Resource allocation method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention provides a resource allocation method and device, electronic equipment and a storage medium, relates to the technical field of data processing, and is suitable for the fields of financial science and technology and medical health. The method comprises the following steps: acquiring target medical data in at least two target time periods; risk factor prediction is carried out according to the target medical data in each target time period, and medical risk factor prediction features are obtained; according to the medical risk factor prediction feature of each target time period, calculating to obtain an attention weight; performing weighted summation on the medical risk factor prediction features according to the attention weight to obtain target fusion medical factor features; predicting a preset insurance risk factor according to the target fusion medical factor feature to obtain insurance influence degree prediction data; and performing prediction according to the insurance influence degree prediction data to obtain resource gap prediction data, and performing resource allocation according to the resource gap prediction data. According to the embodiment of the invention, the reliability of resource allocation can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and is suitable for the fields of financial technology and medical health, and in particular relates to a resource allocation method and device, an electronic device and a storage medium. BACKGROUND

[0002] In the process of resource allocation, risk prediction is often needed. Risk prediction is a method of predicting potential risks (such as the claim rate of an insurance company) by analyzing data, and risk prediction is very important in asset liability matching. Asset liability matching is a resource allocation method, specifically by reasonably configuring the term of a company's resources (such as assets) and liabilities to ensure solvency, that is, to ensure that assets are sufficient to pay liabilities. For example, in the medical insurance asset management scenario in the fields of financial technology and medical health, the risk (such as claim expenditure) of medical insurance or life insurance can be predicted to allocate resources (such as equipment, insurance funds, medical funds, etc.) of the insurance company based on the risk prediction result, such as asset liability matching.

[0003] Current risk prediction methods mainly predict risks according to insurance-related indicators (such as mortality rate). However, these indicators are currently generally considered as static values that do not change over time, making it difficult to cope with changes in external factors (such as sudden changes in medical costs and changes in population health levels), resulting in poor accuracy of risk prediction and unreliable resource allocation.

[0004] Therefore, how to improve the reliability of resource allocation has become a technical problem to be solved. SUMMARY

[0005] The main purpose of the embodiments of the present application is to provide a resource allocation method and device, an electronic device and a storage medium, which aims to improve the accuracy of risk prediction and thus improve the reliability of resource allocation.

[0006] To achieve the above purpose, a first aspect of the embodiments of the present application provides a resource allocation method, which comprises:

[0007] obtaining target medical data in at least two target periods; wherein the length of any two target periods is different;

[0008] performing risk factor prediction according to the target medical data in each target period to obtain a medical risk factor prediction feature corresponding to each target period;

[0009] performing attention calculation according to the medical risk factor prediction feature of each target period to obtain an attention weight corresponding to each target period;

[0010] weighting and summing the at least two medical risk factor prediction features according to at least two of the attention weights, to obtain a target fusion medical factor feature;

[0011] performing influence degree prediction on a preset insurance risk factor according to the target fusion medical factor feature, to obtain insurance influence degree prediction data;

[0012] performing resource gap prediction according to the insurance influence degree prediction data, to obtain resource gap prediction data, so as to perform resource allocation according to the resource gap prediction data; wherein the resource gap prediction data is used to represent a probability distribution of a resource gap.

[0013] In some embodiments, the risk factor prediction according to the target medical data in each of the target time periods, to obtain a medical risk factor prediction feature corresponding to each of the target time periods, comprises:

[0014] performing time sequence feature extraction on the target medical data in each of the target time periods, to obtain a medical time sequence feature;

[0015] performing risk factor judgment on the medical time sequence feature by a target medical rule decision tree model, to obtain the medical risk factor prediction feature corresponding to each of the target time periods.

[0016] In some embodiments, before the risk factor judgment on the medical time sequence feature by the target medical rule decision tree model, to obtain the medical risk factor prediction feature corresponding to each of the target time periods, the method further comprises:

[0017] obtaining historical medical data, and performing incidence rate judgment on the time sequence feature of the historical medical data by a pre-constructed initial medical rule decision tree model, to obtain an initial prediction sub-incidence rate;

[0018] performing reward value calculation on the initial prediction sub-incidence rate and a preset label incidence rate by a Bayesian optimization algorithm, to obtain a target reward value;

[0019] performing parameter adjustment on the initial medical rule decision tree model according to the target reward value, to obtain the target medical rule decision tree model.

[0020] In some embodiments, the target medical data is at least two, and the geographical areas of any two of the target medical data are different, and the target fusion medical factor feature is predicted based on the target medical data of the same geographical area;

[0021] After the influence degree prediction on the preset insurance risk factor according to the target fusion medical factor feature, to obtain the insurance influence degree prediction data, the method further comprises:

[0022] If the target fusion medical factor feature of the geographic area is greater than or equal to the preset medical risk factor threshold value, the geographic area is determined as a high-risk area;

[0023] According to a preset first increasing ratio, the insurance influence degree prediction data corresponding to the high-risk area is increased.

[0024] In some embodiments, after the increasing processing of the insurance influence degree prediction data corresponding to the high-risk area according to the preset first increasing ratio, the method further comprises:

[0025] According to the distance calculation between the high-risk area and each other geographic area, an inter-regional interval distance is obtained.

[0026] According to the inter-regional interval distance and a preset distance threshold value, a neighboring area of the high-risk area is determined from other geographic areas.

[0027] According to a preset second increasing ratio, the insurance influence degree prediction data corresponding to the neighboring area is increased; wherein the second increasing ratio is less than the first increasing ratio.

[0028] In some embodiments, the insurance influence degree prediction data includes claim rate influence degree data corresponding to at least two target periods.

[0029] The resource gap prediction according to the insurance influence degree prediction data to obtain resource gap prediction data, so as to perform resource allocation according to the resource gap prediction data, comprises:

[0030] According to a preset first time length threshold value range, a first period is determined from at least two target periods, and according to a preset second time length threshold value range, a second period is determined from at least two target periods; wherein the time length of the first period is less than the time length of the second period.

[0031] According to the claim rate influence degree data of the first period, resource gap prediction is performed to obtain short-term resource gap prediction data.

[0032] According to the claim rate influence degree data of the second period, resource gap prediction is performed to obtain long-term resource gap prediction data, so as to perform resource allocation according to the short-term resource gap prediction data and the long-term resource gap prediction data.

[0033] In some embodiments, the attention calculation according to the medical risk factor prediction feature of each target period to obtain the attention weight corresponding to each target period comprises:

[0034] multiply the preset query weight matrix with the medical risk factor prediction feature to obtain a medical query vector;

[0035] multiply the preset key weight matrix with the medical risk factor prediction feature to obtain a medical key vector;

[0036] calculate a ratio of a product of the medical query vector and a transpose vector of the medical key vector to a square root of a dimension of the medical key vector to obtain an attention score;

[0037] normalize the attention score to obtain the attention weight corresponding to each target time period.

[0038] To achieve the above object, a second aspect of the embodiment of the present application provides a resource allocation device, which comprises:

[0039] a data acquisition module configured to acquire target medical data in at least two target time periods; wherein the length of any two target time periods is different;

[0040] a medical risk prediction module configured to predict a risk factor according to the target medical data in each target time period to obtain a medical risk factor prediction feature corresponding to each target time period;

[0041] an attention calculation module configured to calculate an attention according to the medical risk factor prediction feature of each target time period to obtain an attention weight corresponding to each target time period;

[0042] a risk factor fusion module configured to perform weighted summation on at least two medical risk factor prediction features according to at least two attention weights to obtain a target fusion medical factor feature;

[0043] an influence degree prediction module configured to predict an influence degree of a preset insurance risk factor according to the target fusion medical factor feature to obtain insurance influence degree prediction data;

[0044] a resource gap prediction module configured to predict a resource gap according to the insurance influence degree prediction data to obtain resource gap prediction data, and to perform resource allocation according to the resource gap prediction data; wherein the resource gap prediction data is used to represent a probability distribution of a resource gap.

[0045] To achieve the above object, a third aspect of the embodiment of the present application provides an electronic device, which comprises a memory and a processor, the memory stores a computer program, and the processor implements the method of the first aspect when executing the computer program.

[0046] To achieve the above object, a computer readable storage medium is provided in a fourth aspect of the embodiments of the present application, the computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the method in the first aspect.

[0047] The resource allocation method, device, electronic device and storage medium provided in the present application can capture the risk change trend in different time scales. For example, the medical risk factor prediction feature in a short time period can reflect the risk impact of an acute disease, and the medical risk factor prediction feature in a long time period can reflect the risk impact of a chronic disease. Moreover, the attention weight is calculated according to the medical risk factor prediction feature in each target time period. In this way, the importance difference of medical risk factors in different time periods to the overall risk prediction can be adaptively adjusted, and the risk prediction error caused by the change of target medical data can be reduced, instead of relying on static indicators to predict risks. For example, if the short-term medical expenses rise, the attention weight of the short time period is increased. Then, the multiple medical risk factor prediction features are weighted and summed according to the multiple attention weights to obtain a target fusion medical factor feature, and the influence of the insurance risk factor is predicted according to the target fusion medical factor feature to obtain insurance influence degree prediction data (such as the change of claim rate, the change of interest rate, etc.), thereby improving the accuracy of risk prediction. Then, the resource gap prediction data is predicted according to the insurance influence degree prediction data, so that the causal relationship between the dynamic change of medical risk and the insurance resource gap can be established, and the resource allocation can be performed according to the resource gap prediction data, thereby improving the reliability of resource allocation. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 is a flowchart of the resource allocation method provided in the embodiments of the present application;

[0049] Figure 2 is a flowchart of step 102 in Figure 1 ;

[0050] Figure 3 is a flowchart of the resource allocation method provided in another embodiment of the present application;

[0051] Figure 4 is a flowchart of the resource allocation method provided in another embodiment of the present application;

[0052] Figure 5 is a flowchart of the resource allocation method provided in another embodiment of the present application;

[0053] Figure 6 is a flowchart of step 103 in Figure 1 ;

[0054] Figure 7 is Figure 1 a flowchart of step 106 in

[0055] Figure 8 is a structural schematic diagram of a resource allocation apparatus provided by an embodiment of the present application;

[0056] Figure 9 is a hardware structural schematic diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0057] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application is further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

[0058] It should be noted that although the functional modules are divided in the apparatus schematic diagram, and the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a manner different from the module division in the apparatus or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and above-mentioned drawings are used to distinguish similar objects, and do not necessarily describe a specific order or sequence.

[0059] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs. The terms used herein are only for the purpose of describing the embodiments of the present application, and are not intended to limit the present application.

[0060] First, several terms involved in the present application are analyzed:

[0061] Artificial Intelligence (AI): is a new technical science that studies, develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence; Artificial intelligence is a branch of computer science, artificial intelligence attempts to understand the essence of intelligence, and produce a new intelligent machine that can react in a similar way to human intelligence, including robots, language recognition, image recognition, natural language processing and expert systems. Artificial intelligence can be a simulation of human consciousness and thinking information process. Artificial intelligence can also be a theory, method, technology and application system that uses digital computers or digital computer controlled machines to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. Artificial intelligence basic technologies generally include technologies such as sensors, special artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction system, mechatronics, etc. Artificial intelligence software technology mainly includes computer vision technology, robot technology, biometric technology, speech processing technology, natural language processing technology, and machine learning / deep learning, etc. The present application can acquire and process related data based on artificial intelligence technology.

[0062] Long Short-Term Memory (LSTM): also known as Long Short-Term Memory, is a kind of time recurrent neural network. Neural network is one of the core technologies of artificial intelligence (AI). Long Short-Term Memory can effectively process long time series data.

[0063] In the resource allocation process, risk prediction is often needed. For example, in the life insurance industry in the field of financial technology, the risk can be predicted by an asset liability matching model according to the traditional actuarial method, historical financial information and static assumption values of some long-term indicators (such as mortality rate, interest rate, expense rate). However, since the long-term indicators are regarded as static values that remain unchanged for a long time, the current risk prediction method is difficult to cope with the influence of external environmental factors in the real world, such as changes in medical costs, changes in population health levels, etc. Moreover, it is currently difficult to mine the relationship between medical data and insurance risk factors, such as the nonlinear correlation between the incidence of chronic diseases and the claim rate. In addition, the current risk prediction usually relies on a data set divided at a fixed period, which is difficult to capture short-term problems such as sudden increase in medical expenses during an outbreak of infectious diseases, and long-term trends of rising medical costs due to aging, resulting in limited accuracy of risk prediction and medical expenditure prediction, for example, it is difficult to capture the sudden increase in claim rate caused by a certain explosive infectious disease.

[0064] Based on this, the embodiment of the application provides a resource allocation method and device, electronic equipment and storage medium, aiming to improve the accuracy of risk prediction, and then improve the reliability and flexibility of resource allocation.

[0065] The resource allocation method, device, electronic equipment and storage medium provided by the embodiment of the application are specifically described through the following embodiments. First, the resource allocation method in the embodiment of the application is described.

[0066] The resource allocation method provided by the embodiment of the application can be applied to a terminal, can also be applied to a server end, and can also be software running in the terminal or the server end. In some embodiments, the terminal can be a smart phone, a tablet computer, a notebook computer, a desktop computer, etc.; the server end can be configured as an independent physical server, can also be configured as a server cluster or a distributed system composed of multiple physical servers, can also be configured as a cloud server providing basic cloud computing services such as cloud service, cloud database, cloud computing, cloud function, cloud storage, network service, cloud communication, middleware service, domain name service, security service, CDN (Content Delivery Network), and big data and artificial intelligence platform; and the software can be an application for implementing the resource allocation method, but is not limited to the above forms.

[0067] The application can be used in many general or specific computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The application can also be practiced in a distributed computing environment, in which tasks are performed by remote processing devices connected by a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media, including storage devices.

[0068] It should be noted that in various specific embodiments of the present application, when relevant processing needs to be performed on data related to the identity or characteristics of the user, such as user information, user behavior data, user history data, and user location information, the user's permission or consent will be obtained first, and the collection, use, and processing of such data will comply with relevant laws, regulations, and standards. In addition, when the embodiments of the present application need to obtain sensitive personal information of the user, the separate permission or separate consent of the user will be obtained through a pop-up window or a jump to a confirmation page, and after obtaining the separate permission or separate consent of the user, the necessary user-related data for enabling the embodiments of the present application to function normally will be obtained.

[0069] Figure 1 is an optional flowchart of the resource allocation method provided by the embodiments of the present application, Figure 1 The method in the method can include but is not limited to steps 101-106.

[0070] Step 101, obtaining target medical data in at least two target periods; wherein the length of any two target periods is different;

[0071] Step 102, predicting a risk factor according to the target medical data in each target period to obtain a medical risk factor prediction feature corresponding to each target period;

[0072] Step 103, calculating attention according to the medical risk factor prediction feature of each target period to obtain an attention weight corresponding to each target period;

[0073] Step 104, performing weighted summation on at least two medical risk factor prediction features according to at least two attention weights to obtain a target fusion medical factor feature;

[0074] Step 105, predicting the influence degree of a preset insurance risk factor according to the target fusion medical factor feature to obtain insurance influence degree prediction data;

[0075] Step 106, predicting a resource gap according to the insurance influence degree prediction data to obtain resource gap prediction data, and performing resource allocation according to the resource gap prediction data; wherein the resource gap prediction data is used to represent the probability distribution of the resource gap.

[0076] The beneficial effects of the embodiments of the present application include but are not limited to: by obtaining multiple target time periods of different lengths, predicting medical risk factor prediction features according to target medical data (such as short-term medical data or long-term medical data) in each target time period. In this way, the risk change trend at different time scales can be captured, for example, the medical risk factor prediction features of the short time period can reflect the risk impact of the sudden disease, and the medical risk factor prediction features of the long time period can reflect the risk impact of the chronic disease. And calculate the attention weight according to the medical risk factor prediction feature of each target time period. In this way, the importance difference of medical risk factors at different time periods to the overall risk prediction can be adaptively adjusted, and the risk prediction error caused by the change of target medical data is reduced, instead of relying on static indicators to predict risks. For example, if the short-term medical expenses rise, the attention weight of the short time period is increased. Then, the multiple medical risk factor prediction features are weighted and summed according to the multiple attention weights to obtain a target fusion medical factor feature, and the influence of the insurance risk factor is predicted according to the target fusion medical factor feature to obtain insurance influence degree prediction data (such as change of claim rate, change of interest rate, etc.), thereby improving the accuracy of risk prediction. Then, the resource gap prediction data is predicted according to the insurance influence degree prediction data, so that the causal relationship between the dynamic change of medical risk and the insurance resource gap can be established, and the resource allocation can be performed according to the resource gap prediction data, thereby improving the reliability of resource allocation.

[0077] In step 101 of some embodiments, the target medical data is data used to predict medical risk factors. For example, the target medical data can include any one or more of the following data: data of electronic health records (EHR), medical text (medical records, diagnosis reports, etc.), data of public health monitoring (such as the incidence of infectious diseases), health data collected by wearable devices, genetic testing data, etc. For example, in the insurance scenario of the financial technology field, the target medical data can also include data of medical insurance (or life insurance) premium reimbursement. The target medical data can also be obtained in other ways, without being limited thereto.

[0078] In some embodiments, it should be noted that in at least two target time periods, the length of any two target time periods is different. The target time period can belong to any one of a short time period, a medium time period and a long time period. For example, the at least two target time periods can include time period T1, time period T2 and time period T3. Assuming that time period T1 belongs to a short time period, time period T2 belongs to a medium time period, and time period T3 belongs to a long time period, the length of time period T1 can be any length in 1 to 3 years, the length of time period T2 can be any length in 3 to 10 years, and the length of time period T3 can be any length of more than 10 years.

[0079] In step 102 of some embodiments, the medical risk factor prediction feature is a variation feature of the predicted medical risk factor. For example, the medical risk factor can include any one or more of the following: morbidity, cure rate, mortality rate, etc. In some embodiments, assuming the medical risk factor is morbidity, the medical risk factor prediction feature can be the degree of variation of the morbidity.

[0080] In step 103 of some embodiments, the attention weight is a weight calculated according to the medical risk factor prediction feature of each target period. Specifically, the attention weight can be calculated by a self-attention algorithm on the medical risk factor prediction feature. The attention weight can also be calculated by other types of attention algorithms, without being limited thereto. As for the calculation process of the attention weight, reference can be made to the specific description of steps 601 to 604 below, which will not be repeated here.

[0081] In step 104 of some embodiments, the target fusion medical factor feature is a feature vector obtained by attention-weighted fusion of the plurality of medical risk factor prediction features.

[0082] In step 105 of some embodiments, the insurance impact prediction data refers to a predicted value of the impact of the target fusion medical factor feature on the insurance risk factor. Specifically, the insurance risk factor can include the claim rate, interest rate, rate of the insurance, etc. It should be noted that the claim rate refers to the percentage of claim expenditure to premium income within a certain period. For example, assuming the insurance risk factor is the claim rate, the insurance impact prediction data can be the degree of variation of the claim rate.

[0083] In step 106 of some embodiments, the resource gap prediction data is a resource gap, such as the estimated resource expenditure, predicted based on the degree of variation of the insurance risk factor (i.e., the insurance impact prediction data). For example, in the insurance scenario of the financial technology field, assuming the insurance impact prediction data is the degree of variation of the claim rate of the medical insurance (or life insurance), the resource gap prediction data can be the probability distribution of the numerical value of the claim cash flow. In some embodiments, specifically, the probability distribution of the claim cash flow (i.e., the resource gap prediction data) can be predicted based on the insurance impact prediction data by a Monte Carlo simulation method.

[0084] In some embodiments, specifically, the resource allocation optimization strategy can be generated according to the resource gap prediction data by a reinforcement learning algorithm, such as a proximal policy optimization (PPO) algorithm. In another embodiment, the resource allocation optimization strategy can also be generated by other policy gradient algorithms, without being limited thereto.

[0085] In some embodiments, for example, in an insurance scenario in the field of financial technology, the resource allocation can refer to adjusting the medical asset allocation strategy of an insurance company, such as allocating resources of the insurance company, such as assets, equipment, etc. For example, assuming that the resources of the insurance company include, and the resource gap prediction data indicates that the incidence of target diseases (such as cancer, cardiovascular disease, infectious disease, etc.) is rising, the allocation proportion of medical assets in the resources of the insurance company can be increased, such as increasing the allocation proportion of real estate investment trusts (REITs) of medical infrastructure, biopharmaceutical industry stocks, and medical technology stocks. For another example, assuming that the resource gap prediction data indicates that the probability of short-term claim payment is rising, the allocation proportion of high-liquidity assets (such as short-term treasury bonds) in the resources of the insurance company can be increased to hedge risks.

[0086] In some embodiments, based on the risk prediction results (such as target fusion medical factor features, insurance influence degree prediction data, resource gap prediction data, etc.) and the strategy of resource allocation, a visual resource allocation chart can be generated, such as a heat map showing the predicted changes in medical resource expenditure in different geographical areas, so as to be displayed to users, improve the visualization of resource allocation, and also make corresponding resource allocation suggestions according to the resource allocation chart.

[0087] In some embodiments, incremental learning technology can be used to obtain new medical data (such as new real-time medical records, new drug prices, etc.), so as to update the risk prediction results in real time.

[0088] Please refer to Figure 2 In some embodiments, step 102 can include, but is not limited to, steps 201 to 202:

[0089] Step 201: performing time series feature extraction according to target medical data in each target period to obtain medical time series features;

[0090] Step 202: performing risk factor judgment on the medical time series features by using a target medical rule decision tree model to obtain medical risk factor prediction features corresponding to each target period.

[0091] The embodiment has the advantages that the medical time sequence features are extracted according to the target medical data in each target period, so that the dynamic change law of the medical risk factor (such as seasonal fluctuation of morbidity) can be captured. The risk factor is judged by the target medical rule decision tree model on the medical time sequence features, so that the medical knowledge and the data-driven features (such as time sequence fluctuation amplitude) can be fused to generate the medical risk factor prediction features corresponding to each target period. In this way, the short-term sudden risk and the long-term trend change can be coped with at the same time, the limitation of response lag to external factors is reduced, the accuracy of risk prediction is improved, and the reliability of resource allocation is improved.

[0092] In step 201 of some embodiments, the medical time sequence features are time sequence features of the target medical data in the target period. For example, the medical time sequence features can represent the change trend of the number of patients in the target period. Specifically, the medical time sequence features can be extracted from the target medical data in each target period by a long short-term memory (LSTM) model. The medical time sequence features can also be extracted by other models or algorithms, without limitation.

[0093] In step 202 of some embodiments, the target medical rule decision tree model is a decision tree model for predicting medical risk factor features. In some embodiments, the medical risk factor prediction features can be predicted by a decision tree model such as an XGBoost model. The medical risk factor features can also be predicted by other models or algorithms, without limitation.

[0094] Please refer to Figure 3 In some embodiments, before step 202, the resource allocation method can further include but is not limited to steps 301 to 303:

[0095] Step 301: Obtain historical medical data, and judge the time sequence features of the historical medical data by a pre-constructed initial medical rule decision tree model to obtain an initial predicted sub-morbidity;

[0096] Step 302: Calculate the reward value of the initial predicted sub-morbidity and the preset label morbidity by a Bayesian optimization algorithm to obtain a target reward value;

[0097] Step 303: Adjust the parameters of the initial medical rule decision tree model according to the target reward value to obtain the target medical rule decision tree model.

[0098] The embodiment has the advantages that, by acquiring historical medical data and judging the incidence rate through the initial medical rule decision tree model, an initial prediction result can be generated based on historical time sequence characteristics, thereby providing a reference for subsequent model training. The reward value is calculated for the initial prediction sub-incidence rate and the label incidence rate through the Bayesian optimization algorithm, so that the model prediction bias can be dynamically evaluated and the optimization direction can be quantified. The model parameters of the initial medical rule decision tree model are adjusted according to the target reward value, so that the threshold, depth and other key parameters of the decision tree can be accurately optimized, thereby the target medical rule decision tree model obtained after training can more accurately predict the risk and improve the accuracy of risk prediction, and then the reliability of resource allocation is improved.

[0099] In step 301 of some embodiments, the acquisition time of the historical medical data is earlier than the acquisition time of the target medical data. In some embodiments, specifically, the data type of the historical medical data is consistent with the data type of the target medical data, which will not be repeated here. It should be noted that the initial prediction sub-incidence rate is the incidence rate predicted based on the historical medical data.

[0100] In step 302 of some embodiments, the label incidence rate is the true incidence rate corresponding to the initial prediction sub-incidence rate. In some embodiments, similarity calculation can be performed on the initial prediction sub-incidence rate and the preset label incidence rate, so that the similarity of the initial prediction sub-incidence rate and the label incidence rate is taken as the target reward value through the Bayesian optimization algorithm, so as to perform model training (i.e., model parameter adjustment) on the initial medical rule decision tree model.

[0101] In step 303 of some embodiments, the target medical rule decision tree model refers to the decision tree model obtained after training. In some embodiments, the target medical rule decision tree model can be an XGBoost model, or can be other types of decision tree models, without being limited thereto.

[0102] Please refer to Figure 4 In some embodiments, the target medical data is at least two, any two target medical data are different in geographical area, and the target fusion medical factor feature is predicted based on the target medical data in the same geographical area;

[0103] After step 105, the resource allocation method can further include but is not limited to steps 401 to 402:

[0104] Step 401, if the target fusion medical factor feature of the geographical area is greater than or equal to the preset medical risk factor threshold, the geographical area is determined as a high-risk area;

[0105] Step 402, according to the preset first increasing ratio, the insurance influence degree prediction data corresponding to the high-risk area is increased.

[0106] The embodiment has the advantages that by generating the target fusion medical factor feature respectively for the target medical data of different geographical areas, regional-specific risk differences can be captured, and high-risk areas are identified based on the comparison result of the target fusion medical factor feature and the preset threshold, so that the geographical areas with greater medical risks can be accurately located. Then, the insurance influence degree prediction data of the high-risk areas are processed by increasing by a preset ratio, so that the insurance risk signals of key areas (such as a sharp increase in medical insurance claims caused by regional infectious disease outbreaks) can be dynamically enhanced, the foresight of resource prediction can be improved, and the comprehensiveness of risk prediction can be improved.

[0107] In step 401 of some embodiments, the high-risk area refers to a geographical area with a target fusion medical factor feature greater than or equal to a medical risk factor threshold.

[0108] In some embodiments, specifically, assuming that the target fusion medical factor feature is the incidence rate change, the medical risk factor threshold can be a threshold of the incidence rate change.

[0109] In step 402 of some embodiments, the first increasing ratio can be multiplied by the insurance influence degree prediction data corresponding to the high-risk area, so as to update the insurance influence degree prediction data corresponding to the high-risk area. In some embodiments, for example, in the insurance scenario in the field of financial technology, for users belonging to the high-risk area, insurance resources can be adjusted according to needs, such as modifying the terms of medical insurance, adjusting the reinsurance ratio, and the like.

[0110] Please refer to Figure 5 In some embodiments, after step 402, the resource allocation method can further include but is not limited to steps 501 to 503:

[0111] Step 501: calculating the distance between the high-risk area and each other geographical area to obtain an inter-regional distance;

[0112] Step 502: determining the adjacent area of the high-risk area from other geographical areas according to the inter-regional distance and a preset distance threshold;

[0113] Step 503: increasing the insurance influence degree prediction data corresponding to the adjacent area by a preset second increasing ratio; wherein the second increasing ratio is less than the first increasing ratio.

[0114] The embodiment has the advantages that, considering the cross-regional risk transmission of medical risk factors (such as the incidence of infectious diseases), the adjacent regions are determined according to the distance calculation of the high-risk region and other geographical regions, so that the spatial transmission path of the medical risk (such as the cross-regional spread of infectious diseases and the transfer of compensation pressure caused by cross-regional medical treatment) can be identified, instead of processing the risk of a single region in isolation. The prediction data of the influence degree of the insurance of the adjacent regions is increased by a proportion (that is, a second increasing proportion) less than that of the high-risk region, so that the characteristics of risk transmission can be reflected, and the influence of the edge region is avoided from being excessively enlarged, thereby improving the comprehensiveness of risk prediction and the reliability of resource allocation.

[0115] In step 501 of some embodiments, it should be noted that the other geographical regions refer to geographical regions other than the high-risk region. The region interval distance refers to the distance between each other geographical region and the high-risk region.

[0116] In step 502 of some embodiments, the adjacent regions of the high-risk region refer to other geographical regions with a region interval distance less than or equal to a preset distance threshold.

[0117] In step 503 of some embodiments, the second increasing proportion can be multiplied by the prediction data of the influence degree of the insurance of the adjacent regions, so as to update the prediction data of the influence degree of the insurance of the adjacent regions. It should be noted that the second increasing proportion is less than the first increasing proportion, for example, if the first increasing proportion is 0.5, the second increasing proportion can be 0.3.

[0118] Please refer to Figure 6 In some embodiments, step 103 can include but is not limited to steps 601 to 604:

[0119] Step 601, multiplying the preset query weight matrix and the medical risk factor prediction feature to obtain a medical query vector;

[0120] Step 602, multiplying the preset key weight matrix and the medical risk factor prediction feature to obtain a medical key vector;

[0121] Step 603, calculating the attention score by ratio of the product of the medical query vector and the transpose vector of the medical key vector to the square root of the dimension of the medical key vector;

[0122] Step 604, normalizing the attention score to obtain the attention weight corresponding to each target period.

[0123] The embodiment has the advantages that the medical query vector and the medical key vector are generated by multiplying the query weight matrix and the key weight matrix with the medical risk factor prediction features respectively, so that the original features can be mapped to a vector space in which the similarity can be calculated. Then, the attention score is obtained according to the product of the medical query vector and the transpose of the medical key vector, divided by the square root of the dimension of the key vector, and the attention score is converted into the attention weight. In this way, the limitation that the traditional mean weighted method is not sensitive to the timing fluctuations can be broken through, so as to adaptively adjust the importance difference of the medical risk factors in different time periods on the overall risk prediction, reduce the risk prediction error caused by the change of the target medical data, and predict the insurance risk factors subsequently, so as to improve the accuracy of risk prediction, and further improve the reliability of resource allocation.

[0124] In step 601 of some embodiments, the medical query vector is a query vector in the attention mechanism, specifically, the medical query vector is the product of the query weight matrix and the medical risk factor prediction feature.

[0125] In step 602 of some embodiments, the medical key vector is a key vector in the attention mechanism, specifically, the medical key vector is the product of the key weight matrix and the medical risk factor prediction feature.

[0126] In step 603 of some embodiments, the product of the medical query vector and the transpose vector of the medical key vector can be taken as the numerator, the square root of the dimension of the medical key vector can be taken as the denominator, and then the attention score can be obtained by division.

[0127] In step 604 of some embodiments, the attention score can be normalized by an activation function. For example, the attention score can be normalized by a Softmax function to obtain the attention weight.

[0128] Please refer to Figure 7 In some embodiments, the insurance influence degree prediction data includes claim rate influence degree data corresponding to at least two target time periods.

[0129] Step 106 can include but is not limited to steps 701 to 703:

[0130] Step 701, determining a first time period from the at least two target time periods according to a preset first time length threshold range, and determining a second time period from the at least two target time periods according to a preset second time length threshold range; wherein the time length of the first time period is less than the time length of the second time period;

[0131] Step 702, predicting the resource gap according to the claim rate influence degree data of the first time period to obtain short-term resource gap prediction data;

[0132] At step 703, resource gap prediction is performed according to the loss ratio influence degree data of the second time period to obtain long-term resource gap prediction data, so as to perform resource allocation according to the short-term resource gap prediction data and the long-term resource gap prediction data.

[0133] The embodiment has the advantages that the first time period (i.e., the short-term time period) and the second time period (i.e., the long-term time period) are determined from the multiple target time periods according to the preset time length threshold range (such as the first time length threshold range and the second time length threshold range), so that the influence of risk fluctuations of different time lengths on the loss ratio can be distinguished. Then, the short-term resource gap prediction data is generated according to the short-term loss ratio influence degree data, and the long-term resource gap prediction data is generated according to the long-term data, so that the resource allocation can be performed more comprehensively according to the predicted resource gaps (the short-term resource gap prediction data and the long-term resource gap prediction data) of different time scales, the risk of resource mismatch is reduced, and the resource allocation is more comprehensive.

[0134] In step 701 of some embodiments, the first time period refers to a target time period whose time length belongs to the first time length threshold range, i.e., the short-term time period. The second time period refers to a target time period whose time length belongs to the second time length threshold range, i.e., the long-term time period. It should be noted that the first time length threshold range is smaller than the second time length threshold range. For example, assuming that the first time length threshold range is 1 to 3 years, the second time length threshold range can be more than 5 years.

[0135] In step 702 of some embodiments, the short-term resource gap prediction data is data predicted according to the loss ratio influence degree data (such as loss ratio change data in the next 3 years) of the first time period.

[0136] In step 703 of some embodiments, the long-term resource gap prediction data is data predicted according to the loss ratio influence degree data (such as loss ratio change data in the next 10 years) of the second time period. In some embodiments, the resource gap prediction data can include the short-term resource gap prediction data and the long-term resource gap prediction data. For example, in the insurance scenario in the field of financial technology, the short-term resource gap prediction data (or the long-term resource gap prediction data) can include the probability distribution of the numerical value of the loss payment fund.

[0137] Please refer to Figure 8 The embodiment of the present application also provides a resource allocation device, which can implement the above-mentioned resource allocation method. The device comprises:

[0138] The data acquisition module 801 is configured to acquire target medical data in at least two target time periods; wherein the time lengths of any two target time periods are different.

[0139] The medical risk prediction module 802 is configured to perform risk factor prediction according to the target medical data in each target period, to obtain a medical risk factor prediction feature corresponding to each target period.

[0140] The attention calculation module 803 is configured to perform attention calculation according to the medical risk factor prediction feature of each target period, to obtain an attention weight corresponding to each target period.

[0141] The risk factor fusion module 804 is configured to perform weighted summation on at least two medical risk factor prediction features according to at least two attention weights, to obtain a target fusion medical factor feature.

[0142] The influence degree prediction module 805 is configured to perform influence degree prediction on a preset insurance risk factor according to the target fusion medical factor feature, to obtain insurance influence degree prediction data.

[0143] The resource gap prediction module 806 is configured to perform resource gap prediction according to the insurance influence degree prediction data, to obtain resource gap prediction data, and to perform resource allocation according to the resource gap prediction data; wherein the resource gap prediction data is used to represent a probability distribution of a resource gap.

[0144] In an embodiment, the resource allocation apparatus further includes a model updating module configured to: obtain historical medical data, and perform incidence rate judgment on time sequence features of the historical medical data by using a pre-constructed initial medical rule decision tree model, to obtain an initial prediction sub-incidence rate; perform reward value calculation on the initial prediction sub-incidence rate and a preset label incidence rate by using a Bayesian optimization algorithm, to obtain a target reward value; and perform parameter adjustment on the initial medical rule decision tree model according to the target reward value, to obtain a target medical rule decision tree model.

[0145] In an embodiment, the resource allocation apparatus further includes a first influence degree increasing module configured to: if the target fusion medical factor feature of the geographic area is greater than or equal to a preset medical risk factor threshold, determine the geographic area as a high-risk area; and perform increasing processing on the insurance influence degree prediction data corresponding to the high-risk area according to a preset first increasing ratio.

[0146] In an embodiment, the resource allocation apparatus further includes a second influence degree increasing module configured to: perform distance calculation on the high-risk area and each other geographic area, to obtain an inter-regional interval distance.

[0147] According to the inter-regional interval distance and a preset distance threshold, determine a neighboring area of the high-risk area from the other geographic areas; and perform increasing processing on the insurance influence degree prediction data corresponding to the neighboring area according to a preset second increasing ratio; wherein the second increasing ratio is less than the first increasing ratio.

[0148] The specific implementation of the resource allocation apparatus is basically the same as the specific embodiments of the above resource allocation method, and will not be repeated here.

[0149] The embodiments of the present application also provide an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor implements the above resource allocation method when executing the computer program. The electronic device can include any intelligent terminal such as a tablet computer or a vehicle-mounted computer.

[0150] Please refer to Figure 9 , Figure 9 The hardware structure of the electronic device of another embodiment is illustrated, which includes:

[0151] The processor 901 can be implemented in the form of a general-purpose CPU (Central Processing Unit), a microprocessor, an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits, and is used to execute related programs to implement the technical solutions provided by the embodiments of the present application.

[0152] The memory 902 can be implemented in the form of a ROM (Read-Only Memory), a static storage device, a dynamic storage device, or a RAM (Random Access Memory). The memory 902 can store an operating system and other application programs. When the technical solutions provided by the embodiments of the present application are implemented by software or firmware, the related program codes are stored in the memory 902 and are called and executed by the processor 901 to implement the resource allocation method of the embodiments of the present application.

[0153] The input / output interface 903 is used to realize information input and output.

[0154] The communication interface 904 is used to realize the communication interaction between the device and other devices. The communication can be realized by a wired manner (such as USB, network cable, etc.) or a wireless manner (such as mobile network, WIFI, Bluetooth, etc.).

[0155] The bus 905 transmits information between various components (such as the processor 901, the memory 902, the input / output interface 903, and the communication interface 904) of the device.

[0156] The processor 901, the memory 902, the input / output interface 903, and the communication interface 904 are connected to each other through the bus 905 to realize communication connection within the device.

[0157] The embodiment of the present application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to realize the resource allocation method.

[0158] The memory, as a non-transitory computer readable storage medium, can be used to store non-transitory software programs and non-transitory computer executable programs. In addition, the memory can include a high-speed random access memory, and can also include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state memory device. In some embodiments, the memory can optionally include a memory remotely arranged relative to the processor, and the remote memory can be connected to the processor through a network. Examples of the network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0159] It should be noted that the non-company software tools or components appearing in the embodiments of the present application are only examples for introduction, and do not represent actual use.

[0160] The embodiments described in the present application are used to more clearly illustrate the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. Those skilled in the art can know that, with the evolution of technology and the appearance of new application scenarios, the technical solutions provided by the embodiments of the present application are also applicable to similar technical problems.

[0161] Those skilled in the art can understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and can include more or fewer steps than the figures shown, or combine certain steps, or different steps.

[0162] The device embodiments described above are only schematic, and the units described as separate components can or can not be physically separated, that is, can be located in one place, or can be distributed on multiple network units. According to actual needs, part or all of the modules can be selected to achieve the purpose of the embodiments of the present application.

[0163] Those skilled in the art can understand that all or some steps in the above disclosed method, the functions of the modules / units in the system and the device can be implemented as software, firmware, hardware and their appropriate combinations.

[0164] The terms "first", "second", "third", "fourth", and the like in the description of this application and in the claims hereof, if any, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the use of the terms so termed herein is solely for the convenience of the reader and does not limit the scope of the application. It is also to be understood that the description and examples in this application are intended to cover all possible combinations where any of the several elements can represent one or more elements.

[0165] It should be understood that, in the application, "at least one" refers to one or more, and "multiple" refers to two or more. "And / or" is used to describe the relationship between associated objects, which means that there can be three relationships, for example, "A and / or B" can represent three cases: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally represents an "or" relationship between the associated objects. "At least one of the following" or similar expressions means any combination of these items, including any combination of single or multiple items. For example, at least one of a, b or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0166] In several embodiments provided in the application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only illustrative, for example, the division of the above-mentioned units is only a logical function division, and actual implementation can have another division manner, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. The coupling or direct coupling or communication connection between the displayed or discussed each other can be indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0167] The units described above as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed on multiple network units. According to actual needs, part or all of the units can be selected to achieve the purpose of the embodiment of the application.

[0168] In addition, each function unit in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software function unit.

[0169] When the integrated unit is realized in the form of a software function unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application, essentially or in part, or all or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes multiple instructions used to cause a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods in the embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various other media that can store programs.

[0170] The preferred embodiments of the embodiments of the present application are described above with reference to the accompanying drawings, and are not limited to the scope of the embodiments of the present application. Any modifications, equivalent replacements and improvements made by those skilled in the art without departing from the scope and essence of the embodiments of the present application shall be within the scope of the embodiments of the present application.

Claims

1. A resource allocation method, characterized in that, The method includes: Acquire target medical data within at least two target time periods; wherein the durations of any two target time periods are different; Based on the target medical data within each target time period, risk factors are predicted to obtain the medical risk factor prediction features corresponding to each target time period. Attention is calculated based on the medical risk factor prediction features of each target time period to obtain the attention weight corresponding to each target time period; The target fused medical factor features are obtained by weighted summation of at least two of the medical risk factor prediction features based on at least two of the attention weights. Based on the target fusion medical factor characteristics, the impact of preset insurance risk factors is predicted to obtain insurance impact prediction data; Resource gap prediction is performed based on the insurance impact prediction data to obtain resource gap prediction data, which is then used for resource allocation; wherein, the resource gap prediction data is used to represent the probability distribution of the resource gap.

2. The method according to claim 1, characterized in that, The step of predicting risk factors based on the target medical data within each target time period to obtain the medical risk factor prediction features corresponding to each target time period includes: Based on the target medical data within each target time period, time-series features are extracted to obtain medical time-series features; The medical time series features are assessed for risk factors using a target medical rule decision tree model to obtain the predicted features of the medical risk factors corresponding to each target time period.

3. The method according to claim 2, characterized in that, Before determining the risk factors of the medical time-series features using a target medical rule decision tree model to obtain the predicted medical risk factors for each target time period, the method further includes: Historical medical data is acquired, and the incidence rate is determined by the temporal characteristics of the historical medical data through a pre-constructed initial medical rule decision tree model to obtain the initial predicted sub-incidence rate. The initial predicted sub-incidence rate and the preset labeled incidence rate are calculated using a Bayesian optimization algorithm to obtain the target reward value. The parameters of the initial medical rule decision tree model are adjusted according to the target reward value to obtain the target medical rule decision tree model.

4. The method according to any one of claims 1 to 3, characterized in that, The target medical data consists of at least two data points, and any two data points are located in different geographical regions. The target fusion medical factor features are predicted based on the target medical data points located in the same geographical region. After predicting the impact of a preset insurance risk factor based on the target fused medical factor characteristics to obtain insurance impact prediction data, the method further includes: If the target fused medical factor feature of the geographical region is greater than or equal to the preset medical risk factor threshold, the geographical region is identified as a high-risk region. The insurance impact prediction data corresponding to the high-risk area is increased according to a preset first increase ratio.

5. The method according to claim 4, characterized in that, After amplifying the insurance impact prediction data corresponding to the high-risk area according to a preset first amplification ratio, the method further includes: The regional interval distance is obtained by calculating the distance between the high-risk area and each of the other geographical areas. Based on the regional interval distance and a preset distance threshold, the adjacent areas of the high-risk area are determined from the other geographical areas; The insurance impact prediction data corresponding to the adjacent areas is increased according to a preset second increase ratio; wherein the second increase ratio is less than the first increase ratio.

6. The method according to any one of claims 1 to 3, characterized in that, The insurance impact prediction data includes loss ratio impact data corresponding to at least two of the target time periods; The step of predicting resource gaps based on the insurance impact prediction data to obtain resource gap prediction data, and then allocating resources based on the resource gap prediction data, includes: A first time period is determined from at least two target time periods according to a preset first duration threshold range, and a second time period is determined from at least two target time periods according to a preset second duration threshold range; wherein the duration of the first time period is less than the duration of the second time period; Based on the payout ratio impact data of the first time period, resource gap prediction is performed to obtain short-term resource gap prediction data; Resource gap prediction is performed based on the payout ratio impact data for the second time period to obtain long-term resource gap prediction data, and resource allocation is performed based on the short-term resource gap prediction data and the long-term resource gap prediction data.

7. The method according to any one of claims 1 to 3, characterized in that, The step of calculating attention based on the predicted features of the medical risk factors for each target time period to obtain the attention weight corresponding to each target time period includes: The preset query weight matrix is ​​multiplied by the medical risk factor prediction features to obtain the medical query vector; The preset key weight matrix is ​​multiplied with the medical risk factor prediction features to obtain the medical key vector; The attention score is calculated by taking the product of the medical query vector and the transpose of the medical key vector and then the square root of the dimension of the medical key vector. The attention scores are normalized to obtain the attention weights corresponding to each target time period.

8. A resource allocation device, characterized in that, The device includes: The data acquisition module is used to acquire target medical data within at least two target time periods; wherein the durations of any two target time periods are different. The medical risk prediction module is used to predict risk factors based on the target medical data within each target time period, and to obtain the medical risk factor prediction features corresponding to each target time period. The attention calculation module is used to perform attention calculation based on the medical risk factor prediction features of each target time period to obtain the attention weight corresponding to each target time period. The risk factor fusion module is used to perform a weighted summation of at least two medical risk factor prediction features based on at least two attention weights to obtain the target fused medical factor features. The impact prediction module is used to predict the impact of preset insurance risk factors based on the target fused medical factor characteristics, and obtain insurance impact prediction data. The resource gap prediction module is used to predict the resource gap based on the insurance impact prediction data, and obtain resource gap prediction data, so as to allocate resources based on the resource gap prediction data; wherein, the resource gap prediction data is used to represent the probability distribution of the resource gap.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the resource allocation method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the resource allocation method according to any one of claims 1 to 7.

Images