CN115829061A - Emergency accident disposal method based on historical case and empirical knowledge learning - Google Patents
Emergency accident disposal method based on historical case and empirical knowledge learning Download PDFInfo
- Publication number
- CN115829061A CN115829061A CN202310144018.8A CN202310144018A CN115829061A CN 115829061 A CN115829061 A CN 115829061A CN 202310144018 A CN202310144018 A CN 202310144018A CN 115829061 A CN115829061 A CN 115829061A
- Authority
- CN
- China
- Prior art keywords
- disposal
- emergency
- decision
- treatment
- learning
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses an emergency incident disposal method based on historical case and experience knowledge learning, which is characterized in that operation experiences in a historical emergency disposal scheme are mined in a machine learning mode, and an emergency disposal decision knowledge model containing multidimensional characteristics such as an accident type, an accident position, a disposal method, disposal force, disposal equipment and the like is constructed. Meanwhile, the treatment scheme generation process and the back business logic can be interpreted and characterized in a thinking guide graph form, and the user with the learned decision knowledge can modify and take effect instantly. The method mainly solves the problem that the traditional fixed rule type emergency disposal scheme generation method for the emergency accidents has no learning updating capability, and by the method, the machine can obtain the emergency disposal scheme in second-level reasoning according to the accident types and the position conditions, and the generation accuracy of the emergency disposal scheme can be continuously improved along with the use of a user.
Description
Technical Field
The invention belongs to the technical field of emergency disposal, and particularly relates to an emergency accident disposal method based on historical case and experience knowledge learning.
Background
In the face of fire, debris flow and other sudden accidents, the emergency treatment task can be completed by adopting the treatment method and selecting which treatment force and which treatment equipment are used, so that the optimal treatment effect, the lowest cost, the best timeliness and the like can be ensured, and the emergency treatment task contains rich expert experience and has greater challenge for novice. The traditional scheme generation mode based on manual rule hard coding, such as 'coal mine accident site disposal scheme automatic generation system research based on expert system' (coal engineering, 11 years 2019), and the like, does not have learning and updating capabilities, and the recommendation accuracy is difficult to improve because users cannot see and change the rules; mainstream AI techniques such as deep learning are difficult to land due to lack of sufficient sample data. The emergency treatment experience knowledge learning method capable of learning and explaining needs to be provided, and an emergency treatment scheme is formed by fast and accurately reasoning by using expert experiences contained in historical emergency treatment cases.
Disclosure of Invention
The purpose of the invention is as follows: the invention aims to solve the technical problem of providing an emergency accident disposal method based on historical case and empirical knowledge learning aiming at the defects of the prior art. In order to solve the technical problem, the invention discloses an emergency accident handling method based on historical case and empirical knowledge learning, which comprises the following steps.
And 2, learning to excavate emergency disposal decision experience according to disposal selection results of the user under different scene characteristics, and learning to construct an emergency disposal decision knowledge model.
And 3, generating an emergency disposal scheme suitable for the current situation through the emergency disposal decision knowledge model reasoning according to the scene characteristics of the current emergency disposal task, and executing the emergency disposal scheme.
Further, step 1 comprises:
step 1-1, collecting all reliable historical emergency disposal schemes to form a historical emergency disposal scheme library.
Step 1-2, extracting scene features of each emergency disposal scheme in a historical emergency disposal scheme library and disposal selection result information corresponding to the scene features to form a training data set for emergency accident disposal experience learning; the scene characteristics comprise accident types, accident positions, treatment time limits and meteorological conditions, and the treatment selection result information comprises treatment methods, treatment force and quantity, treatment equipment and quantity results. For each training data, the scene characteristics such as accident type, accident position, disposal time limit, meteorological conditions and the like are equivalent to the input of the model, the combination of disposal selection results is equivalent to the label of the data, and the subsequent emergency disposal decision knowledge model training can be guided by comparing the output between the models and the difference of the label.
Further, in step 2, a Gradient Boosting Decision Tree (GBDT) model is used to learn and mine an emergency disposal Decision experience. GBDT is an integrated learning model based on decision tree algorithm as a classifier, and can classify data by continuously reducing the loss function generated in the training process.
Further, in step 2, the gradient lifting decision Tree uses a CART (Classification And Regression Tree) decision Tree as a base classifier. The CART decision tree can be used for classification and regression, has the characteristics of high training efficiency, good generalization performance and the like compared with decision tree algorithms such as ID3 and C4.5, and can well solve the emergency treatment decision problem related to the method.
Further, step 2 comprises:
and 2-1, reading a training data set learned by emergency accident disposal experience, and extracting disposal selection result information of the user under different emergency disposal conditions.
