A case-based reasoning method for recommending drought emergency measures

Abstract

The application discloses a kind of drought emergency measures recommendation method based on case reasoning, by extracting the key features of this drought, the global similarity between cases is calculated, and the matching cases are retrieved from the historical case library.Then based on matching cases, the emergency response level and drought-resistant investment funds of this drought are obtained by cross variation and other operations.And according to the emergency response level, the emergency measures corresponding to the level in the drought response measures library are connected, and the drought response measures recommendation is realized.The emergency response level and drought-resistant investment funds of the application are based on the response measures correction of historical cases, so they can better fit the actual situation of this drought, and make the drought response more targeted on the basis of reasonable response measures.

Description

Technical Field

[0001] This invention relates to the field of disaster assessment technology, and more specifically, to a method for recommending drought emergency measures based on case-based reasoning. Background Technology

[0002] Drought is one of the most significant natural disasters facing the world, characterized by its high frequency, long duration, and wide-ranging impact. As a meteorological disaster, drought has long plagued industrial and agricultural production. With the increasing risk of drought events, the situation for disaster prevention and mitigation is becoming increasingly severe, and risk prevention and response have become important research topics for the international community in addressing drought disasters.

[0003] Currently, drought event response decisions globally largely rely on the expertise and experience of specialists in relevant fields. However, the range of response measures is insufficient, their adaptability to specific drought events is poor, and they fail to consider the needs of multi-scale and multi-sectoral decision-making. Faced with these challenges, there is an urgent need to strengthen research on drought event decision support and establish an intelligent recommendation system for multi-scale drought event response strategies to provide drought event risk prevention decision support for disaster risk management departments at all levels.

[0004] Case-based reasoning is an early and relatively mature reasoning method in the field of artificial intelligence, and it has received considerable attention since the 1980s. Its ability to reason without requiring specialized knowledge completely overcomes the limitations of retrieving specialized domain knowledge. Case-based reasoning has wide applications in emergency decision-making.

[0005] The solution process for case-based reasoning can be summarized by the R-4 case-based reasoning cognitive model (Retrieve-Reuse-Revise-Retain) proposed by Aamodt and Plaza in 1994. This model is also the most widely used model to describe case-based reasoning theory. This model divides the main process of case-based reasoning into case retrieval, case reuse, case revision, and case preservation. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a case-based reasoning method for recommending drought emergency measures, thereby accelerating the disaster assessment process and providing corresponding emergency measures.

[0007] To achieve the above-mentioned objectives, the present invention provides a method for recommending drought emergency measures based on case-based reasoning, characterized by comprising the following steps:

[0008] (1) Extract the key features of this drought, including: basic drought information, drought meteorological information, drought index information and socio-economic information;

[0009] Among them, basic drought information includes drought type, drought cycle and affected area;

[0010] Drought meteorological information includes precipitation, temperature, wind speed, and relative humidity;

[0011] Drought indicators include the standard precipitation index, the comprehensive drought index, and the drought intensity.

[0012] Socioeconomic information includes gross national product and the number of people affected by disasters;

[0013] (2) Based on the key characteristics of this drought, search for similar cases in the drought case database;

[0014] (2.1) Calculate the similarity between the key features of this drought and the key features of each historical case in the drought case database;

[0015] For key features of numerical data, the standard Euclidean distance between key features is used as the feature similarity.

[0016] ;

[0017] in, The first of the key characteristics of this drought One key feature The first key feature of historical cases One key feature , These represent the first and second cases in the historical case database, respectively. The maximum and minimum values ​​of the key features;

[0018] For features of enumeration type, feature similarity is calculated using the following formula;

[0019] ;

[0020] (2.2) Calculate the weights of key features using the analytic hierarchy process (AHP);

[0021] (2.2.1) Construct a two-layer structure model of drought characteristics, wherein the first layer contains basic drought information, drought meteorological information, drought index information and socio-economic information, and the second layer contains sub-items of the various information in the first layer;

[0022] (2.2.2) Based on the two-layer structure model, the importance of features in the same layer is compared pairwise, and a discriminant matrix is ​​constructed using the 1-9 scaling method. ,in, Indicates the first The first feature and the second The relative importance of each feature;

