to self-learn and self-optimize from historical data. SUMMARY

[0008] The application provides an event automatic distribution method and system based on historical treatment information and a multi-target genetic algorithm to solve the problems of the prior art, which is used for intelligent distribution of event handling tasks in the fields of public service hotlines, city operation and management platforms, and emergency command centers.

[0009] The specific scheme provided by the application is as follows:

[0010] The application provides an event automatic distribution method based on historical treatment information and a multi-target genetic algorithm, comprising:

[0011] Step 1: Extract event data samples of completed handling from a historical event library, each data sample containing event features, historical handling departments, and multiple handling performance indicators, and construct a training data set according to the data samples;

[0012] Step 2: Train an adaptation prediction model based on the dataset. The adaptation prediction model learns the mapping relationship between the combination of event features and handling departments and various performance indicators, and predicts the possible performance results of any combination of event and handling department.

[0013] Step 3: When a new event arrives, use the fitness prediction model to predict the performance results of multiple performance indicators for each candidate department, and establish a multi-objective optimization model.

[0014] Step 4: Use a multi-objective genetic algorithm to solve the multi-objective optimization model and obtain a set of Pareto optimal allocation schemes;

[0015] Step 5: Select the final allocation scheme from the Pareto optimal solution set according to the preset strategy and execute it.

[0016] Furthermore, in step 1 of the aforementioned automatic event allocation method based on historical handling information and a multi-objective genetic algorithm, the event features extracted from each data sample include event type, level, location, time, and a keyword vector of the problem description processed by Natural Language Processing (NLP).

[0017] The performance indicators for each data sample include total processing time, user feedback satisfaction, processing cost, and whether the processing time exceeded the limit.

[0018] Furthermore, in step 3 of the aforementioned automatic event allocation method based on historical disposal information and a multi-objective genetic algorithm, a multi-objective optimization model is established, including:

[0019] When a new event X_new arrives, there are N candidate departments {D1,D2,...,Dn};

[0020] Define decision variables: Define an N-dimensional 0-1 vector [a1, a2, ..., a...] n ], where a i =1 indicates that the event is assigned to department i, and sum(a) = 1, which satisfies the condition that sum(a) = 1. i ) = 1;

[0021] Creating objective functions: Based on the fitness prediction model, several objectives that need to be optimized are generated, including:

[0022] Minimize F_time(X_new,Di) means minimizing the prediction time.

[0023] Maximize F_satisfaction(X_new,Di) means maximizing the predicted satisfaction level.

[0024] Minimize F_cost(X_new,Di) means minimizing the prediction cost.

[0025] Minimize|workload_i+F_workload(X_new,Di)-avg_workload| represents minimizing department workload deviation;

[0026] Set constraints: including department function scope constraints and department state constraints.

[0027] Further, step 4 of the event automatic allocation method based on historical treatment information and multi-objective genetic algorithm adopts a non-dominated sorting genetic algorithm NSGA-II with an elitist strategy to solve the multi-objective optimization model, including:

[0028] Encoding and initialization: integer encoding is adopted, the length of the chromosome is 1, the gene value represents the number of the department, and an initial population of size M is randomly generated,

[0029] Evaluation and sorting: the values of multiple objective functions corresponding to each individual in the population are calculated, each individual is an allocation scheme, and the individuals are sorted according to the objective function values, and divided into multiple Pareto fronts,

[0030] Calculate the crowding degree: calculate the crowding degree of individuals in the same non-dominated layer in the objective space to measure the density of solutions around the individuals in the same non-dominated layer,

[0031] Selection, crossover and mutation operation: based on the non-dominated level and the crowding degree, the tournament selection operator is used to select the parent individuals, and the crossover and mutation operations are performed on the parent individuals to generate the offspring population,

[0032] Merge and selection: merge the parent population and the offspring population, perform non-dominated sorting and crowding degree calculation on the merged population, and select the best M individuals to form a new parent population,

[0033] Until termination and output: repeat the evaluation and sorting process until the maximum evolution generation is reached, and output the first non-dominated front in the final population, which is the Pareto optimal solution set.