Step 2-2, inputting the scene characteristics as a gradient lifting decision tree model, and outputting disposal selection result information as the gradient lifting decision tree model; let the label category number of a certain treatment selection result in the training data set beK，K≧ 2, log-likelihood as loss function:
wherein:yis a value for the class of the tag,f(x) For the gradient boosting decision tree model prediction function,p k (x) For training samplesxBelonging to the class of labelskThe probability of (a) of (b) being,d k = {0,1} represents whether or not to belong to the second placekClass label, 1 means yes, 0 means no.
2-3, training a CART decision tree for each type of label, and obtaining a strong classifier of each type of label through a multi-round iteration modef k (x) And then, the prediction probability values of all the classifiers are compared to determine the final output treatment decision result.
Step 2-4, according to scene characteristicsKAnd integrating the decision branches obtained by the classifiers to obtain a gradient lifting decision tree model of the treatment selection result.
And 2-5, repeatedly executing the steps 2-3 to 2-4 to obtain gradient lifting decision tree models of other treatment selection results.
And 2-6, respectively inputting all scene characteristics in the training data set into each trained gradient lifting decision tree model, and outputting a corresponding emergency disposal scheme, thereby constructing an emergency disposal decision knowledge model.
Through the step 2, the decision knowledge can be continuously updated, and the defect that the traditional fixed rule type emergency disposal scheme generation method for the emergency accidents has no learning and updating capability is overcome.
Further, step 2-3 comprises:
step 2-3-1, initializing a classifier:
wherein:f k0 (x) Is a firstkInitial classifier of class labels (i.e., number of iteration rounds)m=0），KIs the label category number.
Step 2-3-2, for given iteration round numberMCalculating samples at each iterationx i Corresponding categorykNegative gradient error ofr mik ，i=1,2,…,N，NIn order to train the number of samples,m=1,2,…,M，Mis an integer greater than 1.
Step 2-3-3, calculating the optimal negative gradient fitting value of each leaf node in the classifierc mjk ：
Wherein:j=1,2,…,J，Jthe number of the leaf nodes is the number,R mjk is as followsmIn round iteration belong tokThe set of leaf nodes of the class label,R mk is as followsmIn round iteration belong tokAll sample sets of class labels.
Step 2-3-4, updating the classifier:
when the preset number of iteration rounds of the algorithm is reachedMThen stopping iteration to obtain strong classifierf k (x)=f kM (x)。
And 2-3-5, comparing the prediction probability values of all the classifiers and determining the final output treatment decision result.
Further, each iteration of step 2-3-2 is a samplex i Corresponding categorykNegative gradient error representation ofr mik Comprises the following steps:
wherein:y ik is a samplex i The corresponding tag class value is set to the tag class value,Tis the current date of the day of the week,t i is a samplex i The modification or selection date of the representative treatment case,p k (x i ) Is a samplex i Belonging to each label categorykProbability of (c):
wherein:f km (x i ) Is as followsmWhen the wheel iterates, for the secondkAnd a classifier for classifying the class labels.
With the continuous accumulation and improvement of the emergency disposal decision knowledge, part of the historical cases formed by using the outdated decision knowledge have negative influence on model training and are not suitable for being used as training bases. Therefore, the method calculates the sample on the basis of the gradient lifting decision tree combined with the CART algorithmx i Corresponding categorykA forgetting function is introduced when the gradient error is negative.
Is obviously whent i =TWhile, the samplex i The negative gradient error of (2) will all contribute to the fitting of the model when the sample is takenx i The date of modification/selection of (d) is currently longer,t i <<Twhile, the samplex i The effect of the negative gradient error is close to 1/2 of the original.
Further, step 4 is included, the emergency treatment decision knowledge model may interpret the characterization, including:
step 4-1, displaying emergency disposal schemes under different scene characteristics in the emergency disposal decision knowledge model in the step 2 in a thinking graph form;
and 4-2, automatically positioning and displaying a decision branch according to the emergency disposal scheme generated in the step 3 on a thinking map of the emergency disposal decision knowledge model, so as to explain the business logic for representing the emergency disposal scheme generated in the step 3.
Step 4 can show the business logic generated by the emergency disposal scheme in a visual mode, and has interpretability advantages compared with methods such as pure deep learning.
Further, the method also comprises a step 5 of supporting user-defined decision-making variable modification when a user thinks that the business logic needs to be adjusted, modifying the decision branch on the thinking diagram of the emergency disposal decision-making knowledge model, and taking effect immediately after modification; the decision variables comprise scene characteristics including accident type, accident position, disposal time limit and meteorological conditions, and disposal selection results including disposal method, disposal force and quantity, disposal equipment and quantity results.