[0023] (2.2.3) Find the discriminant matrix The largest eigenvalue in the middle Corresponding feature vector Then, for the feature vector After normalization, the weights are used as the weights of the corresponding layers, denoted as... ;

[0024] (2.2.4) Calculate the discriminant matrix Consistency index ;

[0025] ;

[0026] in, This indicates the number of key features in each layer;

[0027] (2.2.5) Calculate the random consistency ratio ;

[0028] ;

[0029] in, The introduced average random consistency index;

[0030] (2.2.6), judgment Is it less than the preset threshold? ,like Then the weight If valid, otherwise return to step (2.2.2);

[0031] (2.2.7) Weights of key features of drought;

[0032] Following the method in steps (2.2.2)-(2.2.6), let the weight of the first layer in the two-layer structure model be... The weights of the second layer are Then the weight vector of key features of drought ;

[0033] (2.3) Calculate the structural similarity coefficient ;

[0034] ;

[0035] Where A represents the set of key features of this drought, and B represents the set of key features of historical cases. The sum of the weights of all key features at the intersection of A and B. This represents the sum of the weights of all key features of the union of A and B;

[0036] (2.4) Calculate global similarity;

[0037] ;

[0038] in, Represents the weight vector The weight value of the i-th key feature;

[0039] (3) Search the drought case database through global similarity and select the N historical cases with the highest similarity as matching cases;

[0040] (4) Revise the emergency measures for matching cases;

[0041] (4.1) Population initialization;

[0042] Set matching case population , , Indicates the first One matching case;

[0043] Set up emergency measures population , , Indicates the first The emergency measures corresponding to each matched case include drought response level and drought relief investment funds;

[0044] Let this drought be set as the target individual, denoted as... ;

[0045] (4.2) Calculate the population of matching cases Each individual and target individuals fitness between ;

[0046] ;

[0047] in, Represents an individual The Middle One key feature Represents the target individual The Middle Key features;

[0048] (4.3) Determining the optimal individual and : Matching case population In this process, the individual with the highest fitness value is designated as the best individual. At the same time, the best individual The corresponding emergency measures are recorded as follows: ;

[0049] (4.4) Determining the best individual If the fitness value is greater than the preset threshold or the current iteration number reaches the preset maximum iteration number, proceed to step (4.7); otherwise, proceed to step (4.5).

[0050] (4.5) Update the matching case population and emergency response population ;

[0051] ;

[0052] in, To match the case population Any two individuals in the set, and satisfying ; For emergency response populations Any two individuals in the set, and satisfying ;

[0053] (4.6) Increment the current iteration count by 1, and then return to step (4.2) based on the updated population.

[0054] (4.7) Output the best individual And the corresponding emergency measures are recorded as ;

[0055] (5) Provide emergency measures for this drought;

[0056] Read the best individual Corresponding emergency measures Extract emergency measures The drought response level and drought relief investment funds are used to select corresponding emergency measures from the drought response measures database based on the drought response level. Then, the emergency measures and drought relief investment funds are used as recommended measures for this drought and stored in the drought case database.

[0057] The objective of this invention is achieved as follows:

[0058] This invention presents a case-based reasoning method for recommending drought emergency response measures. It extracts key features of the current drought, calculates the global similarity between cases, and retrieves matching cases from a historical case database. Then, based on these matching cases, operations such as crossover and mutation are used to obtain the emergency response level and drought relief investment funds for the current drought. Finally, based on the emergency response level, corresponding emergency measures in the drought response measure database are matched to achieve drought response measure recommendations. Since the emergency response level and drought relief investment funds of this invention are derived from adjustments made to response measures based on historical cases, they can better align with the actual situation of the current drought, making drought response more targeted while ensuring the rationality of the measures.

[0059] Meanwhile, the drought emergency response recommendation method based on case-based reasoning of this invention also has the following beneficial effects:

[0060] (1) Case reasoning is introduced into the recommendation of drought response measures. The recommendation of drought response measures can be realized by relying only on historical cases and response measure database. This reduces the reliance on complex judgment rules and drought resistance expertise, reduces the complexity of the method, and improves the operability of the method.