[0034] Further, step 5 of the event automatic allocation method based on historical treatment information and multi-objective genetic algorithm performs strategy presetting, including:

[0035] Preset artificial decision: show 2-3 optimal schemes and performance effects to the dispatcher on the man-machine interaction interface, and let the dispatcher decide which scheme to use as the final allocation scheme;

[0036] Preset automatic decision: according to the pre-configured weight vector, the weighted sum of the predicted values of multiple objectives is calculated, and the scheme with the highest comprehensive score is selected as the final allocation scheme.

[0037] This invention also provides an automatic event allocation system based on historical disposal information and a multi-objective genetic algorithm, including an extraction module, a fitness prediction model management module, a multi-objective optimization model management module, a calculation module, and an execution module.

[0038] The extraction module extracts data samples of completed events from the historical event database. Each data sample contains event characteristics, historical handling departments, and multiple handling performance indicators. A training dataset is constructed based on the data samples.

[0039] The fit prediction model management module trains a fit prediction model based on the dataset. Through the fit prediction model, it learns the mapping relationship between the combination of event characteristics and handling departments and various performance indicators, and predicts the possible performance results of any combination of event and handling department.

[0040] When a new event arrives, the multi-objective optimization model management module uses the fitness prediction model to predict the performance results of multiple performance indicators for each candidate department and establishes a multi-objective optimization model.

[0041] Step 4: Use a multi-objective genetic algorithm to solve the multi-objective optimization model and obtain a set of Pareto optimal allocation schemes;

[0042] Step 5: Select the final allocation scheme from the Pareto optimal solution set according to the preset strategy and execute it.

[0043] Furthermore, the event features extracted by the extraction module of the event automatic allocation system based on historical handling information and multi-objective genetic algorithm in each data sample include event type, level, location, time, and a keyword vector of the problem description processed by natural language processing (NLP).

[0044] The performance indicators for each data sample include total processing time, user feedback satisfaction, processing cost, and whether the processing time exceeded the limit.

[0045] Furthermore, the multi-objective optimization model management module of the event automatic allocation system based on historical handling information and multi-objective genetic algorithm establishes a multi-objective optimization model, including:

[0046] When a new event X_new arrives, there are N candidate departments {D1,D2,...,Dn};

[0047] Define decision variables: Define an N-dimensional 0-1 vector [a1, a2, ..., a...] n ], where a i =1 indicates that the event is assigned to department i, and sum(a) = 1, which satisfies the condition that sum(a) = 1. i ) = 1;

[0048] Create objective functions: based on the fitness prediction model, generate multiple objectives to be optimized, including:

[0049] Minimize F_time(X_new, Di) means minimizing the predicted time consumption,

[0050] Maximize F_satisfaction(X_new, Di) means maximizing the predicted satisfaction,

[0051] Minimize F_cost(X_new, Di) means minimizing the predicted cost,

[0052] Minimize |workload_i+F_workload(X_new, Di)-avg_workload| means minimizing the department workload deviation;

[0053] Set constraints: including department function range constraints and department state constraints.

[0054] Further, the computing module of the event automatic distribution system based on historical treatment information and multi-objective genetic algorithm adopts a non-dominated sorting genetic algorithm NSGA-II with an elite strategy to solve the multi-objective optimization model, including:

[0055] Encoding and initialization: integer encoding is adopted, the chromosome length is 1, the gene value represents the number of the department, and an initial population of size M is randomly generated,

[0056] Evaluation and sorting: calculate the multiple objective function values corresponding to each individual in the population, each individual is a distribution scheme, sort the individuals according to the objective function values, and divide them into multiple Pareto Fronts,

[0057] Calculate the crowding degree: calculate the crowding degree of individuals in the same non-dominated layer in the objective space to measure the density of solutions around the individuals in the same non-dominated layer,

[0058] Selection, crossover and mutation operations: based on the non-dominated level and the crowding degree, the tournament selection operator is used to select the parent individuals, and the crossover and mutation operations are performed on the parent individuals to generate the offspring population,

[0059] Merge and selection: merge the parent population and the offspring population, perform non-dominated sorting and crowding degree calculation on the merged population, and select the best M individuals to form a new parent population,

[0060] Until termination and output: repeat the process of evaluation and sorting to merging and selection until reaching the maximum evolution number, and output the first non-dominated front in the final population, that is, the Pareto optimal solution set.