Further, step 1 further comprises: and 1-3, recording corresponding scene characteristics and disposal selection result information by a historical emergency disposal scheme library every time emergency disposal operation is performed.
And 1-4, when the user needs to update the emergency disposal decision knowledge model, extracting historical data from the historical emergency disposal scheme library again, recording the historical data item by item according to scene characteristics and disposal selection results, and reorganizing to generate a training set.
The principle of the invention is as follows: the emergency treatment decision-making knowledge model is characterized in that emergency accident treatment experience knowledge of a user is integrated, operation experience in a historical emergency treatment scheme is mined in a machine learning mode, and an emergency treatment decision-making knowledge model comprising multidimensional characteristics such as accident types, accident positions, treatment methods, treatment force and treatment equipment is constructed. Meanwhile, the treatment scheme generation process and the back business logic can be interpreted and characterized in a thinking guide graph form, and the user with the learned decision knowledge can modify and take effect instantly. By using the method, the machine can obtain an emergency disposal scheme by second-level reasoning according to the accident type and the position condition.
Has the advantages that: compared with the prior art, the invention has the following remarkable advantages: the emergency disposal decision experience learning and the business logic custom modification adopted by the method are researched aiming at how the emergency disposal scheme for the emergency accident can be learned and generated, the decision knowledge can be continuously updated, and the defect that the traditional fixed rule type emergency disposal scheme for the emergency accident generation method has no learning and updating capability is overcome. The business logic interpretable representation mode adopted by the method can display the business logic generated by the emergency disposal scheme in a visual mode, automatically position and display the decision branch on which the scheme is based, and has interpretable advantages compared with methods such as pure deep learning.
Drawings
The foregoing and/or other advantages of the invention will become further apparent from the following detailed description of the invention when taken in conjunction with the accompanying drawings.
Fig. 1 is a schematic structural diagram of a GBDT model in an emergency incident handling method based on historical case and empirical knowledge learning according to an embodiment of the present disclosure.
Fig. 2 is a schematic diagram illustrating an integration of a decision-making knowledge model in an emergency incident handling method based on historical cases and empirical knowledge learning according to an embodiment of the present disclosure.
Fig. 3 is a schematic diagram of an emergency disposal decision knowledge model in an emergency incident disposal method based on historical case and empirical knowledge learning according to an embodiment of the present application.
Fig. 4 is a schematic diagram of an emergency disposal scheme in an emergency accident disposal method based on historical cases and empirical knowledge learning according to an embodiment of the present application.
Fig. 5 is a schematic diagram illustrating decision branches according to which an emergency treatment plan generated by automatic positioning display in an emergency accident treatment method based on historical cases and empirical knowledge learning according to an embodiment of the present application is based.
Fig. 6 is a schematic diagram illustrating custom modification of decision variables in an emergency incident handling method based on historical cases and empirical knowledge learning according to an embodiment of the present application.
Fig. 7 is a schematic view illustrating a visualized modification of a decision branch in an emergency incident handling method based on historical case and empirical knowledge learning according to an embodiment of the present application.
Fig. 8 is a schematic flowchart of an emergency incident handling method based on historical case and empirical knowledge learning according to an embodiment of the present application.
Detailed Description
Embodiments of the present invention will be described below with reference to the accompanying drawings.
The embodiment of the application discloses an emergency accident handling method based on historical case and empirical knowledge learning, and as shown in fig. 8, the emergency accident handling method comprises the following steps.
In the disposal experience data acquisition step, information such as accident types, accident positions, disposal methods, disposal forces and quantities, disposal equipment and quantities and the like is acquired and extracted from a historical emergency disposal scheme library, and a training data set for emergency accident disposal experience learning is constructed.
In the process of training data set preparation, a training set needs to be generated by collecting all reliable historical emergency treatment schemes. The content of the training data is shown in table 1, which includes scene features at the time of emergency treatment and treatment selection results including treatment methods, treatment forces and quantities, treatment devices, and quantity results corresponding to the scene features. Taking highway debris flow emergency rescue as an example, the scene characteristics mainly considered during emergency treatment comprise accident type, accident position, treatment time limit, meteorological conditions and the like, and the treatment selection result adopts an aerial emergency method, and an XX aerial rescue team and a rescue helicopter x 1 are used for going out.
