Insurance loan trial calculation optimization method based on intelligent adaptive load scheduling

Abstract

The invention discloses an insurance loan trial calculation optimization method based on intelligent adaptive load scheduling. The method comprises the following steps: obtaining a to-be-trial calculation insurance policy set; and on the basis of the to-be-tried insurance policy set and historical load data of the system, utilizing a machine learning model to predict a future task load, and calculating a calculation demand of each insurance policy trial task. And generating a task priority sequence for the insurance policy trial task based on the predicted task load and calculation demand. And the system executes insurance policy trial calculation tasks according to priorities through the thread pool, dynamically adjusts parameters of the thread pool in combination with a reinforcement learning algorithm, and optimizes system resource configuration. For task execution, the short-time-consuming tasks preferentially return results, and the long-time-consuming tasks are returned in an incremental summarization mode. And the system integrates results of all insurance policy trial calculation tasks to generate a structured loan trial calculation report. According to the method, the processing efficiency, the response speed and the resource utilization rate of the system can be remarkably improved, and the overall performance of insurance loan trial calculation is optimized especially in a high-concurrency scene.

Description

Technical Field

[0001] This invention belongs to the field of insurance loan trial calculation technology, specifically relating to an insurance loan trial calculation optimization method based on intelligent adaptive load scheduling. Background Technology

[0002] With the rapid development of the insurance industry, insurance loan business has played a vital role in the diversification of insurance products and the innovation of financial services. Insurance loan trial calculation, as a prerequisite for insurance loan business, typically requires detailed calculation and analysis of the customer's insurance policies. This process involves complex calculations across multiple dimensions based on the characteristics of the customer's insurance products and loan conditions to determine the loanable amount and other relevant parameters. Due to the large number and computational complexity of policy trial calculation tasks, traditional methods often face problems such as low processing efficiency, uneven system load, and significant resource waste. Especially in high-concurrency scenarios, system performance bottlenecks become a key factor limiting its further development.

[0003] In existing insurance systems, when a customer requests a loan calculation using their insurance policies, the system typically needs to obtain information on all of the customer's policies and then sequentially call the calculation service for each policy. However, when a customer has a large number of policies, this sequential processing leads to a significant increase in interface response time and a sharp decline in performance. In extreme cases, it may cause interface timeouts or service unavailability, severely impacting user experience and system business processing efficiency, becoming a technical bottleneck that urgently needs to be addressed.

[0004] In the prior art, Chinese patent CN120743705A discloses an adaptive data processing optimization method, apparatus, device and medium, including: collecting the operation data, host performance data, network status data and historical task data of the target data source, constructing an analysis model by combining recovery parameters and strategy preferences, predicting the task load status, resource consumption and execution time, generating task execution strategy and completing task scheduling and execution monitoring, and collecting execution feedback data for iterative optimization of the analysis model.

[0005] However, this method has the following limitations: 1. Existing systems still use a serial computing approach, which cannot effectively improve processing efficiency in high-concurrency scenarios. When a customer has a large number of insurance policies, the computation time for a single task will increase significantly, leading to a longer overall system response time and a sharp decline in user experience. 2. Traditional insurance loan calculation systems typically rely on fixed thread pool parameters (such as the number of core threads, the maximum number of threads, etc.) and do not flexibly adjust them according to the actual system load. The system cannot adapt to load fluctuations, resulting in excessive consumption of computing resources under high load and underutilization under low load, thereby reducing resource allocation efficiency and system stability. 3. Existing systems often ignore task priority and computational requirements, which may lead to the delayed execution of high-priority critical tasks. 4. Traditional insurance loan calculation systems often encounter computational bottlenecks when facing high-concurrency tasks. Under high concurrency, the system's task processing efficiency decreases significantly. Due to the lack of a flexible resource allocation mechanism, the system cannot dynamically adjust computing resources according to real-time load conditions, resulting in uneven resource allocation and ultimately affecting system performance and processing efficiency.

[0006] Therefore, there is an urgent need for a technical solution that can address the performance bottlenecks, resource waste, and unreasonable load scheduling issues existing in current insurance loan calculation systems. Summary of the Invention

[0007] The purpose of this invention is to overcome the shortcomings of the existing technology and provide an insurance loan trial calculation optimization method based on intelligent adaptive load scheduling.

[0008] The objective of this invention can be achieved through the following technical solutions: This invention provides an insurance loan trial calculation optimization method based on intelligent adaptive load scheduling, comprising the following steps: The system retrieves basic information on all insurance policies under a customer's name through batch database queries, and excludes policies that do not meet the loan calculation conditions according to preset rules, thereby generating a set of policies to be calculated. Based on the set of policies to be tested and the system's historical load data, a machine learning model is used to predict future task load and calculate the computational requirements for each policy test task. Based on the predicted future task load and computing requirements, generate a task priority sequence for the policy trial calculation task; The thread pool is used to execute the policy calculation task according to the task priority sequence, and the thread pool parameters are dynamically adjusted by combining reinforcement learning algorithm; The processing results of the policy trial calculation task are executed asynchronously, including prioritizing the return of results for short-duration tasks based on the task time estimation, and returning results for long-duration tasks using an incremental aggregation method; The results of all policy trial calculations are integrated, verified, and formatted to generate a structured loan trial calculation report.

[0009] Furthermore, the step of obtaining basic information on all insurance policies under the customer's name through batch database queries, and excluding policies that do not meet the loan calculation conditions according to preset rules, to generate a set of policies to be calculated, specifically includes: Under the customer's unique identifier, basic policy information is obtained through batch database queries, including policy number, policy type, policy status, policy effective date, insured amount, and loan-related information; The retrieved basic policy information is filtered according to the following pre-filtering criteria: Exclude policies that have expired, been surrendered, or been terminated; Exclude policies with existing outstanding loans; Exclude policy types that do not support loan business; The basic information of the policies that meet the pre-filtering criteria is compiled into a set of policies to be tested.

[0010] Furthermore, the machine learning model is a long short-term memory network.

[0011] Furthermore, the training process for the machine learning model includes: Construct a training dataset by using historical system load data and business logs to create sample sequences in time series format. : in, Indicates time t The feature vector includes the number of policies to be tested. System CPU utilization Memory usage Average time spent on historical tasks Number of concurrent tasks ; Indicates time Task load value, referring to the time... The system needs to process a large volume of policy calculation tasks. The length of the time window; The total number of samples; For sample sequences Standardization processing is required; Input the standardized training sample sequence into the machine learning model. Train the system and predict future task loads using the following formula: in, Indicates the predicted future task load; These are the model parameters, including weights and biases; The time after standardization t eigenvectors; Based on the predicted future task load and loss function, the model parameters are... Update the model to obtain the trained machine learning model.

[0012] Furthermore, the loss function is: in, For loss function, Preset weighting coefficients; This is the time decay coefficient.

[0013] Furthermore, the computational requirements for each task specifically include: Treating the trial calculation policy set Each policy To obtain the core characteristics of the policy, including the policy amount, policy type, policy effective date, historical calculation time, and customer level; Calculate the total number of tasks for the policy trial; Using policy characteristics and historical system load data as input, a machine learning model is used to predict the average computation time for each policy's trial calculation task. ; Based on the predicted single-task time Based on the task characteristics and the system's currently available computing resources, calculate the comprehensive computing requirements for each policy task. : in, For the policy The computational requirements indicators , Representing time respectively t The system's CPU and memory usage.

[0014] Furthermore, the step of generating a task priority sequence for the policy trial calculation task based on the predicted future task load and computing requirements specifically includes: For each policy task to be trial calculated Based on its comprehensive computing requirements and business priority indicators Calculate task weights : in, and Preset weighting coefficients; Calculate the total number of tasks for the policy trial; For the policy The computational requirements indicators; Prioritize policy business; Based on the calculated task weights Sort all policy calculation tasks in descending order to generate a task priority sequence, where policy tasks with higher weights have higher priority in the sequence.

[0015] Furthermore, the step of using a thread pool to execute policy calculation tasks according to task priority sequence, and dynamically adjusting thread pool parameters in conjunction with reinforcement learning algorithms, specifically includes: Define thread pool parameters, including: Set the core number of threads in the thread pool. Maximum number of threads and thread lifespan And initialize the reinforcement learning agent; Constructing the reinforcement learning state space and action space includes: Define state vector ,in, For the current moment t Number of pending policy tasks This refers to the system CPU utilization rate. This refers to system memory usage. This represents the current number of active threads. Define action space This indicates the adjustment range of the thread pool parameters; , , These represent the dynamic adjustments to the core thread count, maximum thread count, and thread lifespan, respectively. Based on a task priority sequence, each policy calculation task is submitted to the thread pool for execution in order of priority, and system load status parameters, including CPU utilization, are collected in real time. Memory usage and number of active threads ; Based on the current state of the system As input to the reinforcement learning model; the reinforcement learning model outputs action vectors. And based on the action vector Update thread pool parameters; The thread pool, after updating its parameters, continues to execute the policy calculation task based on the task priority sequence, and records the average task response time. System throughput and CPU utilization ; Average task response time based on records System throughput and CPU utilization Calculate the reinforcement learning reward function, optimize the reinforcement learning model policy based on the calculated reward function, and continue to dynamically adjust the thread pool parameters and optimize the thread pool resource configuration according to the updated reinforcement learning model policy.

[0016] Furthermore, the step based on the action vector The formula for updating thread pool parameters is: in, , , Representing time respectively The core number of threads, the maximum number of threads, and the thread lifespan.

[0017] Furthermore, the reinforcement learning reward function is defined by the following formula: in, For a moment t The reinforcement learning reward function; , , These are the preset weighting coefficients.

[0018] Compared with the prior art, the present invention has the following advantages: (1) In the prior art, insurance loan calculation systems typically employ serial computation when handling a large number of policy calculation tasks, which leads to a significant increase in interface response time. When a customer has a large number of policies, system performance drops sharply, and in extreme cases, interface timeouts or service unavailability may occur, severely impacting user experience and system processing efficiency. To address this issue, this invention improves system processing efficiency and stability by executing policy calculation tasks according to task priority sequence and dynamically adjusting thread pool parameters using reinforcement learning algorithms. This enables more efficient task scheduling, significantly improves system response speed, and avoids performance bottlenecks in high-concurrency scenarios.

[0019] (2) In the prior art, traditional insurance loan calculation systems generally use fixed thread pool parameters, which cannot be dynamically adjusted according to the actual system load, easily leading to waste or insufficient computing resources, especially under high load, which may cause system crashes or performance degradation. This invention, by combining reinforcement learning algorithms, adjusts parameters such as the core number of threads, the maximum number of threads, and the thread lifespan in the thread pool in real time according to the system load status, effectively coping with load fluctuations. This technical feature, by dynamically adjusting the thread pool parameters, not only optimizes resource allocation but also improves the utilization rate of the system's computing resources, avoids excessive resource waste, and ensures the stability of the system under high load conditions.

[0020] (3) In existing technologies, the prediction of computational requirements for policy trial calculation tasks largely relies on manual experience or static settings, lacking an intelligent load prediction mechanism. This results in inaccurate computational requirements for the trial calculation tasks, thereby affecting the efficiency and accuracy of task scheduling. This invention uses machine learning models, particularly the Long Short-Term Memory (LSTM) algorithm, to predict the computational requirements of each policy trial calculation task. Combined with historical load data and business characteristics, it accurately calculates the computational requirements of each task. This technical feature optimizes the task scheduling strategy through accurate computational requirement prediction, avoids unreasonable resource allocation, and improves the overall efficiency and processing capacity of the system.

[0021] (4) Traditional task scheduling methods typically do not consider task priority and computational requirements, which may lead to delays in the execution of critical tasks, thereby affecting the normal operation of business. Existing systems lack intelligent priority management and load scheduling mechanisms, making it impossible to achieve reasonable resource allocation among different tasks. This invention combines the computational requirements and business priorities of the policy calculation task and prioritizes tasks based on task weights to ensure that critical tasks can be executed in a timely manner. Through this technical feature, this invention can dynamically adjust the task execution order, ensuring the timely processing of important tasks and improving business processing efficiency and system response speed.

[0022] (5) Traditional insurance loan calculation systems often face computational bottlenecks under high concurrency, resulting in low task processing efficiency. This invention combines task scheduling with dynamic adjustment of thread pool parameters, monitors the system's load status (including CPU utilization, memory utilization, etc.) in real time, and dynamically optimizes the thread pool configuration based on these parameters. This allows for automatic expansion of computing resources under high load and reduction of computing resource consumption under low load. This technical feature, by adaptively adjusting computing resources according to the system's real-time load, not only improves the system's concurrent processing capability but also effectively reduces system resource consumption, ensuring efficient task processing and system stability. Attached Figure Description

[0023] Figure 1 This is a flowchart of the insurance loan trial calculation optimization method according to an embodiment of the present invention; Figure 2 This is a model diagram of the insurance loan trial calculation optimization system according to an embodiment of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0025] Example 1: This embodiment provides an insurance loan trial calculation optimization method based on intelligent adaptive load scheduling, such as... Figure 1 As shown, it includes the following steps: Step S1: Obtain basic information on all insurance policies under the customer's name through batch database query, and exclude policies that do not meet the loan calculation conditions according to preset rules, generating a set of policies to be calculated, specifically including: Under the customer's unique identifier, basic policy information is obtained through batch database queries, including policy number, policy type, policy status, policy effective date, insured amount, and loan-related information; The retrieved basic policy information is filtered according to the following pre-filtering criteria: Exclude policies that have expired, been surrendered, or been terminated; Exclude policies with existing outstanding loans; Exclude policy types that do not support loan business; The basic information of the policies that meet the pre-filtering criteria is compiled into a set of policies to be tested.

[0026] Step S2: Based on the set of policies to be calculated and the system's historical load data, use a machine learning model to predict future task load and calculate the computational requirements for each policy calculation task. The machine learning model is a long short-term memory network. The training process for machine learning models includes: Construct a training dataset by using historical system load data and business logs to create sample sequences in time series format. : in, Indicates time t The feature vector includes the number of policies to be tested. System CPU utilization Memory usage Average time spent on historical tasks Number of concurrent tasks ; Indicates time Task load value, referring to the time... The system needs to process a large volume of policy calculation tasks. The length of the time window; The total number of samples; For sample sequences Standardization processing is required; Input the standardized training sample sequence into the machine learning model. Train the system and predict future task loads using the following formula: in, Indicates the predicted future task load; These are the model parameters, including weights and biases; The time after standardization t eigenvectors; Based on the predicted future task load and loss function, the model parameters are... Update the model to obtain the trained machine learning model.

[0027] The loss function is: in, For loss function, Preset weighting coefficients; This is the time decay coefficient.

[0028] Step S2 utilizes machine learning models, particularly Long Short-Term Memory (LSTM) networks, to predict future task loads and calculate the computational requirements for each policy trial calculation task. The core technical feature of this step lies in constructing a training dataset based on historical load data and business logs in a time-series format, and then training and predicting using an LSTM model. By learning from historical data, the system can intelligently capture the complex relationship between task load and resource usage, thereby achieving accurate load prediction. This process not only improves the intelligence level of task scheduling but also optimizes the allocation of future computing resources based on the prediction results, avoiding resource waste or insufficiency in traditional methods and improving task processing efficiency and system response speed. By optimizing the configuration of computing resources, this invention can effectively cope with high load fluctuations, ensuring the stability and efficiency of the system in high-concurrency scenarios, thereby significantly improving user experience and system performance.

[0029] Step S3: Generate a task priority sequence for the policy trial calculation task based on the predicted future task load and computing requirements; Calculate the computational requirements for each task, specifically including: Treating the trial calculation policy set Each policy To obtain the core characteristics of the policy, including the policy amount, policy type, policy effective date, historical calculation time, and customer level; Calculate the total number of tasks for the policy trial; Using policy characteristics and historical system load data as input, a machine learning model is used to predict the average computation time for each policy's trial calculation task. ; Based on the predicted single-task time Based on the task characteristics and the system's currently available computing resources, calculate the comprehensive computing requirements for each policy task. : in, For the policy The computational requirements indicators , Representing time respectively t The system's CPU and memory usage.

[0030] For each policy task to be trial calculated Based on its comprehensive computing requirements and business priority indicators Calculate task weights : in, and Preset weighting coefficients; Calculate the total number of tasks for the policy trial; For the policy The computational requirements indicators; Prioritize policy business; Based on the calculated task weights Sort all policy calculation tasks in descending order to generate a task priority sequence, where policy tasks with higher weights have higher priority in the sequence.

[0031] Step S3 generates a priority sequence for policy trial calculation tasks by combining predicted task load, characteristics of each policy, system load, and task priority, thus optimizing task scheduling and resource allocation. This process first predicts the computational requirements of each policy (including task computation time and resource consumption) based on historical data and machine learning models. Then, it determines the weight of each task by comprehensively considering computational requirements and business priorities. Tasks are ranked according to their weights to ensure that critical tasks are executed first. By combining business characteristics and system load data, the computational requirements of each task can be predicted more accurately, avoiding the neglect of differences between tasks in traditional methods and ensuring reasonable allocation of computing resources. Secondly, by prioritizing tasks, the system can process high-priority policy tasks first, ensuring timely response to critical business needs. Ultimately, the entire task scheduling process is more intelligent and efficient, significantly improving the system's task processing capacity and response speed under high load and high concurrency environments, enhancing user experience, and avoiding resource waste.

[0032] Step S4: Execute the policy calculation task using a thread pool according to task priority sequence, and dynamically adjust the thread pool parameters using reinforcement learning algorithms. Specifically, this includes: Define thread pool parameters, including: Set the core number of threads in the thread pool. Maximum number of threads and thread lifespan And initialize the reinforcement learning agent; Constructing the reinforcement learning state space and action space includes: Define state vector ,in, For the current moment t Number of pending policy tasks This refers to the system CPU utilization rate. This refers to system memory usage. This represents the current number of active threads. Define action space This indicates the adjustment range of the thread pool parameters; , , These represent the dynamic adjustments to the core thread count, maximum thread count, and thread lifespan, respectively. Based on a task priority sequence, each policy calculation task is submitted to the thread pool for execution in order of priority, and system load status parameters, including CPU utilization, are collected in real time. Memory usage and number of active threads ; Based on the current state of the system As input to the reinforcement learning model; the reinforcement learning model outputs action vectors. And based on the action vector Update thread pool parameters; The thread pool, after updating its parameters, continues to execute the policy calculation task based on the task priority sequence, and records the average task response time. System throughput and CPU utilization ; Average task response time based on records System throughput and CPU utilization Calculate the reinforcement learning reward function, optimize the reinforcement learning model policy based on the calculated reward function, and continue to dynamically adjust the thread pool parameters and optimize the thread pool resource configuration according to the updated reinforcement learning model policy.

[0033] Based on action vectors The formula for updating thread pool parameters is: in, , , Representing time respectively The core number of threads, the maximum number of threads, and the thread lifespan.

[0034] The reinforcement learning reward function is defined by the following formula: in, For a moment t The reinforcement learning reward function; , , These are the preset weighting coefficients.

[0035] Step S4 executes policy calculation tasks according to task priority using a thread pool, and dynamically adjusts the thread pool parameters using a reinforcement learning algorithm, achieving more flexible and efficient resource management. Specifically, this step first defines the core parameters of the thread pool (such as the core number of threads, the maximum number of threads, and thread lifespan), and then uses a reinforcement learning algorithm to dynamically adjust these parameters based on system load during real-time task execution to achieve optimal resource allocation.

[0036] During this process, the system monitors status variables in real time, including the number of pending insurance policy tasks, CPU and memory usage, and the number of active threads. Based on the current status and task priorities, it dynamically adjusts the thread pool configuration. In this way, the system can adaptively adjust computing resources under different load conditions, avoiding resource shortages under high load or resource waste under low load, thereby improving task execution efficiency.

[0037] Through reinforcement learning training and optimization, the system can not only automatically adjust thread pool parameters based on the current load, but also dynamically adjust strategies based on feedback from metrics such as average task response time, system throughput, and CPU utilization, further improving system performance and resource utilization. In this way, the system can efficiently handle large-scale, high-concurrency policy calculation tasks while ensuring timely task processing and system stability.

[0038] By intelligently optimizing thread pool parameters through reinforcement learning, the system can adaptively adjust under varying system loads and task priorities, significantly improving the system's concurrent processing capabilities.

[0039] Dynamically adjusting the core thread count, maximum thread count, and thread lifespan of the thread pool avoids resource waste or insufficiency under traditional fixed parameter configurations.

[0040] By combining real-time feedback optimization strategies, the system can continuously improve resource allocation based on actual load conditions, avoiding over- or under-configuration and ensuring efficient operation and stability.

[0041] In high-concurrency environments, it effectively solves the performance bottleneck problem faced by traditional insurance loan calculation systems, improves user experience, and increases business processing efficiency.

[0042] Step S5: Asynchronously execute the processing results of the policy trial calculation task, including prioritizing the return of results for short-time tasks based on the task time estimation, and returning results for long-time tasks using an incremental aggregation method; Step S6: Integrate, verify, and format the results of all policy trial calculation tasks to generate a structured loan trial calculation report.

[0043] Example 2: This embodiment provides an insurance loan trial calculation optimization system based on intelligent adaptive load scheduling, such as... Figure 2 As shown, it includes: Task prediction module: This module predicts future task load based on the set of policies to be calculated and historical system load data, using machine learning models (such as Long Short-Term Memory networks, LSTM). It constructs a training dataset based on historical system load data and business logs, and then uses the trained machine learning model to predict future task load, providing a basis for system load prediction and task scheduling decisions.

[0044] Task Priority Generation Module: Based on the predicted task load and the computational requirements of each task, this module generates a priority sequence for tasks. It calculates the comprehensive computational requirements of each policy task, combines this with the policy's business priority, calculates the weight of each task, and sorts the tasks according to their weights to ensure that critical tasks are processed in a timely manner.

[0045] Task scheduling module: Based on the task priority sequence, the module submits policy calculation tasks to the thread pool for execution in order of priority. The task scheduling module is responsible for allocating tasks to appropriate computing resources and ensuring that high-priority tasks are executed first, thereby improving system response speed and processing efficiency.

[0046] Thread pool dynamic adjustment module: This module combines reinforcement learning algorithms to dynamically adjust thread pool parameters (such as the number of core threads, maximum number of threads, thread lifespan, etc.) based on real-time system load conditions (such as CPU utilization, memory utilization, etc.). Through continuous load monitoring and real-time adjustments, it optimizes thread pool resource configuration and improves the system's task processing capabilities and resource utilization efficiency in high-concurrency environments.

[0047] The system load monitoring module collects and monitors the system's load status in real time, including CPU utilization, memory utilization, and the number of active threads, and passes the monitoring data as input to the reinforcement learning module and task scheduling module. This module ensures that the system has comprehensive load information at all times and can react promptly to optimize system performance.

[0048] Reinforcement Learning Optimization Module: This module uses reinforcement learning algorithms to optimize thread pool parameters based on task execution performance (such as task response time, system throughput, and resource utilization). The reinforcement learning model continuously adjusts its strategy based on environmental feedback, optimizing system resource allocation, improving task execution efficiency, and avoiding resource waste.

[0049] Feedback Optimization Module: Based on real-time data during task execution (such as average task response time, system throughput, etc.), this module provides feedback on the system's execution performance and updates optimization strategies. By optimizing the reinforcement learning model's strategy using feedback data, it ensures the system always maintains its optimal configuration, thereby improving the overall performance and stability of the system.

[0050] User Interface Module: Provides a user interface that displays calculation results, task status, and system performance indicators. It allows users to view the processing progress and results of the current policy calculation task, as well as view and adjust system parameters.

[0051] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0052] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An intelligent adaptive load scheduling based insurance loan trial calculation optimization method, characterized in that, The method comprises the following steps: obtaining all policy basic information under the name of a customer through batch database query, and excluding policies that do not meet the loan trial calculation conditions according to preset rules to generate a set of policies to be trial calculated; based on the set of policies to be trial calculated and historical load data of the system, predicting future task load by using a machine learning model, and calculating the calculation demand of each policy trial calculation task; generating a task priority sequence for the policy trial calculation task according to the predicted future task load and the calculation demand; performing the policy trial calculation task according to the task priority sequence using a thread pool, and dynamically adjusting the thread pool parameters by combining a reinforcement learning algorithm; performing the processing results of the policy trial calculation task asynchronously, including preferentially returning the results of short-time-consuming tasks based on time consumption estimation, and returning the results of long-time-consuming tasks in an incremental summary manner; integrating, checking and formatting the results of all policy trial calculation tasks to generate a structured loan trial calculation report. 2.The insurance loan trial calculation optimization method based on intelligent adaptive load scheduling according to claim 1, characterized in that, The method of obtaining all policy basic information under the name of a customer through batch database query, and excluding policies that do not meet the loan trial calculation conditions according to preset rules to generate a set of policies to be trial calculated, specifically comprises: obtaining policy basic information including policy number, policy type, policy status, policy effective date, insurance amount and loan related information under the unique identification of the customer through batch database query statements; screening the obtained policy basic information according to the following pre-filtering conditions: excluding policies that have been invalidated, canceled or terminated; excluding policies that have existing unsettled loans; excluding policy types that do not support loan business; arranging the policy basic information meeting the pre-filtering conditions into a set of policies to be trial calculated. 3.The insurance loan trial calculation optimization method based on intelligent adaptive load scheduling according to claim 1, characterized in that, The machine learning model is a long short-term memory network.

4. The insurance loan trial calculation optimization method based on intelligent adaptive load scheduling according to claim 1, characterized in that, The training process of the machine learning model comprises: Construct a training data set, based on system historical load data and business logs, construct sample sequences in time series : wherein, represents the eigenvector at time t , including the number of policies to be trial calculated , system CPU occupancy , memory occupancy , historical task average time consumption , the number of concurrent tasks ; represents the task load value at time , indicating the number of policy trial calculation tasks to be processed by the system at time ; is the length of the time window; is the total number of samples; standardizing the sample sequence standardization The standardized training sample sequence is input into a machine learning model Training is performed and future task load is predicted by the following equation: in, Indicates the predicted future task load; These are the model parameters, including weights and biases; The time after standardization t eigenvectors; Based on the predicted future task load and loss function, the model parameters are... Update the model to obtain the trained machine learning model.

5. The insurance loan trial calculation optimization method based on intelligent adaptive load scheduling according to claim 4, characterized in that, The loss function is: wherein, is a loss function, is a preset weight coefficient; is a time decay coefficient.

6. The insurance loan trial calculation optimization method based on intelligent adaptive load scheduling according to claim 1, characterized in that, The calculation of the calculation demand of each task specifically comprises: Treating the trial calculation of policy sets Each policy To obtain the core characteristics of the policy, including the policy amount, policy type, policy effective date, historical calculation time, and customer level; Calculate the total number of tasks for the policy trial; using a machine learning model to predict the average calculation time for a trial task for each policy using policy characteristics and system historical load data as inputs ; Based on the predicted single-task time consumption and task characteristics, in combination with the system's current available computing resources, the integrated computing demand of each policy task is calculated : in, For the policy The computational requirements indicators , Representing time respectively t The system's CPU and memory usage.

7. The insurance loan trial calculation optimization method based on intelligent adaptive load scheduling according to claim 1, characterized in that, The generation of the task priority sequence for the policy trial calculation task according to the predicted future task load and the calculation demand specifically comprises: For each to-be-tested policy task , based on its comprehensive computing demand and business priority index , calculate the task weight : wherein, with is a preset weight coefficient; is a total number of policy trial tasks; is a policy calculation requirement index; is a policy business priority; According to the calculated task weights All the policy trial tasks are sorted in descending order to generate a task priority sequence, wherein the higher the weight of a policy task, the higher the priority of the policy task in the sequence.

8. The insurance loan trial calculation optimization method based on intelligent adaptive load scheduling according to claim 1, characterized in that, The execution of the policy trial calculation task according to the task priority sequence using a thread pool, and the dynamic adjustment of the thread pool parameters by combining a reinforcement learning algorithm specifically comprises: defining thread pool parameters, including: Setting thread pool core thread number , maximum thread number and thread survival time , and initialize reinforcement learning agent; constructing a reinforcement learning state space and an action space, including: Definition of state vector wherein, is the current time t the number of insurance policies to be executed, is the system CPU occupancy rate, is the system memory occupancy rate, is the current number of active threads; Define the action space , which represents the adjustment range of thread pool parameters; 、 、 respectively represent the dynamic adjustment amount of the core thread number, the maximum thread number and the thread survival time. Based on the task priority sequence, each insurance policy trial task is submitted to the thread pool for execution in order of priority, and the system load state parameters are collected in real time, including CPU occupancy rate , memory occupancy rate , and thread activity number ; with the current state of the system as an input to the reinforcement learning model; outputting an action vector by the reinforcement learning model and updating the thread pool parameters in accordance with the action vector and updating the thread pool parameters in accordance with the action vector Based on the thread pool after updating the thread pool parameters, continue to perform the policy trial calculation task based on the task priority sequence, and record the average response time of the task , system throughput , and CPU utilization ; Record-based task average response time System throughput And CPU utilization The reinforcement learning reward function is calculated, and the reinforcement learning model strategy is optimized based on the calculated reinforcement learning reward function, and according to the updated reinforcement learning model strategy, the thread pool parameters are continuously dynamically adjusted, and the thread pool resource configuration is optimized.

9. The insurance loan trial calculation optimization method based on intelligent adaptive load scheduling according to claim 8, characterized in that, The action vector Update the thread pool parameters, formula: in, , , Representing time respectively The core number of threads, the maximum number of threads, and the thread lifespan.

10. The insurance loan trial calculation optimization method based on intelligent adaptive load scheduling according to claim 8, characterized in that, The reinforcement learning reward function is: wherein, is a reinforcement learning reward function at time t , , , is a preset weight coefficient.

Images