[0061] Further, the execution module of the event automatic distribution system based on historical treatment information and a multi-objective genetic algorithm performs strategy presetting, including:

[0062] Preset artificial decision: 2-3 optimal schemes and performance effects are displayed to the dispatcher on the man-machine interaction interface, and the dispatcher determines which scheme is used as the final distribution scheme;

[0063] Preset automatic decision: according to a pre-configured weight vector, the predicted values of multiple targets are weighted and summed, and the scheme with the highest comprehensive score is selected as the final distribution scheme.

[0064] The present application has the following advantages:

[0065] (1) High efficiency and automation: realize second-level automatic distribution, greatly improve efficiency, and liberate manpower.

[0066] (2) Decision-making scientific: use multi-objective optimization theory, consider multiple performance targets, avoid one-sidedness of single target, and realize global optimization.

[0067] (3) Continuous self-learning: can continuously learn and update the prediction model from historical data, so as to continuously evolve the distribution strategy and become more intelligent.

[0068] (4) Flexible configuration: the optimization target can be conveniently added, deleted or adjusted in weight according to the change of management strategy, and has strong adaptability. BRIEF DESCRIPTION OF DRAWINGS

[0069] Figure 1 is a schematic diagram of the application process.

[0070] Figure 2 is a solution flowchart of the multi-objective genetic algorithm (NSGA-II). DETAILED DESCRIPTION

[0071] The present application will be further described below in combination with the drawings and specific embodiments, so that those skilled in the art can better understand the present application and implement it, but the embodiments are not limiting to the present application.

[0072] Embodiment 1

[0073] The present application provides an event automatic distribution method based on historical treatment information and a multi-objective genetic algorithm, including:

[0074] Step 1: Extract event data samples that have been handled from the historical event database. Each data sample contains event features X, historical handling department D, and multiple handling performance indicators Z. Construct a training dataset based on the data samples.

[0075] The event features X extracted from each data sample include event type, level, location, time, and a vector of keywords describing the problem after processing by Natural Language Processing (NLP).

[0076] The performance indicators Z for each data sample include total processing time, user feedback satisfaction, processing cost, and whether the processing time exceeded the limit.

[0077] Step 2: Train one or more fitness prediction models based on the dataset. These models then learn the mapping relationship between the combination of event features X and the handling department D, and various performance indicators.

[0078] F(X,D)->Z

[0079] This invention predicts the various performance outcomes that may arise from any combination of events and handling departments. Preferably, this invention employs multi-task learning or separately trained models such as LightGBM or XGBoost.

[0080] Step 3: When a new event arrives, use the fitness prediction model to predict the performance results of multiple performance indicators for each candidate department, and establish a multi-objective optimization model.

[0081] The establishment of a multi-objective optimization model includes:

[0082] When a new event X_new arrives, there are N candidate departments {D1,D2,...,Dn};

[0083] Define decision variables: Define an N-dimensional 0-1 vector [a1, a2, ..., a...] n ], where a i =1 indicates that the event is assigned to department i, and sum(a) = 1, which satisfies the condition that sum(a) = 1. i ) = 1;

[0084] Creating objective functions: Based on the fitness prediction model, several objectives that need to be optimized are generated, including:

[0085] Minimize F_time(X_new,Di) means minimizing the prediction time.

[0086] Maximize F_satisfaction(X_new,Di) means maximizing the predicted satisfaction level.

[0087] Minimize F_cost(X_new,Di) means minimizing the prediction cost.

[0088] Minimize | workload_i + F_workload(X_new, Di) - avg_workload | represents minimizing department workload deviation;

[0089] Setting constraints: including department function scope constraints and department state constraints.

[0090] Step 4: Solve the multi-objective optimization model using a multi-objective genetic algorithm to obtain a set of Pareto optimal distribution schemes.