TABLE 1 training data Structure for handling empirical learning
And recording scene characteristics such as corresponding accident types, accident positions, disposal time limits and meteorological conditions and disposal selection result information such as disposal methods, disposal forces and disposal equipment from the historical emergency disposal scheme library every time the emergency disposal operation is performed. When a user needs to update the emergency disposal decision knowledge model, historical data can be extracted from the historical emergency disposal scheme library, and are recorded item by item according to scene characteristics and disposal selection results, and a training set is reorganized and generated. For each training data, the scene characteristics such as accident type, accident position, disposal time limit, meteorological conditions and the like are equivalent to the input of the model, the combination of disposal selection results is equivalent to the label of the data, and the subsequent emergency disposal decision knowledge model training can be guided by comparing the output between the models and the difference of the label.
And 2, learning and constructing an emergency disposal decision knowledge model.
The emergency disposal decision-making knowledge model learning and constructing step is used for learning the combined selection of a disposal method, disposal force and disposal equipment of a user under different scene characteristics, learning and excavating emergency disposal decision-making experience by adopting a gradient lifting decision tree combined with a CART algorithm, and learning and constructing an emergency disposal decision-making knowledge model.
The GBDT is an integrated learning model taking a decision tree algorithm as a base classifier, and can achieve an algorithm for classifying or regressing data by continuously reducing a loss function generated in a training process. The emergency treatment decision belongs to a typical multi-classification problem (namely training sample data has a plurality of label classes), and the CART decision tree is used as a base classifier in the method. The CART decision tree can be used for classification and regression, has the characteristics of high training efficiency, good generalization performance and the like compared with decision tree algorithms such as ID3 and C4.5, and can well solve the emergency treatment decision problem related to the method.
The step 2 comprises the following steps:
step 2-1, reading a training data set of emergency accident disposal experience learning, and extracting disposal selection result information of a user under different emergency disposal conditions;
step 2-2, adopting scene characteristics such as accident type, accident position, disposal time limit, meteorological conditions and the like as gradient lifting decision tree model input, and selecting disposal selection nodes such as disposal method, disposal force, disposal equipment and the likeThe result is output as a gradient lifting decision tree model, and the number of label categories of a certain treatment selection result in the training data set is assumed to beK. For example, if the handling method is { air first aid, ground rescue } class 2, it corresponds toK=2。
Log-likelihood as a loss function:
wherein:yis a value of a category of the label,f(x) In order to predict the function for the model,p k (x) For training samplesxBelonging to the class of labelskThe probability of (a) of (b) being,d k = {0,1} represents whether or not to belong to the second placekClass label, 1 means yes, 0 means no.
Step 2-3, the GBDT separately trains a CART decision tree for each type of label, and each iteration generates a weak classifier through a multi-iteration modef km (x) Each classifier is trained on the basis of the loss function of the last classifier, the loss function generated in the training process is reduced, and the strong classifier of each class of labels is obtainedf k (x) Then, the prediction probability values of the classifiers are compared to determine the final output treatment decision result, as shown in fig. 1. The specific process is as follows:
step 2-3-1, initializing a classifier:
wherein:f k0 (x) Is as followskInitial classifier of class labels (i.e., number of iteration rounds)m=0），KIs the label category number.
Step 2-3-2, for given iteration round numberMCalculating samplesx i Corresponding categorykNegative gradient error ofr mik ：
Wherein:i=1,2,…,N，Nin order to train the number of samples,m=1,2,…,M，Mis an integer greater than 1 and is,y ik is a samplex i The corresponding tag class value is set to the tag class value,p k (x i ) Is a samplex i Belonging to each label categorykProbability of (c):
wherein:f km (x i ) Is as followsmWhen the wheel iterates, for the secondkAnd a classifier for classifying the class labels.
Step 2-3-3, calculating the optimal negative gradient fitting value of each leaf node in the classifierc mjk ：
Wherein:j=1,2,…,J，Jthe number of the leaf nodes is the number of the leaf nodes,R mjk is as followsmIn round iteration belong tokThe set of leaf nodes of the class label,R mk is as followsmIn round iteration belong tokAll sample sets of class labels.
Step 2-3-4, updating the classifier:
when the preset number of iteration rounds of the algorithm is reachedMIf so, terminating the iteration to obtain a strong classifierf k (x)=f kM (x)。
And 2-3-5, comparing the prediction probability values of all the classifiers and determining the final output treatment decision result.
With the continuous accumulation of emergency disposition decision knowledge,And in the improvement, part of historical cases formed by using outdated decision knowledge have negative influence on model training and are not suitable for being used as training bases. Therefore, the step is based on the gradient lifting decision tree combined with the CART algorithm and used for calculating samplesx i Corresponding categorykWhen the negative gradient error is detected, a forgetting function is introduced, namely the negative gradient error in the step 2-3-2 is calculated and adjusted as follows:
wherein:Tas a result of the current date,t i is a samplex i The modification/selection date of the representative treatment case. Is obviously whent i =TWhile, the samplex i The negative gradient error of (2) will all contribute to the fitting of the model when the sample is takenx i The date of modification/selection of (d) is currently longer,t i <<Twhile, the samplex i The effect of the negative gradient error is close to 1/2 of the original.