[0061] (2) Introducing structural similarity coefficients in the global similarity calculation of cases can reduce the error caused by the lack of key features of cases, thereby improving the accuracy of case retrieval;

[0062] (3) The mutation operator used when updating the population considers both the best individual in the current population and random individuals. This not only improves the convergence speed, but also avoids the convergence process from getting stuck in local optima to a certain extent, effectively improving the speed and accuracy of case correction. Attached Figure Description

[0063] Figure 1 This is a flowchart of the drought emergency response recommendation method based on case-based reasoning of the present invention;

[0064] Figure 2 This is a hierarchical structure diagram of key features in drought cases;

[0065] Figure 3 This is a comparison chart of case retrieval accuracy. Detailed Implementation

[0066] The specific embodiments of the present invention will now be described with reference to the accompanying drawings to enable those skilled in the art to better understand the invention. It should be particularly noted that in the following description, detailed descriptions of known functions and designs that might obscure the main content of the invention will be omitted here.

[0067] Example

[0068] Figure 1 This is a flowchart of the drought emergency response recommendation method based on case-based reasoning, as described in this invention.

[0069] In this embodiment, as Figure 1 As shown, the present invention provides a method for recommending drought emergency measures based on case-based reasoning, comprising the following steps:

[0070] (1) Extract the key characteristics of this drought;

[0071] In this embodiment, the characteristic elements of drought include: basic drought information, drought meteorological information, drought index information, socio-economic information, and disaster information;

[0072] The basic information on drought includes drought type, start and end time, and affected area;

[0073] Drought meteorological information includes precipitation, temperature, wind speed, relative humidity, etc.

[0074] Drought indicators include standard precipitation index, comprehensive drought index, drought intensity, etc.

[0075] Socioeconomic information includes GDP, disaster-affected population, etc.

[0076] Disaster information includes direct economic losses, affected crop area, and number of people affected;

[0077] The key characteristics of drought mainly include: basic drought information, drought meteorological information and drought indicator information, and socio-economic information;

[0078] Among them, basic drought information includes drought type, drought cycle and affected area;

[0079] Drought meteorological information includes precipitation, temperature, wind speed, and relative humidity;

[0080] Drought indicators include the standard precipitation index, the comprehensive drought index, and the drought intensity.

[0081] Socioeconomic information includes gross national product and the number of people affected by disasters;

[0082] (2) Based on the key characteristics of this drought, search for similar cases in the drought case database;

[0083] (2.1) Calculate the similarity between the key features of this drought and the key features of each historical case in the drought case database;

[0084] For key features of numerical data, the standard Euclidean distance between key features is used as the feature similarity.

[0085] ;

[0086] in, The first of the key characteristics of this drought One key feature The first key feature of historical cases One key feature , These represent the first and second cases in the historical case database, respectively. The maximum and minimum values ​​of the key features;

[0087] For features of enumeration type, feature similarity is calculated using the following formula;

[0088] ;

[0089] (2.2) Calculate the weights of key features using the analytic hierarchy process (AHP);

[0090] (2.2.1) Construct a two-layer structural model of drought characteristics, such as Figure 2 As shown, the first layer contains basic drought information, drought meteorological information, drought index information, and socio-economic information, while the second layer contains sub-items of the various types of information in the first layer.

[0091] (2.2.2) Based on the two-layer structure model, the importance of features in the same layer is compared pairwise, and a discriminant matrix is ​​constructed using the 1-9 scaling method. ,in, Indicates the first The first feature and the second The relative importance of each feature and the definition of the 1-9 annotation method are shown in Table 1.

[0092]

[0093] Table 1

[0094] (2.2.3) Find the discriminant matrix The largest eigenvalue in the middle Corresponding feature vector Then, for the feature vector After normalization, the weights are used as the weights of the corresponding layers, denoted as... ;

[0095] (2.2.4) Calculate the discriminant matrix Consistency index ;

[0096] ;

[0097] in, This indicates the number of key features in each layer;

[0098] (2.2.5) Calculate the random consistency ratio ;

[0099] ;

[0100] in, The introduced average random consistency index, The value of depends on the dimension of the discrimination matrix, as shown in Table 2;

[0101]

[0102] Table 2

[0103] (2.2.6), judgment Is it less than the preset threshold? =0.1, if Then the weight If valid, otherwise return to step (2.2.2);

[0104] (2.2.7) Weights of key features of drought;

[0105] Following the method in steps (2.2.2)-(2.2.6), let the weight of the first layer in the two-layer structure model be... The weights of the second layer are Then the weight vector of key features of drought ;

[0106] (2.3) Calculate the structural similarity coefficient ;

[0107] In similarity calculation, to reduce the errors that may be caused by missing key features and improve the accuracy of case retrieval, a structural similarity coefficient is introduced. The calculation formula is as follows:

[0108] ;

[0109] Where A represents the set of key features of this drought, and B represents the set of key features of historical cases. The sum of the weights of all key features at the intersection of A and B. This represents the sum of the weights of all key features of the union of A and B;

[0110] ;

[0111] in, Represents the weight vector The weight value of the i-th key feature;

[0112] (3) Search the drought case database through global similarity and select the N historical cases with the highest similarity as matching cases;

[0113] (4) Revise the emergency measures for matching cases;

[0114] (4.1) Population initialization;

[0115] Set matching case population , , Indicates the first One matching case;

[0116] Set up emergency measures population , , Indicates the first The emergency measures corresponding to each matched case include drought response level and drought relief investment funds;

[0117] Let this drought be set as the target individual, denoted as... ;

[0118] (4.2) Calculate the population of matching cases Each individual and target individuals fitness between ;

[0119] ;

[0120] in, Represents an individual The Middle One key feature Represents the target individual The Middle Key features;

[0121] (4.3) Determining the optimal individual and : Matching case population In this process, the individual with the highest fitness value is designated as the best individual. At the same time, the best individual The corresponding emergency measures are recorded as follows: ;

[0122] (4.4) Determining the best individual If the fitness value is greater than the preset threshold or the current iteration number reaches the preset maximum iteration number, proceed to step (4.7); otherwise, proceed to step (4.5).

[0123] (4.5) Update the matching case population and emergency response population ;

[0124] To improve convergence speed and avoid getting trapped in local optima, a hybrid update method is used when updating the population. This method includes both the current best individual and random individuals from the population. The update formula is shown below.

[0125] ;

[0126] in, To match the case population Any two individuals in the set, and satisfying ; For emergency response populations Any two individuals in the set, and satisfying ;

[0127] (4.6) Increment the current iteration count by 1, and then return to step (4.2) based on the updated population.

[0128] (4.7) Output the best individual And the corresponding emergency measures are recorded as ;

[0129] (5) Provide emergency measures for this drought;

[0130] Read the best individual Corresponding emergency measures Extract emergency measures The drought response level and drought relief investment funds are used to select corresponding emergency measures from the drought response measures database based on the drought response level. Then, the emergency measures and drought relief investment funds are used as recommended measures for this drought and stored in the drought case database.

[0131] In this embodiment, taking the drought in Province A in 2019 as an example, the five historical cases with the highest similarity were retrieved from the drought case database as follows: drought in Province A in 2015, drought in Province B in 2017, drought in City C in 2017, drought in Province B in 2019, and drought in City C in 2019.

[0132] In the absence of local key features, case retrieval using structural similarity coefficients is more accurate, such as... Figure 3 As shown;

[0133] Finally, the recommended measures for this drought can be given through the particle swarm optimization algorithm, as shown in Table 3.

[0134]

[0135] Table 3

[0136] Although the illustrative specific embodiments of the present invention have been described above to enable those skilled in the art to understand the invention, it should be understood that the invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the invention as defined and determined by the appended claims, and all inventions utilizing the concept of the present invention are protected.

Claims

1. A method for recommending drought emergency response measures based on case-based reasoning, characterized in that, Includes the following steps: (1) Extract the key features of this drought, including: basic drought information, drought meteorological information, drought index information and socio-economic information; Among them, basic drought information includes drought type, drought cycle and affected area; Drought meteorological information includes precipitation, temperature, wind speed, and relative humidity; Drought indicators include the standard precipitation index, the comprehensive drought index, and the drought intensity. Socioeconomic information includes gross national product and the number of people affected by disasters; (2) Based on the key characteristics of this drought, search for similar cases in the drought case database; (2.1) Calculate the similarity between the key features of this drought and the key features of each historical case in the drought case database; For key features of numerical data, the standard Euclidean distance between key features is used as the feature similarity. ； in, The first of the key characteristics of this drought One key feature The first key feature of historical cases One key feature , These represent the first and second cases in the historical case database, respectively. The maximum and minimum values ​​of the key features; For features of enumeration type, feature similarity is calculated using the following formula; ； (2.2) Calculate the weights of key features using the analytic hierarchy process (AHP); (2.2.1) Construct a two-layer structure model of drought characteristics, wherein the first layer contains basic drought information, drought meteorological information, drought index information and socio-economic information, and the second layer contains sub-items of the various information in the first layer; (2.2.2) Based on the two-layer structure model, the importance of features in the same layer is compared pairwise, and a discriminant matrix is ​​constructed using the 1-9 scaling method. ,in, Indicates the first The first feature and the second The relative importance of each feature; (2.2.3) Find the discriminant matrix The largest eigenvalue in the middle Corresponding feature vector Then, for the feature vector After normalization, the weights are used as the weights of the corresponding layers, denoted as... ; (2.2.4) Calculate the discriminant matrix Consistency index ; ； in, This indicates the number of key features in each layer; (2.2.5) Calculate the random consistency ratio ; ； in, The introduced average random consistency index; (2.2.6), judgment Is it less than the preset threshold? ,like Then the weight If valid, otherwise return to step (2.2.2); (2.2.7) Weights of key features of drought; Following the method in steps (2.2.2)-(2.2.6), let the weight of the first layer in the two-layer structure model be... The weights of the second layer are Then the weight vector of key features of drought ; (2.3) Calculate the structural similarity coefficient ; ； Where A represents the set of key features of this drought, and B represents the set of key features of historical cases. The sum of the weights of all key features at the intersection of A and B. This represents the sum of the weights of all key features of the union of A and B; (2.4) Calculate global similarity; ； in, Represents the weight vector The weight value of the i-th key feature; (3) Search the drought case database through global similarity and select the N historical cases with the highest similarity as matching cases; (4) Revise the emergency measures for matching cases; (4.1) Population initialization; Set matching case population , , Indicates the first One matching case; Set up emergency measures population , , Indicates the first The emergency measures corresponding to each matched case include drought response level and drought relief investment funds; Let this drought be set as the target individual, denoted as... ; (4.2) Calculate the population of matching cases Each individual and target individuals fitness between ; ； in, Represents an individual The Middle One key feature Indicating among the target individuals No. Key features; (4.3) Determining the optimal individual and : Matching case population In this process, the individual with the highest fitness value is designated as the best individual. At the same time, the best individual The corresponding emergency measures are recorded as follows: ; (4.4) Determining the best individual If the fitness value is greater than the preset threshold or the current iteration number reaches the preset maximum iteration number, proceed to step (4.7); otherwise, proceed to step (4.5). (4.5) Update the matching case population and emergency response population ; ； in, To match the case population Any two individuals in the set, and satisfying ; For emergency response populations Any two individuals in the set, and satisfying ; (4.6) Increment the current iteration count by 1, and then return to step (4.2) based on the updated population. (4.7) Output the best individual And the corresponding emergency measures are recorded as ; (5) Provide emergency measures for this drought; Read the best individual Corresponding emergency measures Extract emergency measures The drought response level and drought relief investment funds are used to select corresponding emergency measures from the drought response measures database based on the drought response level. Then, the emergency measures and drought relief investment funds are used as recommended measures for this drought and stored in the drought case database.

Images