[0091] Among them, the non-dominated sorting genetic algorithm with elitist strategy NSGA-II is used to solve the multi-objective optimization model, which can include:

[0092] Encoding and initialization: integer coding is used, the length of the chromosome is 1, the gene value represents the number of departments, and an initial population of size M is randomly generated,

[0093] Evaluation and sorting: calculate the multiple objective function values corresponding to each individual in the population, each individual is a distribution scheme, and sort the individuals according to the objective function values, and divide them into multiple Pareto Fronts,

[0094] Calculate the crowding degree: calculate the crowding degree of individuals in the same non-dominated layer in the objective space to measure the density of solutions around individuals in the same non-dominated layer,

[0095] Selection, crossover and mutation operations: based on non-dominated level and crowding degree, use tournament selection operator to select parent individuals, perform crossover and mutation operations on parent individuals to generate offspring population,

[0096] Merge and selection: merge the parent population and the offspring population, perform non-dominated sorting and crowding degree calculation on the merged population, and select the best M individuals to form a new parent population,

[0097] Until termination and output: repeat the evaluation and sorting process until the maximum evolution generation is reached, and output the first non-dominated front in the final population, which is the Pareto optimal solution set.

[0098] Step 5: Select the final distribution scheme from the Pareto optimal solution set according to the preset strategy and execute it.

[0099] Among them, the strategy preset can include:

[0100] Preset artificial decision: show 2-3 optimal schemes and performance effects to the dispatcher on the human-computer interaction interface, and let the dispatcher decide which scheme to use as the final distribution scheme;

[0101] Pre-set automatic decision: according to the pre-configured weight vector, the predicted values of multiple targets are weighted and summed, and the scheme with the highest comprehensive score is selected as the final distribution scheme.

[0102] Embodiment 2

[0103] The application also provides an event automatic distribution system based on historical treatment information and a multi-target genetic algorithm, comprising an extraction module, a fitness prediction model management module, a multi-target optimization model management module, a calculation module and an execution module,

[0104] The extraction module extracts event data samples of completed treatment from a historical event library, each data sample containing event characteristics, historical treatment departments and multiple treatment performance indicators, and a training data set is constructed according to the data samples;

[0105] The fitness prediction model management module trains a fitness prediction model based on the data set, learns the mapping relationship from the combination of event characteristics and treatment departments to each performance indicator through the fitness prediction model, and predicts the performance results of each performance indicator that may be produced by the combination of any event and treatment department;

[0106] The multi-target optimization model management module predicts the performance results of multiple performance indicators produced by each candidate department when a new event arrives, and establishes a multi-target optimization model;

[0107] Step 4: A multi-target genetic algorithm is used to solve the multi-target optimization model to obtain a group of Pareto optimal distribution schemes;

[0108] Step 5: The final distribution scheme is selected from the Pareto optimal solution set according to the pre-set strategy and executed.

[0109] The information interaction and execution process between the modules in the above system, etc. are based on the same concept as the method embodiments of the application, and the specific content can be referred to the description in the method embodiments of the application, which will not be repeated here.

[0110] Similarly, the benefits of the system of the application are:

[0111] (1) High efficiency and automation: realize second-level automatic distribution, greatly improve efficiency and liberate manpower.

[0112] (2) Scientific decision-making: use multi-target optimization theory, consider multiple performance targets, avoid one-sidedness of single target, and achieve global optimization.

[0113] (3) Continuous self-learning: can continuously learn and update the prediction model from historical data, so as to continuously evolve the distribution strategy and become more intelligent.

[0114] (4) Flexible configuration: the optimization target can be easily added, deleted or adjusted in weight according to the change of management strategy, and has strong adaptability.

[0115] It should be noted that not all steps and modules in the above processes and system structures are necessary, and some steps or modules can be ignored according to actual needs. The execution order of each step is not fixed and can be adjusted as needed. The system structure described in the above embodiments can be a physical structure or a logical structure, that is, some modules can be implemented by the same physical entity, or some modules can be implemented by multiple physical entities, or can be implemented by some components in multiple independent devices together.

[0116] The above-described embodiments are only preferred embodiments of the present application, and the protection scope of the present application is not limited thereto. Any equivalent replacement or transformation made by those skilled in the art based on the present application is within the protection scope of the present application. The protection scope of the present application is subject to the claims.