And 2-4, integrating decision branches obtained by the classifiers according to scene characteristics such as accident types, accident positions, handling time limits and meteorological conditions to obtain a gradient lifting decision tree model of the handling selection result.
As shown in fig. 2, the treatment selection result of the treatment method includes 2 kinds of labels for air first aid and ground rescue, 2 classifiers (decision trees) corresponding to the 2 kinds of labels are obtained by training, and the decision tree model integrated with the treatment selection result is obtained by merging the 2 decision tree branches.
And 2-5, repeatedly executing the steps 2-3 to 2-4 to obtain decision tree models of other treatment selection results.
And 2-6, respectively inputting all scene characteristics in the training data set into each trained gradient lifting decision tree model, and outputting a corresponding emergency disposal scheme, thereby constructing an emergency disposal decision knowledge model. An example of a treatment decision model including elements of treatment method, treatment power, treatment equipment, etc. is shown in fig. 3.
And 3, generating an emergency disposal scheme.
The emergency disposal scheme generating step is to generate an emergency disposal scheme suitable for the current situation according to the accident type (road debris flow), the accident position (certain mountain area), the disposal time limit (1.5 hours) and the meteorological condition (light rain) of the current emergency disposal task in an inference mode according to an emergency disposal decision knowledge model, and the emergency disposal scheme is executed as shown in fig. 4.
Step 4, the business logic can explain the representation, including:
step 4-1, displaying emergency disposal schemes under different scene characteristics in the emergency disposal decision knowledge model in the step 2 in a thinking graph form; this step is the prior art, and this embodiment is not specifically limited.
And 4-2, automatically positioning and displaying a decision branch according to the emergency disposal scheme generated in the step 3 on a thinking map of the emergency disposal decision knowledge model, so as to explain the business logic for representing the emergency disposal scheme generated in the step 3. The current accident type, accident location, disposal time limit, weather conditions, etc. scene features may be represented as a multivariate set { highway debris flow, M mountainous area, 1.5 hours, light rain }. When the user selects the emergency disposal scheme, the multi-component message is sent to the mind map display page of the emergency disposal decision knowledge model, and the decision branch according to which is automatically positioned and displayed is shown in fig. 5.
And step 5, when the user considers that the service logic needs to be adjusted, the user can customize the decision variables on the page, modify the decision branch, and take effect immediately after modification, as shown in fig. 6 and 7. The decision variables comprise scene characteristics including accident type, accident position, disposal time limit and meteorological conditions, and disposal selection results including disposal method, disposal force and quantity, disposal equipment and quantity results.
In a specific implementation, the present application provides a computer storage medium and a corresponding data processing unit, where the computer storage medium is capable of storing a computer program, and the computer program, when executed by the data processing unit, may execute the inventive content of the emergency incident handling method based on historical case and empirical knowledge learning provided by the present invention and some or all of the steps in each embodiment. The storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), a Random Access Memory (RAM), or the like.
It is clear to those skilled in the art that the technical solutions in the embodiments of the present invention can be implemented by means of a computer program and its corresponding general-purpose hardware platform. Based on such understanding, the technical solutions in the embodiments of the present invention may be embodied in the form of a computer program, that is, a software product, which may be stored in a storage medium and includes several instructions to enable a device (which may be a personal computer, a server, a single chip, an MUU, or a network device) including a data processing unit to execute the method in each embodiment or some parts of the embodiments of the present invention.
The invention provides an emergency accident disposal method based on historical case and empirical knowledge learning, and a plurality of methods and ways for implementing the technical scheme are provided. All the components not specified in the present embodiment can be realized by the prior art.
Claims (10)
1. An emergency accident handling method based on historical case and experience knowledge learning is characterized by comprising the following steps:
step 1, collecting disposal experience data, and constructing a training data set for emergency accident disposal experience learning, wherein the disposal experience data comprises scene characteristics and corresponding disposal selection result information;
step 2, learning to excavate emergency disposal decision experience according to disposal selection results of users under different scene characteristics, and learning to construct an emergency disposal decision knowledge model;
and 3, generating an emergency disposal scheme suitable for the current situation through the emergency disposal decision knowledge model reasoning according to the scene characteristics of the current emergency disposal task, and executing the emergency disposal scheme.
