Flood disaster early warning decision optimization method, equipment, medium and product

Abstract

The invention discloses a flood disaster early warning decision optimization method and device, a medium and a product, and relates to the field of environmental disaster management and prediction.The method comprises the steps that a decision feature matrix is determined according to a multi-dimensional decision situation; determining certainty and other values according to the decision feature matrix, and dynamically calibrating psychological parameters; a Gini index is improved according to the decision feature matrix and the weight of the flood event occurrence probability, and a split decision tree model is constructed based on the improved Gini index; determining a key split point on the flood event occurrence probability according to the trained split decision tree model; determining an adaptive decision threshold according to the flood event occurrence probability and the key split point; determining a framework effect compensation factor; adjusting an adaptive decision threshold by using the framework effect compensation factor; and determining a flood disaster early warning decision according to the adjusted threshold value. According to the invention, the accuracy of flood early warning decision and the system robustness can be improved.

Description

Technical Field

[0001] The present application relates to the field of environmental disaster management and prediction, and in particular to a flood disaster early warning decision optimization method, equipment, medium and product. Background Art

[0002] Flood warning decision-making technology is a key link in the modern flood prevention and disaster reduction system. It achieves accurate flood prediction, timely warning and scientific decision-making through the comprehensive application of multiple technical means. Existing flood warning decision-making technology mainly relies on the following two methods: (1) Traditional economic decision-making model: such as the cost-loss ratio model, which determines the decision threshold by calculating the ratio of warning cost to potential loss. However, this method has significant defects: for example, the psychological factors are missing and the irrational behavioral characteristics of decision-makers in risk probability perception (such as probability weight function distortion) and loss aversion tendency are not quantified; the static threshold is limited and uses fixed threshold parameters, which cannot reflect the change in risk attitude when the flood probability changes dynamically. (2) Single model application: Commonly used models include behavioral models such as prospect theory. Although they can characterize the risk preferences of decision-makers, they are not coupled with flood probability prediction models, resulting in a lack of real-time hydrological data support for decision results; machine learning classification models represented by decision tree models can handle classification decision problems, but they do not introduce psychological parameters and are not sensitive enough to decision bias caused by framing effects.

[0003] It can be seen that the existing technical problems are mainly concentrated in the following four aspects: (1) Lack of quantification of psychological behavior: Traditional models cannot quantify the dynamic changes in decision makers' risk attitudes in different flood probability intervals (such as risk aversion-seeking reversal in the 25%-50% probability interval), resulting in excessive deviation in the setting of warning thresholds; (2) Insufficient model synergy: In existing technologies, models operate independently, resulting in a break in the correlation between psychological parameters and hydrological prediction data, causing errors in cross-model data transmission; (3) Dynamic adaptability defects: Static threshold methods (such as fixed r=0.5) are difficult to adapt to real-time changes in flood probability, and the prediction accuracy decreases in cross-scenario tests; (4) Framing effect interference: A single model does not have a built-in framing effect correction mechanism. When the same decision problem is expressed in the form of "loss" or "gain", the probability threshold that triggers the risk attitude change varies greatly. Summary of the Invention

[0004] The purpose of this application is to provide a flood disaster warning decision optimization method, equipment, medium and product, which can improve the accuracy and system robustness of flood warning decisions, while reducing the risk of emergency resource misallocation due to decision-making deviations.

[0005] To achieve the above objectives, this application provides the following solutions: In a first aspect, the present application provides a flood disaster early warning decision optimization method, the flood disaster early warning decision optimization method comprising: Based on the flood warning decision-making experiment, a multidimensional decision-making scenario combining probability information and cost-benefit was constructed. A decision feature matrix was determined based on the multidimensional decision-making scenario. The multidimensional decision-making scenario used rainfall forecasts to deeply integrate the probability of flood events, the actual losses caused by floods, and the cost of issuing flood warnings as core parameters. The decision feature matrix included the core parameters and binary decision labels under different decision-making scenarios. Based on the decision feature matrix, the certainty equivalent value is determined, and the psychological parameters are dynamically calibrated to obtain the weight of the probability of flood events. The certainty equivalent value represents the quantitative value of the risk attitude of decision makers when taking different decisions when facing uncertain disaster forecasts. The Gini index is improved according to the decision feature matrix and the weight of the flood event probability, and a split decision tree model is constructed based on the improved Gini index; Determine the key splitting points on the probability of flood events based on the trained split decision tree model; the key splitting points are used to reflect the behavioral changes of decision makers in different probability intervals; Determine the adaptive decision threshold based on the probability of flood events and the corresponding key split points; According to the flood re-investigation report, the semantic feature vector is determined based on the word vector model; and the framing effect compensation factor is determined based on the semantic feature vector; The adaptive decision threshold is adjusted using the framing effect compensation factor to obtain the adjusted threshold; Flood disaster warning decisions are determined based on the adjusted thresholds.

[0006] Optionally, determining the value of certainty based on the decision feature matrix and dynamically calibrating the psychological parameters to obtain the weight of the probability of the flood event specifically includes: Using the formula Determine the value of certainty ;in, is the weight function of the probability of flood event occurrence, , P is the probability of flood events, δ∈(0.28, 1] is the probability perception distortion coefficient, is the cumulative prospect theory value function, , The economic utility of the decision-making consequences and the actual losses caused by floods or the costs of issuing flood warnings, ∈(0,1) is the risk attitude coefficient, is the loss aversion coefficient; The weight of the probability of flood events is obtained by fitting the deterministic equivalent values ​​according to the optimization algorithm.

[0007] Optionally, improving the Gini index according to the decision feature matrix and the weight of the flood event probability, and constructing a split decision tree model based on the improved Gini index, specifically includes: Using the formula Determine the improved Gini index ; in, is the weight of the probability of flood event occurring at the i-th node, is the category in the split point, is the binary decision label of the flood event of the i-th node, A tree node for the split decision tree model.

[0008] Optionally, determining the adaptive decision threshold according to the probability of occurrence of the flood event and the corresponding key split point specifically includes: Using the formula Construct a risk attitude-probability mapping matrix; in, is the adaptive decision threshold, and is the key splitting point, is the probability of flood events.

[0009] Optionally, determining a semantic feature vector based on a word vector model according to the flood re-investigation report; and determining a framing effect compensation factor according to the semantic feature vector specifically includes: Using the formula Determine the framing effect compensation factor ; in, and is a semantic feature vector, and includes flood loss keyword density and flood benefit keyword density.

[0010] Optionally, adjusting the adaptive decision threshold using the framing effect compensation factor to obtain the adjusted threshold specifically includes: Using the formula Determine the adjusted threshold ; in, is the adaptive decision threshold, is the framing effect compensation factor.

[0011] Optionally, determining a flood disaster warning decision based on the adjusted threshold value specifically includes: Determine the real-time cost-loss ratio based on the actual losses caused by the flood and the cost of issuing flood warnings; Determine whether the real-time cost-loss ratio is less than or equal to the adjusted threshold; if so, issue an alert; if not, do not issue an alert; Based on the judgment results, decision instructions with confidence ratings and risk attitude evolution curves are output.

[0012] In a second aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the flood disaster warning decision optimization method.

[0013] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the flood disaster warning decision optimization method.

[0014] In a fourth aspect, the present application provides a computer program product, comprising a computer program, which, when executed by a processor, implements the flood disaster warning decision optimization method.

[0015] According to the specific embodiments provided in this application, this application has the following technical effects: This application provides a flood disaster warning decision optimization method, device, medium, and product. This method constructs a multidimensional decision scenario that combines probability information with cost-benefit information. Based on this multidimensional decision scenario, a decision feature matrix is ​​determined. Based on this matrix, values ​​such as certainty are determined, and psychological parameters are dynamically calibrated to obtain weights for flood event probability. This improves the Gini index. This application implements a coupling algorithm for psychological parameters and the splitting criterion (Gini index), overcoming the dimensionality curse of traditional statistical methods in heterogeneous data processing. A joint calibration method for probability weight functions and value functions is established to address the multi-scale matching problem between psychological parameters and hydrological data. Based on the probability of a flood event and the corresponding key splitting points, an adaptive decision threshold is determined. Specifically, a two-dimensional mapping matrix between risk attitude and flood event probability is constructed, enabling adaptive adjustment of the threshold for different warning levels. The adaptive decision threshold is adjusted using the framing effect compensation factor, that is, a framing effect correction mechanism is set to reduce the decision-making bias caused by the framing effect; this application combines human behavior simulation in psychology with machine learning technology in computer science to quantify the risk preferences and judgment biases of decision makers in an uncertain environment, thereby optimizing the automated decision-making process in the flood warning system; this application is mainly used for early warning and risk analysis of hydrological and meteorological disasters, especially for efficient assessment of flood risks in dynamic environments, improving the accuracy and reliability of disaster response decisions, and reducing the risk of misallocation of emergency resources due to decision-making bias. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 This is a flow chart of a flood disaster early warning decision optimization method in one embodiment of the present application; Figure 2 This is a schematic diagram of the principle of a flood disaster early warning decision optimization method in one embodiment of the present application. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0019] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0020] In an exemplary embodiment, Figure 1 and Figure 2 As shown, a flood disaster early warning decision optimization method is provided, which includes the following S101 to S108. S101: Based on the flood warning decision-making experiment, construct a multidimensional decision-making scenario that combines probability information and cost-benefit; and determine a decision feature matrix based on the multidimensional decision-making scenario; the multidimensional decision-making scenario uses the probability of flood events P∈[0,1] deeply integrated with rainfall forecasts, the actual losses L caused by floods (including direct and indirect losses), and the cost of issuing flood warnings C (including the costs of issuing warnings and the costs of making incorrect decisions) as core parameters; the decision feature matrix includes the core parameters and binary decision labels under different decision-making scenarios; the binary decision label Y∈{0,1}, where 1 indicates that one of the 168 decision makers participating in the experiment issued a warning to the public, and 0 indicates that one of the 168 decision makers participating in the experiment did not issue a warning to the public; The probability of flood event occurrence P is discretized dynamically into four risk sub-intervals (divided into low flood probability (0%-25%), medium flood probability (25%-50%), high flood probability (50%-75%), and high flood probability (75%-100%)), and the cost loss ratio C / L is calculated; Output the decision feature matrix X = [P, C / L, Y], which serves as the basis for subsequent psychological parameter estimation and model training; S102: Determine the certainty equivalent value based on the decision feature matrix and dynamically calibrate the psychological parameters to obtain the weight of the probability of the flood event; the certainty equivalent value represents the quantitative value of the risk attitude of the decision maker when taking different decisions when facing uncertain disaster forecasts; The purpose of S102 is to quantify the risk preferences and behavioral biases of decision makers based on historical data (decision feature matrix); S102 specifically includes: (1) Given a fixed probability p of a flood event, give the decision maker the options of "issuing an alert" or "not issuing an alert" and the costs and losses associated with each decision. Then ask the decision maker which of the following options he or she should choose: 1. Accept the payment of M to avoid this decision; 2. Or bear the risk of facing costs or losses caused by the decision; 3. Adjust M according to the decision maker's choice until it makes no difference between the certain amount and the corresponding risk. The certain amount M is the certainty equivalent value CE of the risk prospect when the probability p of the flood event occurs.

[0021] (2) Calling the cumulative prospect theory value function: ; (3) Update the weight function of the probability of flood events: ; Among them, δ∈(0.28, 1] is the probability perception distortion coefficient, is the weight function of the probability of flood event, δ∈(0.28, 1] is the probability perception distortion coefficient, is the cumulative prospect theory value function, The economic utility of the decision-making consequences and the actual losses caused by floods or the costs of issuing flood warnings, ∈(0,1) is the risk attitude coefficient, is the loss aversion coefficient; (4) Calculating the value of certainty Certainty of value It represents the quantitative value of the risk attitude of decision makers when taking different decisions when facing uncertain disaster forecasts. By fitting the label Y through the optimization algorithm (i.e. gradient descent), it is estimated The value of , making the certainty equal to the value and the actual data collected through experiments Closest.

[0022] ; S103, improving the Gini index according to the decision feature matrix and the weight of the flood event probability, and constructing a split decision tree model based on the improved Gini index; the split decision tree model can reflect the decision maker's psychological bias; S103 specifically includes: (1) Use weighted samples to train the split decision tree model and predict the decision behavior label Y∈{0,1}. The probability weight of the weighted sample , adjust the importance that the split decision tree model attaches to samples with different weights, so that the splitting process reflects the psychological bias of the decision maker.

[0023] (2) When evaluating the split point, calculate the improved Gini index (the traditional split model Gini index formula is ),in is the category in the split point ratio, used to measure the impurity of the node): ; in, is the weight of the probability of flood event occurring at the i-th node, is the category in the split point, is the binary decision label of the flood event of the i-th node, A tree node for the split decision tree model.

[0024] S104, determining key splitting points on the probability of flood events based on the trained split decision tree model; the key splitting points are used to reflect the behavior changes of decision makers in different probability intervals; Using the trained split decision tree model, we identify key split points on the flood event probability P. The key split point serves as the threshold r, reflecting the decision maker's behavioral changes in different probability intervals (such as low risk or high risk).

[0025] S105, based on the probability of flood events and the corresponding key split points, an adaptive decision threshold is determined to achieve a dynamic transition of the threshold in the probability range of 25%-50%, eliminating the static threshold deviation; S105 specifically includes: Using the formula Construct a risk attitude-probability mapping matrix; in, is the adaptive decision threshold, and is the key splitting point, is the probability of flood events.

[0026] As a specific example, =0.81, which is the low risk threshold, indicating that decision makers are more sensitive to whether to issue flood warnings in low-risk situations. =0.53, which is the high-risk threshold, indicating that decision makers are more risk-averse in high-risk situations and are more inclined to take active preventive measures to avoid potential catastrophic losses as much as possible.

[0027] S106: Based on the flood re-investigation report and the word embedding model, a semantic feature vector is determined; a framing effect compensation factor is determined based on the semantic feature vector; the framing effect compensation factor is used to reduce the decision-making bias caused by the framing effect to within ±5%; As a specific embodiment, according to the decision problem, the text is expressed and the semantic feature vector S∈ℝ is extracted based on the word vector model. 6 (Including flood loss keyword density, flood benefit (early warning effect) keyword density) Using the formula Determine the framing effect compensation factor ; in, and is a semantic feature vector, and includes flood loss keyword density and flood benefit keyword density.

[0028] S107, adjusting the adaptive decision threshold using the framing effect compensation factor to obtain an adjusted threshold; Using the formula Determine the adjusted threshold ; S108: Determine a flood disaster warning decision based on the adjusted threshold.

[0029] S108 specifically includes: S1: Determine the real-time cost-loss ratio based on the actual losses caused by the flood and the cost of issuing flood warnings. ; S2, determining whether the real-time cost-loss ratio is less than or equal to the adjusted threshold; if so, issuing an early warning; if not, not issuing an early warning; ; S3, according to the judgment result , outputs decision instructions with confidence ratings and risk attitude evolution curves, and then demonstrates the impact of flood forecast probability, psychological elaboration and framing effects on decision-making.

[0030] Compared with the prior art, this application has the following effects: (1) Improvement of the accuracy of dynamic quantification of risk attitude: Psychological parameters ( ) and the splitting criterion can accurately capture the risk attitude reversal phenomenon in the 25%-50% flood event probability interval. Experimental data show that the decision accuracy in this interval is improved from 68% of the traditional model to 82%; the risk neutrality bias aversion threshold ( =0.81) and the risk aversion threshold ( =0.53) dynamic transition mechanism, which reduces the cross-scenario prediction error to 8.2% (the traditional static threshold model error is 21.5%); (2) Model synergy breakthrough: using the improved Gini index ( ) and the probability weight function ( ) embedded fusion, to achieve real-time interaction between psychological parameters and hydrological data, the data transmission error between models was reduced from 12.7% to 3.4%; and in the Ramos case test, the area under the ROC curve (AUC) of the joint model reached 0.89, which was significantly better than the single CPT model (AUC = 0.76) or DT model (AUC = 0.81); (3) Enhanced anti-interference ability of the frame effect: According to the semantic parsing compensation factor β and the dynamic threshold function ( ) product correction, eliminating the decision bias caused by different problem statements, reducing the trigger difference between the loss framework (40% threshold) and the gain framework (75% threshold) to within ±5%; and in the test set containing 145 groups of semantic interference, the false alarm rate was reduced from 22.3% to 9.1%; (4) Real-time decision efficiency optimization: According to the segmented threshold function ( The linear computing characteristics of the CPT model and pre-trained semantic vectors can reduce the single decision calculation time from 58ms in the traditional CPT model to 12ms, meeting the real-time response requirements of flood warnings (<50ms). It also reduces memory usage by 63% (from 2.1GB to 0.78GB), adapting to the deployment requirements of edge computing devices. The following are some specific examples to illustrate the beneficial effects of this application: Example 1: Core parameter calibration experiment A flood decision-making simulation experiment involving 168 emergency management personnel (including five probability scenarios and three loss frameworks) was conducted. The specific implementation steps were as follows: (1) Behavioral data collection Through the decision-making simulation interface, we collected the subjects' warning selection records (a total of 2520 decision data) under different flood event probabilities (P = 10%, 25%, 40%, 60%, 80%). We also recorded the response time, economic parameters (C / L∈[0.2, 1.8]), and problem statement framework (loss / benefit) for each decision. (2) Psychological parameter fitting The maximum likelihood estimation method was used to solve the cost function parameters, and the optimal parameters were obtained: α = 0.88 ± 0.03, γ = 0.65 ± 0.05, λ = 2.25 (95% confidence interval); (3) Split decision tree model training Input feature matrix X = [P, C / L, framework type], label Y = actual decision; use the improved Gini index (embedding w(p)) to split nodes, generate a tree structure with a depth of 5; obtain the key threshold r1 = 0.81 (split criterion when P < 25%); (4) Verification results 4.1 Risk Attitude Reversal Interval Test: In the P = 25%-50% interval, the decision maker's risk aversion rate dropped sharply from 82% to 37% (χ² = 29.7, p < 0.001); 4.2 Model goodness of fit: McFadden's R²=0.63, significantly better than the traditional logit model (R²=0.41); Example 2: Cross-scenario Verification (Ramos Case) Connect to the real-time hydrological monitoring network of the Tagus River Basin in Spain and obtain historical flood event records (145 warning decision data from 2010-2020). The specific implementation steps are as follows: (1) Data migration adaptation The original cost data (C∈[23,000,51,000]€) and loss data (L∈[87,000,324,000€) were converted with exchange rates and adjusted for purchasing power parity; a cross-cultural framing effect vocabulary was established: the Spanish expression "dañospotenciales" (potential losses) was mapped to vector; (2) Dynamic threshold generation Calculate T(P) based on the real-time flood event probability P(t): When P = 35%, T(P) = 0.81 + (0.63-0.81) × (35-25) / 25 = 0.738; The semantic parser detects that the warning text contains "evitar pérdidas" (avoid losses), triggering =1.15; (3) Decision instruction output Calculate the final threshold: T_final =0.738×1.15=0.849; When C / L=0.79 (C=42,000, L=53,000), 0.79<0.849 triggers an early warning; (4) Verification results Early warning accuracy: 76.4% (68.2% for traditional models), especially in the P=30%-45% range, where the accuracy increases to 81.3%; Response time: The average time from meteorological data input to decision output is 23ms, meeting the real-time warning requirements; Example 3: Verification of Framing Effect Correction The same decision problem is expressed in a loss framework ("possible loss of L million yuan") and a benefit framework ("possible savings of L million yuan"). The specific implementation process is as follows: (1) Semantic parser detects keyword frequency differences Loss frame: "loss" appears 3.2 times per sentence, and "disaster" appears 1.8 times per sentence; Benefit frame: "saving" appears 2.7 times per sentence, and "safety" appears 2.1 times per sentence; (2) Calculation of compensation factor In practical applications, normalization is used to limit ∈[0.9, 1.1]); Adjust the threshold: When P=40%, the original threshold T(P)=0.72, then =0.72×1.1=0.792; (3) Verification results The C / L threshold for triggering an early warning under the loss framework has increased from 0.72 to 0.792, and the difference with the threshold under the gain framework has narrowed from 35% to 8.7%; In a double-blind test, the decision-makers' agreement on different frameworks increased from 61% to 89%. Based on the above examples, the application has been verified to have good parameter calibration accuracy (α error < ± 0.05), real-time response capability (< 50ms) and anti-interference performance ( Correction effectiveness>87%) and other aspects of technological progress have provided a feasible intelligent solution for flood warning decision-making.

[0031] In an exemplary embodiment, a computer device is provided, which may be a server or a terminal. The computer device includes a processor, memory, an input / output (I / O) interface, and a communication interface. The processor, memory, and I / O interface are connected via a system bus, and the communication interface is connected to the system bus via the I / O interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The I / O interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal via a network connection. When executed by the processor, the computer program implements a flood disaster early warning decision optimization method.

[0032] In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0033] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0034] In an exemplary embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0035] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0036] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. In particular, any reference to memory, database, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0037] The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units based on quantum computing, and the like.

[0038] In this application, all actions to obtain signals, information or data are carried out in compliance with the relevant data protection laws and policies of the country where they are located and with the authorization given by the owner of the corresponding device.

[0039] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0040] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A flood disaster early warning decision optimization method, characterized in that: The flood disaster early warning decision optimization method includes: Based on the flood warning decision-making experiment, a multidimensional decision-making scenario combining probability information and cost-benefit was constructed. A decision feature matrix was determined based on the multidimensional decision-making scenario. The multidimensional decision-making scenario used rainfall forecasts to deeply integrate the probability of flood events, the actual losses caused by floods, and the cost of issuing flood warnings as core parameters. The decision feature matrix included the core parameters and binary decision labels under different decision-making scenarios. Based on the decision feature matrix, the certainty equivalent value is determined, and the psychological parameters are dynamically calibrated to obtain the weight of the probability of flood events. The certainty equivalent value represents the quantitative value of the risk attitude of decision makers when taking different decisions when facing uncertain disaster forecasts. The Gini index is improved according to the decision feature matrix and the weight of the flood event probability, and a split decision tree model is constructed based on the improved Gini index; Determine the key splitting points on the probability of flood events based on the trained split decision tree model; the key splitting points are used to reflect the behavioral changes of decision makers in different probability intervals; Determine the adaptive decision threshold based on the probability of flood events and the corresponding key split points; According to the flood re-investigation report, the semantic feature vector is determined based on the word vector model; and the framing effect compensation factor is determined based on the semantic feature vector; The adaptive decision threshold is adjusted using the framing effect compensation factor to obtain the adjusted threshold; Flood disaster warning decisions are determined based on the adjusted thresholds.

2. The flood disaster early warning decision optimization method according to claim 1 is characterized in that: The above method determines the value of certainty based on the decision feature matrix, and dynamically calibrates the psychological parameters to obtain the weight of the probability of flood event occurrence, which specifically includes: Using the formula Determine the value of certainty ;in, is the weight function of the probability of flood event occurrence, , P is the probability of flood events, δ∈(0.28, 1] is the probability perception distortion coefficient, is the cumulative prospect theory value function, , The economic utility of the decision-making consequences and the actual losses caused by floods or the costs of issuing flood warnings, ∈(0,1) is the risk attitude coefficient, is the loss aversion coefficient; The weight of the probability of flood events is obtained by fitting the deterministic equivalent values ​​according to the optimization algorithm.

3. The flood disaster early warning decision optimization method according to claim 1 is characterized in that: The method of improving the Gini index according to the decision feature matrix and the weight of the flood event probability and constructing a split decision tree model based on the improved Gini index specifically includes: Using the formula Determine the improved Gini index ; in, is the weight of the probability of flood event occurring at the i-th node, is the category in the split point, is the binary decision label of the flood event of the i-th node, A tree node for the split decision tree model.

4. The flood disaster early warning decision optimization method according to claim 1, characterized in that: Determining the adaptive decision threshold based on the probability of a flood event and the corresponding key split point specifically includes: Using the formula Construct a risk attitude-probability mapping matrix; in, is the adaptive decision threshold, and is the key splitting point, is the probability of flood events occurring.

5. The flood disaster early warning decision optimization method according to claim 1 is characterized in that: Determining semantic feature vectors based on the flood re-disaster report and the word vector model; The framing effect compensation factor is determined based on the semantic feature vector, specifically including: Using the formula Determine the framing effect compensation factor ; in, and is a semantic feature vector, and includes flood loss keyword density and flood benefit keyword density.

6. The flood disaster early warning decision optimization method according to claim 1 is characterized in that: The method of adjusting the adaptive decision threshold using the framing effect compensation factor to obtain the adjusted threshold specifically includes: Using the formula Determine the adjusted threshold ; in, is the adaptive decision threshold, is the framing effect compensation factor.

7. The flood disaster early warning decision optimization method according to claim 1 is characterized in that: Determining flood disaster warning decisions based on the adjusted thresholds specifically includes: Determine the real-time cost-loss ratio based on the actual losses caused by the flood and the cost of issuing flood warnings; Determine whether the real-time cost-loss ratio is less than or equal to the adjusted threshold; if so, issue an alert; if not, do not issue an alert; Based on the judgment results, decision instructions with confidence ratings and risk attitude evolution curves are output.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the flood disaster warning decision optimization method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the flood disaster early warning decision optimization method according to any one of claims 1 to 7 is implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the flood disaster early warning decision optimization method according to any one of claims 1 to 7 is implemented.

Images