Claims

1. An event automatic distribution method based on historical treatment information and multi-objective genetic algorithm, characterized in that The method comprises the following steps: Step 1: extracting event data samples of completed disposal from a historical event library, each data sample containing event features, historical disposal departments and multiple disposal performance indicators, and constructing a training data set according to the data samples; Step 2: training an adaptation prediction model based on the data set, learning the mapping relationship from the combination of event features and disposal departments to each performance indicator through the adaptation prediction model, and predicting the performance results of each performance indicator generated by the combination of any event and disposal department; Step 3: when a new event arrives, predicting the performance results of multiple performance indicators generated by each candidate department by using the adaptation prediction model, and establishing a multi-objective optimization model; Step 4: solving the multi-objective optimization model by using a multi-objective genetic algorithm to obtain a set of Pareto optimal distribution schemes; Step 5: selecting a final distribution scheme from the Pareto optimal solution set according to a preset strategy and executing the final distribution scheme.

2. The event automatic distribution method based on historical treatment information and multi-objective genetic algorithm according to claim 1, characterized in that The event features in each data sample extracted in step 1 include event type, level, location, time, and problem description keyword vector processed by a natural language processing method (NLP), The disposal performance indicators in each data sample include disposal total time, user revisit satisfaction, disposal cost, and whether overtime.

3. The event automatic distribution method based on historical treatment information and multi-objective genetic algorithm according to claim 1, The feature is that in step 3, a multi-objective optimization model is established, including: When a new event X_new arrives, there are N candidate departments {D1, D2,..., Dn}; Define decision variables: Define an N-dimensional 0-1 vector [a1, a2,..., aN], where ai = 1 means assign event i to department j, and satisfy sum(ai) = 1. n ] i i ​​ Creating a target function: generating multiple optimization targets based on the adaptation prediction model, including: Minimize F_time(X_new, Di) represents minimizing the predicted time, Maximize F_satisfaction(X_new, Di) represents maximizing the predicted satisfaction, Minimize F_cost(X_new, Di) represents minimizing the predicted cost, Minimize|workload_i+F_workload(X_new, Di)-avg_workload| represents minimizing the department workload deviation; Setting constraint conditions: including department function range constraints and department state constraints.

4. The event automatic distribution method based on historical disposal information and a multi-objective genetic algorithm according to claim 1, wherein in step 4, a non-dominated sorting genetic algorithm II (NSGA-II) with an elitist strategy is used to solve the multi-objective optimization model, including: Encoding and initialization: integer encoding is used, the chromosome length is 1, the gene value represents the number of the department, and an initial population with a size of M is randomly generated, Evaluation and sorting: the multiple target function values corresponding to each individual in the population are calculated, each individual is a distribution scheme, the individuals are non-dominantly sorted according to the target function values, and the individuals are divided into multiple Pareto fronts, Calculate the crowding degree: the crowding degree of the individuals in the same non-dominant layer in the target space is calculated to measure the density of the solutions around the individuals in the same non-dominant layer, Selection, crossover and mutation operations are performed: parent individuals are selected by using a tournament selection operator based on non-dominated rank and crowding degree, crossover and mutation operations are performed on the parent individuals to generate a child population, Merging and selection are performed: the parent population and the child population are merged, non-dominated sorting and crowding degree calculation are performed on the merged population, and the best M individuals are selected to form a new parent population, Until termination and output: repeat the process of evaluation and sorting to merging and selection until the maximum evolution generation is reached, and output the first non-dominated front in the final population, that is, the Pareto optimal solution set.

5. The method of claim 1, wherein the method is characterized by Strategy presetting is performed in step 5, including: Manual decision presetting: 2-3 optimal schemes and performance effects are displayed to the dispatcher on the man-machine interaction interface, and the dispatcher decides which scheme to adopt as the final distribution scheme; Automatic decision presetting: according to a pre-configured weight vector, the predicted values of multiple targets are weighted and summed, and the scheme with the highest comprehensive score is selected as the final distribution scheme.

6. An event automatic distribution system based on historical treatment information and multi-objective genetic algorithm, characterized in that The system comprises an extraction module, an adaptability prediction model management module, a multi-objective optimization model management module, a calculation module and an execution module, The extraction module extracts event data samples of completed disposal from a historical event library, each data sample comprising event features, historical disposal departments and multiple disposal performance indicators, and a training data set is constructed according to the data samples; The adaptability prediction model management module trains an adaptability prediction model based on the data set, learns the mapping relationship from the combination of event features and disposal departments to each performance indicator through the adaptability prediction model, and predicts the performance results of each performance indicator that may be generated by the combination of any event and disposal department; When a new event arrives, the multi-objective optimization model management module predicts the performance results of multiple performance indicators generated by each candidate department by using the adaptability prediction model, and establishes a multi-objective optimization model; Step 4: a multi-objective genetic algorithm is used to solve the multi-objective optimization model to obtain a group of Pareto optimal distribution schemes; Step 5: the final distribution scheme is selected from the Pareto optimal solution set according to the preset strategy and executed.

7. The event automatic distribution system based on historical treatment information and multi-objective genetic algorithm according to claim 6, characterized in that The event features in each data sample extracted by the extraction module include event type, level, location, time, and problem description keyword vector processed by a natural language processing method NLP, The disposal performance indicators in each data sample include disposal total time, user revisit satisfaction, disposal cost, and whether overtime.

8. The event automatic distribution system based on historical treatment information and multi-objective genetic algorithm according to claim 6, characterized in that The multi-objective optimization model management module establishes a multi-objective optimization model, including: When a new event X_new arrives, there are N candidate departments {D1, D2,..., Dn}; Define decision variables: Define an N-dimensional 0-1 vector [a1, a2,..., aN] where a n = 1 means assign event to department i, and satisfy sum(a i ) = 1. i ], where a n = 1 means assign event to department i, and satisfy sum(a i ) = 1. i ], where a n = 1 means assign event to department i, and satisfy sum(a i ) = Create objective function: based on the adaptability prediction model, generate multiple objectives to be optimized, including: Minimize F_time(X_new, Di) represents minimizing predicted time, Maximize F_satisfaction(X_new, Di) represents maximizing predicted satisfaction, Minimize F_cost(X_new, Di) represents minimizing predicted cost, Minimize | workload_i + F_workload(X_new, Di) - avg_workload | represents minimizing the department workload deviation; Set constraints: including department function scope constraints and department state constraints.

9. The event automatic distribution system based on historical treatment information and multi-objective genetic algorithm according to claim 6, characterized in that The calculation module uses the non-dominated sorting genetic algorithm with elitism NSGA-II to solve the multi-objective optimization model, including: Coding and initialization: integer coding is used, the chromosome length is 1, and the gene value represents the number of the department. An initial population of size M is randomly generated, Evaluation and sorting: calculate the multiple objective function values corresponding to each individual in the population, each individual is a distribution scheme, and sort the individuals according to the objective function values, and divide them into multiple Pareto Fronts, Calculate the crowding degree: calculate the crowding degree of individuals in the same non-dominated layer in the objective space to measure the density of solutions around the individuals in the same non-dominated layer, Selection, crossover and mutation operations: based on the non-dominated level and the crowding degree, the tournament selection operator is used to select the parent individuals, and the crossover and mutation operations are performed on the parent individuals to generate the offspring population, Merge and select: merge the parent population and the offspring population, and perform non-dominated sorting and crowding degree calculation on the merged population, and select the best M individuals to form a new parent population, Until termination and output: repeat the evaluation and sorting process until the maximum evolution generation is reached, and output the first non-dominated front in the final population, which is the Pareto optimal solution set.

10. The event automatic distribution system based on historical treatment information and multi-objective genetic algorithm according to claim 6, characterized in that The execution module presets the strategy, including: Preset artificial decision: show 2-3 optimal schemes and performance effects to the dispatcher on the human-computer interaction interface, and let the dispatcher decide which scheme to use as the final distribution scheme; Preset automatic decision: according to the pre-configured weight vector, the predicted values of multiple objectives are weighted and summed, and the scheme with the highest comprehensive score is selected as the final distribution scheme.

Images