2. The emergency incident handling method based on historical case and empirical knowledge learning according to claim 1, wherein the step 1 comprises:
step 1-1, collecting all reliable historical emergency disposal schemes to form a historical emergency disposal scheme library;
step 1-2, extracting scene features of each emergency disposal scheme in a historical emergency disposal scheme library and disposal selection result information corresponding to the scene features to form a training data set for emergency accident disposal experience learning; the scene characteristics comprise accident types, accident positions, treatment time limits and meteorological conditions, and the treatment selection result information comprises treatment methods, treatment force and quantity, treatment equipment and quantity results.
3. The emergency incident treatment method based on historical case and experience knowledge learning of claim 2, wherein the gradient boosting decision tree model learning is adopted in step 2 to mine the emergency treatment decision experience.
4. The emergency incident handling method based on historical case and empirical knowledge learning of claim 3, wherein the gradient boosting decision tree in step 2 uses CART decision tree as a base classifier.
5. The emergency incident handling method based on historical case and empirical knowledge learning according to claim 4, wherein the step 2 comprises:
step 2-1, reading a training data set of emergency accident disposal experience learning, and extracting disposal selection result information of a user under different emergency disposal conditions;
step 2-2, inputting the scene characteristics as a gradient lifting decision tree model, and outputting disposal selection result information as the gradient lifting decision tree model; and setting the label category number of a certain treatment selection result in the training data set as K, wherein K is more than or equal to 2, and taking the log-likelihood as a loss function:
wherein: y is the label class value, f (x) is the gradient boosting decision tree model prediction function, p k (x) To train the probability that a sample x belongs to the label class k, δ k = {0,1} indicates whether it belongs to the kth class label, 1 indicates yes, 0 indicates no;
2-3, training a CART decision tree for each type of label, and obtaining a strong classifier f of each type of label through a multi-round iteration mode k (x) Then, the prediction probability values of all the classifiers are compared to determine the final output treatment decision result;
step 2-4, integrating decision branches obtained by K classifiers according to scene characteristics to obtain a gradient lifting decision tree model of the disposal selection result;
step 2-5, repeatedly executing the step 2-3 to the step 2-4, and obtaining gradient lifting decision tree models of other treatment selection results;
and 2-6, respectively inputting all scene characteristics in the training data set into each trained gradient lifting decision tree model, and outputting a corresponding emergency disposal scheme, thereby constructing an emergency disposal decision knowledge model.
6. The emergency incident handling method based on historical case and empirical knowledge learning according to claim 5, wherein the steps 2-3 comprise:
step 2-3-1, initializing a classifier:
wherein: f. of k0 (x) An initial classifier of the kth class label, wherein K is the label class number;
step 2-3-2, calculating a sample x in each iteration for a given iteration round number M i Negative gradient error r for class k mik I =1,2, \8230, N, N is the number of training samples, M =1,2, \8230, M, M is an integer greater than 1;
step 2-3-3, calculating the optimal negative gradient fitting value c of each leaf node in the classifier mjk ：
Wherein: j =1,2, \ 8230, J is the number of leaf nodes, R mjk For the set of leaf nodes belonging to the kth class label in the mth iteration, R mk All sample sets belonging to the kth class label in the mth iteration are collected;
step 2-3-4, updating the classifier:
when the iteration number M preset by the algorithm is reached, the iteration is terminated, and a strong classifier f is obtained k (x)=f kM (x)；
And 2-3-5, comparing the prediction probability values of all the classifiers and determining the final output treatment decision result.
7. The emergency incident handling method based on historical case and empirical knowledge learning of claim 6, wherein the sample x is obtained in each iteration of the steps 2-3-2 i Negative gradient error representation r for class k mik Comprises the following steps:
wherein: y is ik Is a sample x i Corresponding label category value, T is the current date, T i Is a sample x i Modification or selection date, p, of representative treatment cases k (x i ) Is a sample x i Probability of belonging to each label category k:
wherein: f. of km (x i ) And the classifier is used for classifying the kth class label in the mth iteration.
8. The emergency incident treatment method based on historical case and empirical knowledge learning of claim 7, further comprising step 4, wherein the emergency treatment decision knowledge model interpretable characterization comprises:
step 4-1, displaying emergency disposal schemes under different scene characteristics in the emergency disposal decision knowledge model in the step 2 in a thinking graph form;
and 4-2, automatically positioning and displaying a decision branch according to the emergency disposal scheme generated in the step 3 on a thinking map of the emergency disposal decision knowledge model, so as to explain the business logic for representing the emergency disposal scheme generated in the step 3.
9. The emergency incident handling method based on historical case and empirical knowledge learning according to claim 8, further comprising step 5, when the user considers that the business logic needs to be adjusted, supporting custom modification of decision variables, and modifying decision branches on the mind map of the emergency handling decision knowledge model, wherein the modified decision variables take effect immediately; the decision variables comprise scene characteristics including accident type, accident position, disposal time limit and meteorological conditions, and disposal selection results including disposal method, disposal force and quantity, disposal equipment and quantity results.
10. The emergency incident handling method based on historical case and empirical knowledge learning according to claim 9, wherein the step 1 further comprises:
step 1-3, recording corresponding scene characteristics and disposal selection result information by a historical emergency disposal scheme library every time emergency disposal operation is carried out;
and 1-4, when the user needs to update the emergency disposal decision-making knowledge model, extracting the historical data from the historical emergency disposal scheme library again, recording the historical data item by item according to the scene characteristics and disposal selection results, and reorganizing to generate a training set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310144018.8A CN115829061B (en) | 2023-02-21 | 2023-02-21 | Emergency accident handling method based on historical case and experience knowledge learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310144018.8A CN115829061B (en) | 2023-02-21 | 2023-02-21 | Emergency accident handling method based on historical case and experience knowledge learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115829061A true CN115829061A (en) | 2023-03-21 |
| CN115829061B CN115829061B (en) | 2023-04-28 |
Family
ID=85522006
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310144018.8A Active CN115829061B (en) | 2023-02-21 | 2023-02-21 | Emergency accident handling method based on historical case and experience knowledge learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115829061B (en) |
Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103870879A (en) * | 2014-04-09 | 2014-06-18 | 中国电子科技集团公司第二十八研究所 | Method for selecting knowledge of computer-assisted decision making system based on undistinguishable relation |
| CN107977670A (en) * | 2017-10-09 | 2018-05-01 | 中国电子科技集团公司第二十八研究所 | Accident classification stage division, the apparatus and system of decision tree and bayesian algorithm |
| CN109726766A (en) * | 2019-01-04 | 2019-05-07 | 三峡大学 | A kind of online dynamic secure estimation method of electric system based on Integrated Decision tree |
| CN109840690A (en) * | 2018-12-29 | 2019-06-04 | 国网山东省电力公司日照供电公司 | A kind of artificial intelligence electric power first-aid system and method based on big data |
| CN109858647A (en) * | 2018-12-21 | 2019-06-07 | 河海大学 | A kind of Regional Flooding Hazard Risk Assessment and predictor method coupling GIS and GBDT algorithm |
| CN110119891A (en) * | 2019-04-28 | 2019-08-13 | 华南理工大学 | A kind of traffic safety influence factor discrimination method suitable for big data |
| CN110188979A (en) * | 2019-04-15 | 2019-08-30 | 广州智联研究院有限公司 | Water industry Emergency decision generation method and device |
| CN110675626A (en) * | 2019-09-27 | 2020-01-10 | 汉纳森（厦门）数据股份有限公司 | Traffic accident black point prediction method, device and medium based on multidimensional data |
| CN111324657A (en) * | 2020-02-12 | 2020-06-23 | 广州奥格智能科技有限公司 | Emergency plan content optimization method and computer equipment |
| CN111784084A (en) * | 2020-08-17 | 2020-10-16 | 北京市城市规划设计研究院 | Travel generation prediction method, system and device based on gradient lifting decision tree |
| CN113326432A (en) * | 2021-06-04 | 2021-08-31 | 温州大学 | Model optimization method based on decision tree and recommendation method |
| CN115409100A (en) * | 2022-08-23 | 2022-11-29 | 国能大渡河大岗山发电有限公司 | Method, device, medium and equipment for generating power accident emergency scheme |
| CN115587295A (en) * | 2021-06-22 | 2023-01-10 | 国网上海市电力公司 | Intelligent generation method and system of supply chain strategy based on machine self-learning |
| CN115659011A (en) * | 2022-10-14 | 2023-01-31 | 上海东方明珠数字电视有限公司 | Decision case module recommendation implementation method based on autonomous recommendation mechanism |
-
2023
- 2023-02-21 CN CN202310144018.8A patent/CN115829061B/en active Active
Patent Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103870879A (en) * | 2014-04-09 | 2014-06-18 | 中国电子科技集团公司第二十八研究所 | Method for selecting knowledge of computer-assisted decision making system based on undistinguishable relation |
| CN107977670A (en) * | 2017-10-09 | 2018-05-01 | 中国电子科技集团公司第二十八研究所 | Accident classification stage division, the apparatus and system of decision tree and bayesian algorithm |
| CN109858647A (en) * | 2018-12-21 | 2019-06-07 | 河海大学 | A kind of Regional Flooding Hazard Risk Assessment and predictor method coupling GIS and GBDT algorithm |
| CN109840690A (en) * | 2018-12-29 | 2019-06-04 | 国网山东省电力公司日照供电公司 | A kind of artificial intelligence electric power first-aid system and method based on big data |
| CN109726766A (en) * | 2019-01-04 | 2019-05-07 | 三峡大学 | A kind of online dynamic secure estimation method of electric system based on Integrated Decision tree |
| CN110188979A (en) * | 2019-04-15 | 2019-08-30 | 广州智联研究院有限公司 | Water industry Emergency decision generation method and device |
| CN110119891A (en) * | 2019-04-28 | 2019-08-13 | 华南理工大学 | A kind of traffic safety influence factor discrimination method suitable for big data |
| CN110675626A (en) * | 2019-09-27 | 2020-01-10 | 汉纳森（厦门）数据股份有限公司 | Traffic accident black point prediction method, device and medium based on multidimensional data |
| CN111324657A (en) * | 2020-02-12 | 2020-06-23 | 广州奥格智能科技有限公司 | Emergency plan content optimization method and computer equipment |
| CN111784084A (en) * | 2020-08-17 | 2020-10-16 | 北京市城市规划设计研究院 | Travel generation prediction method, system and device based on gradient lifting decision tree |
| CN113326432A (en) * | 2021-06-04 | 2021-08-31 | 温州大学 | Model optimization method based on decision tree and recommendation method |
| CN115587295A (en) * | 2021-06-22 | 2023-01-10 | 国网上海市电力公司 | Intelligent generation method and system of supply chain strategy based on machine self-learning |
| CN115409100A (en) * | 2022-08-23 | 2022-11-29 | 国能大渡河大岗山发电有限公司 | Method, device, medium and equipment for generating power accident emergency scheme |
| CN115659011A (en) * | 2022-10-14 | 2023-01-31 | 上海东方明珠数字电视有限公司 | Decision case module recommendation implementation method based on autonomous recommendation mechanism |
Non-Patent Citations (1)
| Title |
|---|
| ZHANG W等: "Predictive model of cooling load for ice storage air-conditioning system by using GBDT" * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115829061B (en) | 2023-04-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Geetha et al. | Data mining for meteorological applications: Decision trees for modeling rainfall prediction | |
| CN109829236A (en) | A kind of Compressor Fault Diagnosis method based on XGBoost feature extraction | |
| CN113096388B (en) | Short-term traffic flow prediction method based on gradient lifting decision tree | |
| CN111754345B (en) | Bit currency address classification method based on improved random forest | |
| CN111950585A (en) | XGboost-based underground comprehensive pipe gallery safety condition assessment method | |
| CN106897821A (en) | A kind of transient state assesses feature selection approach and device | |
| CN111008337B (en) | Deep attention rumor identification method and device based on ternary characteristics | |
| CN110781297B (en) | Classification method of multi-label scientific research papers based on hierarchical discriminant trees | |
| CN107977670A (en) | Accident classification stage division, the apparatus and system of decision tree and bayesian algorithm | |
| CN103886030B (en) | Cost-sensitive decision-making tree based physical information fusion system data classification method | |
| CN103324954A (en) | Image classification method based on tree structure and system using same | |
| Nhita | A rainfall forecasting using fuzzy system based on genetic algorithm | |
| CN111967675A (en) | Photovoltaic power generation amount prediction method and prediction device | |
| CN106600046A (en) | Multi-classifier fusion-based land unused condition prediction method and device | |
| CN110942098A (en) | Power supply service quality analysis method based on Bayesian pruning decision tree | |
| CN105808706A (en) | Space object identification method based on application ontology | |
| CN105808582A (en) | Parallel generation method and device of decision tree on the basis of layered strategy | |
| CN113128769A (en) | Intelligent flight delay prediction method based on deep learning | |
| CN110287995B (en) | Multi-feature learning network model method for grading all-day overhead traffic jam conditions | |
| Shabani et al. | Modeling the susceptibility of an uneven-aged broad-leaved forest to snowstorm damage using spatially explicit machine learning | |
| CN112380945B (en) | Method and system for identifying potential hazards of power transmission channel environment based on satellite remote sensing image | |
| CN112101313B (en) | Machine room robot inspection method and system | |
| CN117995027A (en) | Flight plan drop-preparation recommendation method, system, equipment and medium | |
| CN115829061A (en) | Emergency accident disposal method based on historical case and empirical knowledge learning | |
| CN111985782A (en) | Automatic tramcar driving risk assessment method based on environment perception |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |