CN111324657A - Emergency plan content optimization method and computer equipment - Google Patents
Emergency plan content optimization method and computer equipment Download PDFInfo
- Publication number
- CN111324657A CN111324657A CN202010088012.XA CN202010088012A CN111324657A CN 111324657 A CN111324657 A CN 111324657A CN 202010088012 A CN202010088012 A CN 202010088012A CN 111324657 A CN111324657 A CN 111324657A
- Authority
- CN
- China
- Prior art keywords
- emergency
- attribute
- emergency plan
- data
- item set
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/26—Visual data mining; Browsing structured data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the field of safety emergency management, in particular to an emergency plan content optimization method and computer equipment. The method comprises the following steps: collecting data and establishing a historical safety emergency plan database; processing and sorting the acquired data to form a pattern suitable for data analysis; and (3) data mining and analyzing, namely, setting a proper algorithm model, putting training set data into the algorithm model for training, and continuously optimizing the algorithm model to finally obtain a calculation result value with the minimum error with an actual result so as to realize the optimization of the emergency plan content.
Description
Technical Field
The invention relates to the field of safety emergency management, in particular to an emergency plan content optimization method and computer equipment based on historical case big data mining.
Background
The safety emergency plan is a scientific and effective plan and arrangement which is made in advance for specific equipment, facilities, places and environments, on the basis of safety evaluation, in order to reduce the loss of personnel, property and environment caused by accidents, on the basis of emergency rescue mechanisms and personnel after the accidents occur, emergency rescue equipment, facilities, conditions and environments, action steps and outlines, accident development control methods and programs and the like. The safety emergency plans comprise a comprehensive emergency plan for comprehensive treatment, a special emergency plan for special treatment, a field disposal plan for field treatment, a department emergency plan for disposal of a certain department, a unit emergency plan for emergency disposal of each unit and the like. The following four categories can be classified according to event types: natural disasters, accident disasters, public health events, social security events.
Along with the continuous accumulation of the historical data of the safety emergency plan, the application of the big data analysis technology to the optimization of the safety emergency plan becomes a brand-new direction, and the whole process from the prior to the subsequent of the safety emergency plan is optimized by establishing different algorithm models, wherein the process comprises the steps of making the content of the safety emergency plan, setting certain parameters in the plan, scheduling during event disposal, optimizing an evacuation disposal mode and the like. The safety emergency plan content optimization based on big data analysis provides reliable technical support for the establishment of a safety emergency plan system.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides an emergency plan content optimization method, which optimizes the content of an emergency plan on the basis of historical emergency case data by utilizing technologies such as big data analysis, data mining and the like, so that the content attribute setting of the safety emergency plan can better accord with the actual case condition, and the aim of intelligently setting the attribute parameters of the safety emergency plan is fulfilled.
The invention also proposes a computer device which, when executing an executable program, carries out the steps of the optimization method of the invention.
The optimization method is realized by adopting the following technical scheme: the emergency plan content optimization method comprises the following steps:
s1, collecting data, and establishing a historical safety emergency plan database;
s2, processing and sorting the collected data to form a pattern suitable for data analysis;
and S3, data mining and analyzing, putting training set data into the algorithm model for training by establishing a proper algorithm model, and finally obtaining a calculation result value with the minimum error with an actual result by continuously optimizing the algorithm model, so as to realize the optimization of the emergency plan content.
In a preferred embodiment, step S3 includes optimizing attribute parameters of the safety emergency plan based on a random forest algorithm:
analyzing historical emergency command case data, preprocessing the historical emergency command case data, extracting attribute parameter influence characteristics of a safety emergency plan, and dividing the attribute parameters into category attribute parameters for qualitatively analyzing the emergency plan and numerical attribute parameters for quantitatively determining the disposal mode of the emergency plan;
and selecting a random forest algorithm to establish a regression model and a classification model, predicting the attribute parameter values, continuously optimizing the regression model and the classification model according to the error between the prediction result of the attribute parameter values and the actual case parameter values, and finally enabling the category attribute parameter values and the numerical attribute parameter values predicted according to the regression model and the classification model to be closer to the actual case parameter values so as to realize the setting of the attribute parameters of the intelligent safety emergency plan.
In a preferred embodiment, step S3 is to repeatedly and randomly extract k samples from the original training sample set N to generate a new training sample set by using a random forest algorithm, and then generate k decision trees from the new training sample set to form a random forest, where the prediction result of the sample to be tested is determined by an average of the prediction values of the k decision trees.
The step S3 is based on the process of predicting the attribute parameter values by the random forest algorithm, and comprises the following steps:
s31, acquiring emergency case data in a historical safety emergency plan database, and constructing an initial sample data set;
s32, processing the data of the safety emergency plan;
s33, processing the characteristics of the emergency data, screening the characteristics of which the correlation is greater than a set threshold value, and combining the characteristics of which the correlation is strong;
s34, extracting the characteristics of the emergency data, randomly selecting j characteristics from the i characteristics for each training set, wherein j < < i, and constructing a single decision tree by using the randomly selected j characteristics;
s35, traversing each feature in each training set and the feature value corresponding to the feature value, selecting the optimal feature to segment the training sets, and segmenting each training set into a data space R1 and a data space R2;
s36, recursively calling the step S35 for the two segmented data spaces R1 and R2 until no optimal features can be used for segmenting the training set or the deepest depth of the tree is reached, and generating a single decision tree;
s37, repeating the steps S35 and S36 by using n training sets respectively to construct a single decision tree, and finally constructing a random forest consisting of n decision trees;
and S38, predicting the attribute parameter values by using the random forest.
In a preferred embodiment, step S3 includes optimizing the attribute structure of the safety emergency protocol based on a correlation algorithm:
analyzing the attribute structure of the safety emergency plan, establishing an association item set table of the contents and the attributes of the safety emergency plan, finding a frequent item set of the contents and the attributes of the safety emergency plan, and finding a strong association rule of the contents and the attributes of the safety emergency plan according to the set support degree and confidence; and adding the corresponding attributes into the safety emergency plan model by combining the found strong association rules, and continuously optimizing the attribute structure of the safety emergency plan model, so that the setting of the content attributes of the safety emergency plan can better accord with the actual case situation.
Preferably, the step S3 of optimizing the attribute structure of the safety emergency plan based on the association algorithm includes:
s31', acquiring a data set, analyzing the association model of the emergency resource scheduling class and the corresponding attribute in the safety emergency plan, and discovering the attribute of the emergency resource scheduling class; or the safety emergency plan is selected as a class, and the corresponding attribute is found.
S32', establishing a frequent item set, firstly setting a support degree and a confidence degree threshold value, wherein the support degree represents the times of the emergency resource scheduling class and a certain attribute appearing in a safety emergency plan at the same time, and the confidence degree represents the times of the emergency resource scheduling class and a certain attribute appearing in a safety emergency plan/the number of historical emergency plans;
s33', according to the confidence coefficient, extracting all strong association rules in the frequent item set established in the step S32' to find the strong association relation between the class and the attribute, thereby optimizing the attribute structure of a certain class of the safety emergency plan.
Wherein, the process of establishing the frequent item set in step S32' is as follows:
establishing an initial plan item set table, traversing all cases in the initial plan item set table, and counting each attribute to obtain a candidate attribute item set table; deleting the item set with the support degree smaller than a preset support degree threshold value in the candidate attribute item set table, reserving the item set with the support degree larger than or equal to the preset support degree threshold value to obtain a first frequent item set, and obtaining a second candidate attribute item set table by utilizing self-connection of the first frequent item set; deleting all the item sets with the support degrees smaller than a preset support degree threshold value in the second candidate attribute item set table to obtain a second frequent item set table; traversing the second frequent item set table, and self-connecting the second frequent item set table to obtain a third candidate attribute item set table; deleting the item set with the support degree smaller than a preset support degree threshold value in the third candidate attribute item set table to obtain a third complex item set table;
and repeating in a circulating mode, acquiring a new candidate attribute item set table by traversing the previous frequent item set table and connecting the previous frequent item set table, and deleting an item set with the support degree smaller than a preset support degree threshold value in the new candidate attribute item set table to obtain a new frequent item set table until the candidate attribute item set is empty, thereby obtaining all frequent item sets.
The computer equipment comprises a memory and a processor, wherein the memory is provided with an executable program, and the processor executes the executable program to realize the steps of the optimization method.
Compared with the prior art, the invention has the following beneficial effects: the method extracts general information of the emergency plan from the existing emergency plan case, wherein the general information comprises basic information of emergency plan category of emergency events, emergency plan level of emergency events, emergency plan organization mechanism of emergency events, emergency plan resource allocation scheme, emergency plan disposal process and the like, so that a historical safety emergency plan database is established; the emergency plan content optimization method based on big data analysis is provided by combining a large amount of historical emergency plan data, and comprises the safety emergency plan attribute parameter optimization based on a random forest algorithm and the safety emergency plan attribute structure optimization based on an associated algorithm, so that the safety emergency plan content attribute setting can better accord with the actual case situation, and the goal of intelligent safety emergency plan attribute parameter setting is realized.
Drawings
FIG. 1 is a flow chart of an emergency plan content optimization method of the present invention;
FIG. 2 is an illustration of a historical safety emergency plan database;
FIG. 3 is an exemplary graph of impact range prediction based on random forests.
Detailed Description
The following describes in detail an implementation of the present invention in conjunction with examples and drawings, but the present invention is not limited to these.
Examples
The embodiment provides an emergency plan content optimization method, which is used for mining and analyzing safety emergency plan data by using a big data analysis technology and a data mining technology, and mainly optimizes the content and partial attribute parameters of a safety emergency plan aiming at historical safety emergency plan data, wherein the safety emergency plan content optimization mainly comprises two parts, namely safety emergency plan attribute parameter optimization and safety emergency plan attribute structure optimization, and as shown in fig. 1, the method specifically comprises the following steps:
step 1, collecting data, and establishing a historical safety emergency plan database, namely a safety emergency plan historical database and a historical emergency command case database.
In the embodiment, a text form emergency plan is analyzed, and a content structure and attribute parameters of the text form emergency plan are extracted to form a primary historical safety emergency plan database; when the actual emergency type is the same as the safety emergency plan type, carrying out automatic emergency intelligent command according to the emergency plan designed by the safety emergency plan model, and simultaneously adding the emergency content into a historical safety emergency plan database; if the actual emergency type is different from the safety emergency plan type, adding the contents of each part of the actual emergency into a historical safety emergency plan database so as to complete the establishment of the historical safety emergency plan database in the step. The design use graph of the historical safety emergency plan database is shown in figure 2.
And 2, processing data. The data processing is to process and arrange the acquired data to form a pattern suitable for data analysis, so that the consistency and the effectiveness of the data are ensured. Such as conversion of data types in safety emergency plans, integration of different data, and processing of data loss and errors.
After data processing, data mining analysis is carried out, and a proper algorithm model is mainly established, and training set data are put into the algorithm model for training to obtain a calculation result. Aiming at different analysis purposes, such as classification, prediction and the like, different algorithm models are established, and a calculation result value with the minimum error with an actual result is finally obtained by continuously optimizing the algorithm models, so that the optimization of the emergency plan content is realized. The invention mainly optimizes the attribute parameters and the attribute results of the safety emergency plan, and the detailed description is as follows:
and 3, optimizing the attribute parameters of the safety emergency plan based on the random forest algorithm.
The method comprises the steps of analyzing historical emergency command case data, preprocessing the historical emergency command case data, extracting attribute parameter influence characteristics of a safety emergency plan, and dividing the attribute parameters into category attribute parameters (namely, classification attribute parameters) and numerical attribute parameters, wherein the classification attribute parameters are used for qualitatively analyzing the emergency plan, such as an affected object of an emergency event, an emergency scheduling resource category and the like, and the numerical attribute parameters are used for quantitatively determining a disposal mode of the emergency plan. The attribute values of the attribute parameters are usually set according to manual experience, such as the quantity setting, the type setting, the setting of an emergency influence object, the setting of an influence range and the like of certain types of emergency resources, and the setting is subjective depending on human, so that the invention analyzes the content data of the historical safety emergency plan and provides a safety emergency plan attribute parameter optimization method based on a random forest algorithm.
The method selects a random forest algorithm to establish a regression model and a classification model, predicts the attribute parameter values, continuously optimizes the regression model and the classification model according to the error between the prediction result of the attribute parameter values and the actual case parameter values, and finally enables the category attribute parameter values and the numerical attribute parameter values predicted according to the regression model and the classification model to be closer to the actual case parameter values, so that the aim of setting the attribute parameters of the intelligent safety emergency plan is fulfilled.
The principle of the random forest algorithm is that k samples are repeatedly and randomly extracted from an original training sample set N in a replacement mode to generate a new training sample set, then k decision trees are generated according to the new training sample set to form a random forest, and the prediction result of a sample to be tested is determined through the average value of the predicted values of the k decision trees. As shown in equation (1):
the left side of the equation in equation (1) represents the final predicted value, hi(x) The predicted values for each decision tree are represented. Similarly, when a random forest is used for predicting the classification attribute parameters, the classification of the classification attribute parameters is determined based on a voting principle of a plurality of decision trees.
The invention takes a sudden flood disaster as an example, and explains the specific steps of the invention aiming at the prediction of two different attribute parameters of an affected object and an affected range. The influence range is subjected to regression prediction based on random forests, and the influenced objects are subjected to prediction based on random forest classification. The influence range prediction method based on the random forest algorithm comprises the following steps:
and 311, constructing a sample data set. Acquiring emergency case data such as flood disasters and the like in a historical safety emergency plan database, constructing an initial sample data set, setting an influence range as y, setting other description information of the emergency as X, setting X as a vector and setting dimensionality as a characteristic number, and constructing the initial sample data set in the shape shown in table 1:
TABLE 1 initial sample data set
| Case m | Event rank x1 | Time x2 | Affected object x3 | Wind direction x3 | Wind speed x4 | Weather x5 | .....xn | Extent of influence y |
| Case 1 | Level 1 | 17:20 | Construction of buildings | Northern wind | 4km/s | All-weather | 1000m | |
| …… | …… | …… | …… | …… | …… | …… | …… | …… |
| Case m | Grade 3 | 9:20 | Personnel | Northwest wind | 5km/s | Yin (kidney) | …… | 3000m |
And (3) randomly putting back samples from the initial sample data set by using a Bootstrap (random resampling) method to take m samples, carrying out sampling for n times in total to generate n training sets with the same size as the initial sample data set, and using the residual data as test data to form a test set for algorithm evaluation.
And step 312, processing emergency plan data. Normalization and standardization processing are carried out on the emergency plan data, so that the data cannot be influenced by magnitude. For example, the event occurrence time is labeled by using a number, such as the morning, afternoon and evening, and then converted into a digital label. Since the characteristics of each case are different, there may be a case where the characteristic value is missing, and therefore interpolation processing needs to be performed on the missing value.
Step 313, processing the characteristics of the emergency data, screening the characteristics with the correlation larger than a set threshold value, and combining the characteristics with strong correlation. As shown in table 1, the correlations between the n eigenvalues and the influence range of the emergency are different, so that the relationship between the independent variable and the single dependent variable needs to be analyzed, a correlation coefficient matrix is established, and features with correlations greater than 0.7 are screened. Meanwhile, there may be strong correlation between features, so it is necessary to merge features with strong correlation. The features are processed to obtain a data set as shown in table 2:
TABLE 2 data set after screening for features
| Case m | Event rank x1 | Time x2 | Affected object x3 | Wind direction x3 | Wind speed x4 | Weather x5 | .....xi | Extent of influence y |
| Case 1 | Level 1 | 17:20 | Construction of buildings | Northern wind | 4km/s | All-weather | 1000m | |
| …… | …… | …… | …… | …… | …… | …… | …… | …… |
| Case m | Grade 3 | 9:20 | Personnel | Northwest wind | 5km/s | Yin (kidney) | …… | 3000m |
I < < n in Table 2.
And step 314, extracting the characteristics of the emergency data so as to construct a single decision tree. After the features are screened, j features are randomly selected from the i features for each training set, wherein j < < i, and a single decision tree is constructed by using the randomly selected j features.
Step 315, traversing each feature in each training set and the feature value corresponding to the feature, selecting the optimal feature to segment the training sets, and segmenting each training set into a data space R1 and a data space R2; the principle of optimal feature selection is shown in equation (2):
wherein j represents the j-th feature, s represents the s-th feature value of the j-th feature, x represents the sample, y represents the actual value of the influence range, c1 represents the predicted value of the influence range under the data space R1, and similarly, c2 represents the predicted value of the influence range under the data space R2. Namely, the influence range prediction error can be minimized after the training set is divided according to the s-th eigenvalue of the j-th characteristic. The training set is divided into two parts, for example, wind speed is selected as the optimal characteristic in table 2, 5km/s is selected as the optimal characteristic value, the training set can be segmented into two training sets with wind speed >5km./s and wind speed <5km/s, and meanwhile, for each segmented training set, the predicted value can be obtained according to formula (3).
Wherein R ismRepresenting the divided training set, yiShows the influence range of each caseThe left side of the equation represents the predicted impact range, i.e., the predicted value of the impact range, and the right side of the equation represents the mean of the impact ranges of all cases in the training set.
Step 316, recursively invoking step 315 on the two segmented data spaces R1 and R2, stopping until no optimal features are available for segmenting the training set or the deepest depth of the tree is reached, thereby generating a single decision tree.
And 317, repeating the steps 315 and 316 by using n training sets respectively to construct a single decision tree, and finally constructing a random forest consisting of n decision trees. The random forest construction process is shown in fig. 3.
And 318, predicting attribute parameter values by using the random forest.
Randomly extracting a flood disaster case in the test set as a case to be predicted, firstly putting the flood disaster case into a decision tree, segmenting the data set according to a first optimal characteristic (the wind speed is more than 5km/s), and if the sample is classified into a leaf node data set after segmentation, taking the mean value of the influence range of the data set as the predicted value of the influence range of the case to be predicted; if the sample is not the leaf node data set, continuing segmentation until the sample is included in the leaf node data set, and taking the mean value of the influence range of the final data set as a predicted value of the influence range of the case to be predicted; and putting the case to be predicted into the other decision trees, repeating the steps, and solving the mean value of the n predicted values to obtain the predicted value of the final influence range.
The above process is only an optimization method of one numerical attribute parameter in the safety emergency plan, and other numerical attribute parameters, such as the quantity of emergency scheduling resources, the evacuation range and the like, can be predicted by using the algorithm to give a predicted value.
The method for predicting the affected objects based on the random forest algorithm comprises the following steps:
step 321, constructing a sample data set. Acquiring flood disaster case data in a historical safety emergency plan database, constructing an initial sample data set, setting an affected object as y, setting other description information of an emergency as X, setting X as a vector and dimensionality as a characteristic number, and constructing the initial sample data set in the form shown in Table 3:
TABLE 3 initial sample data set
| Case m | Event rank x1 | Time x2 | Event location x3 | Water flow rate x4 | Weather x5 | .....xn | Affected object y |
| Case 1 | Level 1 | 17:20 | Country | 4km/s | All-weather | Construction of buildings | |
| …… | …… | …… | …… | …… | …… | …… | …… |
| Case m | Grade 3 | 9:20 | Suburb | 5km/s | Yin (kidney) | …… | Personnel, livestock |
And (3) randomly putting back samples from the initial sample data set by using a Bootstrap (random resampling) method to take m samples, carrying out sampling for n times in total to generate n training sets with the same size as the initial sample data set, and using the residual data as test data to form a test set for algorithm evaluation.
And 322, processing emergency plan data. Normalization and standardization processing are carried out on the emergency plan data, so that the data cannot be influenced by magnitude. For example, the event occurrence time is labeled by using a number, such as the morning, afternoon and evening, and then converted into a digital label. Since the characteristics of each case are different, there may be a case where the characteristic value is missing, and therefore interpolation processing needs to be performed on the missing value.
Step 323, processing the characteristics of the emergency data, screening the characteristics with the correlation greater than a set threshold value, and combining the characteristics with strong correlation. As shown in table 3, the correlations between the n eigenvalues and the affected objects of the emergency are different, so that the relationship between the independent variable and the single dependent variable needs to be analyzed, a correlation coefficient matrix is established, and features with correlations greater than 0.7 are screened. Meanwhile, there may be strong correlation between features, so it is necessary to merge features with strong correlation. The features are processed to obtain a data set as shown in table 4:
TABLE 4 data set after screening for features
| Case m | Event rank x1 | Time x2 | Location x3 | Rainfall x4 | Weather x5 | .....xi | Affected object y |
| Case 1 | Level 1 | 17:20 | Country | 5mm | Heavy rain | Construction of buildings | |
| …… | …… | …… | …… | …… | …… | …… | …… |
| Case m | Grade 3 | 9:20 | Suburb | 10mm | Storm rain | …… | Personnel, livestock |
I < < n in Table 4.
And 324, extracting the characteristics of the emergency data so as to construct a single decision tree. After the features are screened, j features are randomly selected from i features for each training set, wherein j < < i, and a single decision tree is constructed by using the j features.
Step 325, traversing each feature in each training set and the feature value corresponding to the feature value, selecting the optimal feature to segment the training sets, and segmenting each training set into a data space R1 and a data space R2; the principle of optimal feature selection is shown in equation (4):
the formula (4) is a damping coefficient of the selected characteristic, the damping coefficient represents the impure degree of the model, and the smaller the damping coefficient is, the lower the impure degree is, and the better the characteristic is. And K represents the category of the affected object, namely the prediction error of the affected range can be minimized after the training set is segmented according to the K-th optimal characteristic. The training set is divided into two parts, for example, the rainfall is selected as the optimal characteristic in table 4, and 5mm is selected as the optimal characteristic value, so that the training set can be divided into two subsets, namely, the rainfall >5mm and the rainfall <5 mm.
Step 326, recursively invoking step 325 on the two segmented data spaces R1 and R2 until no optimal features are available for segmenting the training set or the deepest depth of the tree is reached, thereby generating a single decision tree.
And 327, repeating the step 325 and the step 326 by using n training sets respectively to construct a single decision tree, and finally constructing a random forest consisting of n decision trees.
And 328, predicting attribute parameter values by using the random forest.
Randomly extracting a flood disaster case in the test set as a case to be predicted, firstly putting the flood disaster case into a decision tree, segmenting the data set according to a first optimal characteristic (rainfall is greater than 5mm), and if the sample is classified into a leaf node data set after segmentation, taking the maximum sample number probability in the data set as a predicted value of the affected range of the case to be predicted; if the sample is not a leaf node dataset, continuing to cut until the sample is included in the leaf node dataset; and putting the cases to be predicted into the other decision trees, repeating the steps to obtain the predicted values of the n affected objects, and taking the category with the most sample prediction as the final predicted value of the affected objects according to the voting principle.
Similarly, the above process is only an optimization method of one category attribute parameter in the safety emergency plan, and other category attribute parameters can be predicted by using the algorithm to give a predicted value.
And 4, optimizing the attribute structure of the safety emergency plan based on the association algorithm.
Analyzing the attribute structure of the safety emergency plan, establishing an association item set table of the contents and the attributes of the safety emergency plan, finding a frequent item set of the contents and the attributes of the safety emergency plan, and finding a strong association rule of the contents and the attributes of the safety emergency plan according to the set support degree and confidence; and adding the corresponding attributes into the safety emergency plan model by combining the found strong association rules, and continuously optimizing the attribute structure of the safety emergency plan model, so that the setting of the content attributes of the safety emergency plan can better accord with the actual case situation.
The safety emergency plan model can be established in advance. In the process of establishing a safety emergency plan model, the attributes of each part of the emergency plan are manually increased or reduced, a large amount of historical safety emergency plan data are analyzed, an emergency plan class-attribute association model is established by combining an association algorithm, and the attribute structure of the safety emergency plan is optimized, and the method specifically comprises the following steps:
and step 41, acquiring a data set. The invention analyzes the correlation model of the emergency resource scheduling class and the corresponding attribute in the safety emergency plan and finds the attribute of the emergency resource scheduling class. Similarly, the safety emergency plan can be selected as a class, and the corresponding attribute can be found.
And step 42, establishing a frequent item set. Firstly, setting a support degree and a confidence degree threshold value, wherein the support degree represents the times of the emergency resource scheduling class and a certain attribute (such as the maximum scheduling radius) appearing in an emergency plan at the same time, and the confidence degree represents the times of the emergency resource scheduling class and a certain attribute (such as the maximum scheduling radius) appearing in an emergency plan/the number of historical emergency plans. An initial plan entry set table is then established, as shown in table 5:
TABLE 5 initial plan entry set Table
In the table, L1 represents the emergency resource scheduling scheme class, L2 represents the emergency resource category, L3 represents the emergency resource name, L4 represents the number of emergency resources, and L5 represents the maximum scheduling radius. All cases in table 3 are traversed and each attribute is counted, resulting in a first candidate attribute item set table C1, as shown in table 6:
table 6 candidate table C1
| Item set | Degree of support |
| {L1} | 10 |
| {L2} | 5 |
| {L3} | 4 |
| {L4} | 4 |
| {L5} | 5 |
The support degree in the table indicates the number of times the set of items occurs in all cases. Assuming that the preset support threshold is 2, the item set greater than or equal to 2 is reserved, the item set smaller than 2 is deleted, and all the item sets in C1 are reserved because the support of all the item sets is greater than 2. At this time, C1 was designated as the first frequent item set Lable 1. After getting Lable1, self-join using the first frequent item set Lable1 gets a second candidate property item set Table C2, as shown in Table 7:
table 7 candidate table C2
Deleting all the item sets with the support degrees smaller than the preset support degree threshold 2 in the second candidate attribute item set table to obtain a second frequent item set table Lable2, as shown in Table 8:
TABLE 8 frequent item set Table Lable2
| Item set | Degree of support |
| {L1,L2} | 5 |
| {L1,L3} | 4 |
| {L1,L4} | 4 |
| {L1,L5} | 5 |
| {L2,L4} | 2 |
| {L2,L5} | 2 |
| {L4,L5} | 2 |
Traversing the second frequent item set table Lable2, and simultaneously self-connecting the second frequent item set table Lable2 to obtain a third candidate attribute item set table C3, as shown in Table 9:
table 9 candidate table C3
| Item set |
| {L1,L2,L3} |
| {L1,L2,L4} |
| {L1,L2,L5} |
| {L1,L3,L4} |
| {L1,L3,L5} |
| {L1,L4,L5} |
| {L2,L4,L5} |
In table 7, since non-frequent item sets exist in the item sets { L1, L2, L3}, { L1, L3, L4}, { L1, L3, and L5}, deletion is required, and the remaining item sets are retained to obtain the final third candidate set table C3, as shown in table 10
Table 10 candidate table C3
Deleting the item set with the support degree smaller than the preset support degree threshold value 2 in the third candidate set table to obtain a third frequent item set table lab 3 shown in table 11:
TABLE 11 frequent item set Table L3
| Item set | Degree of support |
| {L1,L2,L4} | 2 |
| {L1,L2,L5} | 2 |
Traversing the third frequent item set table Lable3, and self-connecting the third frequent item set table to obtain { L1, L2, L4, L5}, deleting the item set { L2, L4, L5} which does not belong to Lable3, so that the candidate item set C4 is empty, and ending the algorithm to obtain all frequent item sets.
And 43, discovering the strong association relationship between the classes and the attributes. Extracting all strong association rules in the third frequent item set Lable3 according to the confidence coefficient to obtain the strong association rule discovery result shown in Table 12:
table 12 strong association rule discovery
| Association rules | Confidence level |
| {L2,L5}→L1 | Support of { L1, L2, L5 }/{ L2, L5}, 100% |
The calculation method of confidence in the table is shown in formula (5):
as can be seen from table 12, when the attributes L2 (emergency resource category) and L5 (maximum scheduling radius) occur simultaneously, they must be in the emergency resource scheduling scheme class. With the increase of the number of cases in the safety emergency plan historical library, the method can be used for discovering the strong association relationship between the class and the attribute, so that the attribute structure of a certain class of the safety emergency plan is optimized.
Based on the same inventive concept, this embodiment further provides a computer device, which includes a memory and a processor, where the memory has an executable program, and when the processor executes the executable program, the computer device implements the above steps of the optimization method of the present invention.
The above embodiments are preferred embodiments of the present invention, but the present invention is not limited to the above embodiments, and any other changes, modifications, substitutions, combinations, and simplifications which do not depart from the spirit and principle of the present invention should be construed as equivalents thereof, and all such changes, modifications, substitutions, combinations, and simplifications are intended to be included in the scope of the present invention.
Claims (10)
1. The emergency plan content optimization method is characterized by comprising the following steps:
s1, collecting data, and establishing a historical safety emergency plan database;
s2, processing and sorting the collected data to form a pattern suitable for data analysis;
and S3, data mining and analyzing, putting training set data into the algorithm model for training by establishing a proper algorithm model, and finally obtaining a calculation result value with the minimum error with an actual result by continuously optimizing the algorithm model, so as to realize the optimization of the emergency plan content.
2. The emergency plan content optimization method according to claim 1, wherein step S1 is to form a historical security emergency plan database by analyzing the text form emergency plan, and extracting content structure and attribute parameters of the text form emergency plan; when the actual emergency type is the same as the safety emergency plan type, carrying out automatic emergency intelligent command according to the emergency plan designed by the safety emergency plan model, and simultaneously adding the emergency content into a historical safety emergency plan database; and if the actual emergency type is different from the safety emergency plan type, adding the contents of each part of the actual emergency into a historical safety emergency plan database.
3. The emergency plan content optimization method according to claim 1, wherein the step S3 includes optimizing attribute parameters of the safety emergency plan based on a random forest algorithm:
analyzing historical emergency command case data, preprocessing the historical emergency command case data, extracting attribute parameter influence characteristics of a safety emergency plan, and dividing the attribute parameters into category attribute parameters for qualitatively analyzing the emergency plan and numerical attribute parameters for quantitatively determining the disposal mode of the emergency plan;
and selecting a random forest algorithm to establish a regression model and a classification model, predicting the attribute parameter values, continuously optimizing the regression model and the classification model according to the error between the prediction result of the attribute parameter values and the actual case parameter values, and finally enabling the category attribute parameter values and the numerical attribute parameter values predicted according to the regression model and the classification model to be closer to the actual case parameter values so as to realize the setting of the attribute parameters of the intelligent safety emergency plan.
4. The emergency plan content optimization method according to claim 3, wherein step S3 is implemented by using a random forest algorithm, repeatedly and randomly extracting k samples from the original training sample set N in a replacement manner to generate a new training sample set, and then generating k decision trees according to the new training sample set to form a random forest, wherein the prediction result of the sample to be tested is determined by an average value of predicted values of the k decision trees.
5. The emergency plan content optimization method of claim 3, wherein the step S3 is based on a process of predicting attribute parameter values by a random forest algorithm, and comprises:
s31, acquiring emergency case data in a historical safety emergency plan database, and constructing an initial sample data set;
s32, processing the data of the safety emergency plan;
s33, processing the characteristics of the emergency data, screening the characteristics of which the correlation is greater than a set threshold value, and combining the characteristics of which the correlation is strong;
s34, extracting the characteristics of the emergency data, randomly selecting j characteristics from the i characteristics for each training set, wherein j < < i, and constructing a single decision tree by using the randomly selected j characteristics;
s35, traversing each feature in each training set and the feature value corresponding to the feature value, selecting the optimal feature to segment the training sets, and segmenting each training set into a data space R1 and a data space R2;
s36, recursively calling the step S35 for the two segmented data spaces R1 and R2 until no optimal features can be used for segmenting the training set or the deepest depth of the tree is reached, and generating a single decision tree;
s37, repeating the steps S35 and S36 by using n training sets respectively to construct a single decision tree, and finally constructing a random forest consisting of n decision trees;
and S38, predicting the attribute parameter values by using the random forest.
6. The method for optimizing the contents of an emergency plan according to claim 5, wherein the step S38 of predicting the numerical attribute parameter values by using a random forest comprises the following steps:
randomly extracting an emergency case in the test set as a case to be predicted, firstly putting the emergency case into a decision tree, segmenting the data set according to a first optimal characteristic, and if the sample is classified into a leaf node data set after segmentation, taking the mean value of the influence range of the data set as a numerical attribute parameter prediction value of the case to be predicted; if the sample is not the leaf node data set, continuing segmentation until the sample is included in the leaf node data set, and taking the numerical attribute parameter mean value of the final data set as a predicted value of the numerical attribute parameter of the case to be predicted; putting the case to be predicted into other decision trees, repeating the steps, and solving the mean value of the n predicted values to obtain the predicted value of the final numerical attribute parameter;
the process of predicting the category attribute parameter values by using the random forest in step S38 is as follows:
randomly extracting an emergency case in the test set as a case to be predicted, firstly putting the emergency case into a decision tree, segmenting the data set according to a first optimal characteristic, and if the sample is classified into a leaf node data set after segmentation, taking the maximum sample number probability in the data set as a category attribute parameter prediction value of the case to be predicted; if the sample is not a leaf node dataset, continuing to cut until the sample is included in the leaf node dataset; and putting the cases to be predicted into the other decision trees, repeating the steps to obtain predicted values of the n category attribute parameters, and taking the category with the most prediction of the sample as the final predicted value of the category attribute parameters according to a voting principle.
7. The emergency protocol content optimization method according to claim 1, wherein the step S3 includes optimizing an attribute structure of the safety emergency protocol based on a correlation algorithm:
analyzing the attribute structure of the safety emergency plan, establishing an association item set table of the contents and the attributes of the safety emergency plan, finding a frequent item set of the contents and the attributes of the safety emergency plan, and finding a strong association rule of the contents and the attributes of the safety emergency plan according to the set support degree and confidence; and adding the corresponding attributes into the safety emergency plan model by combining the found strong association rules, and continuously optimizing the attribute structure of the safety emergency plan model, so that the setting of the content attributes of the safety emergency plan can better accord with the actual case situation.
8. The emergency protocol content optimization method according to claim 7, wherein the step S3 is a process of optimizing an attribute structure of the security emergency protocol based on a correlation algorithm, and includes:
s31', acquiring a data set, analyzing the association model of the emergency resource scheduling class and the corresponding attribute in the safety emergency plan, and discovering the attribute of the emergency resource scheduling class; or the safety emergency plan is selected as a class, and the corresponding attribute is found.
S32', establishing a frequent item set, firstly setting a support degree and a confidence degree threshold value, wherein the support degree represents the times of the emergency resource scheduling class and a certain attribute appearing in a safety emergency plan at the same time, and the confidence degree represents the times of the emergency resource scheduling class and a certain attribute appearing in a safety emergency plan/the number of historical emergency plans;
s33', according to the confidence coefficient, extracting all strong association rules in the frequent item set established in the step S32' to find the strong association relation between the class and the attribute, thereby optimizing the attribute structure of a certain class of the safety emergency plan.
9. The emergency plan content optimization method according to claim 8, wherein the step S32' creates a frequent item set as follows:
establishing an initial plan item set table, traversing all cases in the initial plan item set table, and counting each attribute to obtain a candidate attribute item set table; deleting the item set with the support degree smaller than a preset support degree threshold value in the candidate attribute item set table, reserving the item set with the support degree larger than or equal to the preset support degree threshold value to obtain a first frequent item set, and obtaining a second candidate attribute item set table by utilizing self-connection of the first frequent item set; deleting all the item sets with the support degrees smaller than a preset support degree threshold value in the second candidate attribute item set table to obtain a second frequent item set table; traversing the second frequent item set table, and self-connecting the second frequent item set table to obtain a third candidate attribute item set table; deleting the item set with the support degree smaller than a preset support degree threshold value in the third candidate attribute item set table to obtain a third complex item set table;
and repeating in a circulating mode, acquiring a new candidate attribute item set table by traversing the previous frequent item set table and connecting the previous frequent item set table, and deleting an item set with the support degree smaller than a preset support degree threshold value in the new candidate attribute item set table to obtain a new frequent item set table until the candidate attribute item set is empty, thereby obtaining all frequent item sets.
10. Computer device comprising a memory and a processor, the memory having an executable program thereon, characterized in that the processor, when executing the executable program, performs the steps of the optimization method according to any one of claims 1 to 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010088012.XA CN111324657B (en) | 2020-02-12 | 2020-02-12 | Emergency plan content optimization method and computer equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010088012.XA CN111324657B (en) | 2020-02-12 | 2020-02-12 | Emergency plan content optimization method and computer equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111324657A true CN111324657A (en) | 2020-06-23 |
| CN111324657B CN111324657B (en) | 2023-09-08 |
Family
ID=71172675
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010088012.XA Active CN111324657B (en) | 2020-02-12 | 2020-02-12 | Emergency plan content optimization method and computer equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111324657B (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112258068A (en) * | 2020-10-30 | 2021-01-22 | 武汉理工光科股份有限公司 | Security activity scheme generation and dynamic update method and device |
| CN112258367A (en) * | 2020-11-13 | 2021-01-22 | 珠海大横琴科技发展有限公司 | Monitoring processing method and device |
| CN113034017A (en) * | 2021-03-31 | 2021-06-25 | 重庆城市管理职业学院 | Application security service platform based on cloud computing |
| CN114997754A (en) * | 2022-08-03 | 2022-09-02 | 广东省电信规划设计院有限公司 | Emergency plan analysis method and device based on cloud model and entropy weight method |
| CN115547055A (en) * | 2022-11-30 | 2022-12-30 | 广东省科技基础条件平台中心 | Traffic signal lamp coordination control method and device, storage medium and equipment |
| CN115544267A (en) * | 2022-09-27 | 2022-12-30 | 广东师大维智信息科技有限公司 | Knowledge graph-based simulation dynamic verification method for multiple emergency treatments |
| CN115829061A (en) * | 2023-02-21 | 2023-03-21 | 中国电子科技集团公司第二十八研究所 | Emergency accident disposal method based on historical case and empirical knowledge learning |
| CN116934127A (en) * | 2023-09-19 | 2023-10-24 | 中国铁塔股份有限公司吉林省分公司 | Emergency plan generation method and system based on intelligent platform |
| CN117371933A (en) * | 2023-10-08 | 2024-01-09 | 广州市科颐轩电子科技有限公司 | Intelligent laboratory management system based on Internet of things |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106202561A (en) * | 2016-07-29 | 2016-12-07 | 北京联创众升科技有限公司 | Digitized contingency management case library construction methods based on the big data of text and device |
| CN108268460A (en) * | 2016-12-30 | 2018-07-10 | 广东精点数据科技股份有限公司 | A kind of method for automatically selecting optimal models based on big data |
| CN108573078A (en) * | 2017-03-09 | 2018-09-25 | 中国石油化工股份有限公司 | Post-frac effect forecasting method based on data mining |
| CN109584564A (en) * | 2018-12-24 | 2019-04-05 | 上海羡通交通科技有限公司 | A kind of letter prosecutor case being applicable in more equipment implements optimization method and its system and device |
-
2020
- 2020-02-12 CN CN202010088012.XA patent/CN111324657B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106202561A (en) * | 2016-07-29 | 2016-12-07 | 北京联创众升科技有限公司 | Digitized contingency management case library construction methods based on the big data of text and device |
| CN108268460A (en) * | 2016-12-30 | 2018-07-10 | 广东精点数据科技股份有限公司 | A kind of method for automatically selecting optimal models based on big data |
| CN108573078A (en) * | 2017-03-09 | 2018-09-25 | 中国石油化工股份有限公司 | Post-frac effect forecasting method based on data mining |
| CN109584564A (en) * | 2018-12-24 | 2019-04-05 | 上海羡通交通科技有限公司 | A kind of letter prosecutor case being applicable in more equipment implements optimization method and its system and device |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112258068A (en) * | 2020-10-30 | 2021-01-22 | 武汉理工光科股份有限公司 | Security activity scheme generation and dynamic update method and device |
| CN112258367A (en) * | 2020-11-13 | 2021-01-22 | 珠海大横琴科技发展有限公司 | Monitoring processing method and device |
| CN113034017A (en) * | 2021-03-31 | 2021-06-25 | 重庆城市管理职业学院 | Application security service platform based on cloud computing |
| CN114997754A (en) * | 2022-08-03 | 2022-09-02 | 广东省电信规划设计院有限公司 | Emergency plan analysis method and device based on cloud model and entropy weight method |
| CN115544267B (en) * | 2022-09-27 | 2023-06-06 | 广东师大维智信息科技有限公司 | Knowledge graph-based simulation dynamic verification method for multiple emergency treatments |
| CN115544267A (en) * | 2022-09-27 | 2022-12-30 | 广东师大维智信息科技有限公司 | Knowledge graph-based simulation dynamic verification method for multiple emergency treatments |
| CN115547055B (en) * | 2022-11-30 | 2023-03-28 | 广东省科技基础条件平台中心 | Traffic signal lamp coordination control method and device, storage medium and equipment |
| CN115547055A (en) * | 2022-11-30 | 2022-12-30 | 广东省科技基础条件平台中心 | Traffic signal lamp coordination control method and device, storage medium and equipment |
| CN115829061A (en) * | 2023-02-21 | 2023-03-21 | 中国电子科技集团公司第二十八研究所 | Emergency accident disposal method based on historical case and empirical knowledge learning |
| CN115829061B (en) * | 2023-02-21 | 2023-04-28 | 中国电子科技集团公司第二十八研究所 | Emergency accident handling method based on historical case and experience knowledge learning |
| CN116934127A (en) * | 2023-09-19 | 2023-10-24 | 中国铁塔股份有限公司吉林省分公司 | Emergency plan generation method and system based on intelligent platform |
| CN116934127B (en) * | 2023-09-19 | 2023-11-24 | 中国铁塔股份有限公司吉林省分公司 | Emergency plan generation method and system based on intelligent platform |
| CN117371933A (en) * | 2023-10-08 | 2024-01-09 | 广州市科颐轩电子科技有限公司 | Intelligent laboratory management system based on Internet of things |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111324657B (en) | 2023-09-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111324657A (en) | Emergency plan content optimization method and computer equipment | |
| JP7090936B2 (en) | ESG-based corporate evaluation execution device and its operation method | |
| TWI723528B (en) | Computer-executed event risk assessment method and device, computer-readable storage medium and computing equipment | |
| US8090743B2 (en) | Document management system and method | |
| EP1677213A2 (en) | Data object association based on graph theory techniques | |
| CN110458324B (en) | Method and device for calculating risk probability and computer equipment | |
| CA2144255A1 (en) | Training apparatus and method | |
| CN106708738B (en) | Software test defect prediction method and system | |
| Sikder et al. | Application of rough set and decision tree for characterization of premonitory factors of low seismic activity | |
| JP2000339351A (en) | System for identifying selectively related database record | |
| Ciflikli et al. | Enhancing product quality of a process | |
| CN111160959A (en) | User click conversion estimation method and device | |
| CN113052225A (en) | Alarm convergence method and device based on clustering algorithm and time sequence association rule | |
| CN115794803A (en) | Engineering audit problem monitoring method and system based on big data AI technology | |
| CN113177644B (en) | Automatic modeling system based on word embedding and depth time sequence model | |
| Lasantha et al. | Data‐Driven versus Köppen–Geiger Systems of Climate Classification | |
| CN116842174A (en) | Agricultural resource database platform building method based on network data | |
| CN115619539A (en) | Pre-loan risk evaluation method and device | |
| US20040193573A1 (en) | Downward hierarchical classification of multivalue data | |
| CN112100165B (en) | Traffic data processing method, system, equipment and medium based on quality assessment | |
| CN110097126B (en) | Method for checking important personnel and house missing registration based on DBSCAN clustering algorithm | |
| KR102217092B1 (en) | Method and apparatus for providing quality information of application | |
| DE112021001565T5 (en) | SORTING DATA ITEMS OF A SPECIFIC SET OF DATA ITEMS | |
| CN112749990B (en) | Data analysis method and system based on tourist identity | |
| CN118447459B (en) | Landslide accumulation monitoring method and system based on deep learning |
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 | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Address after: Room 201, No. 1029 and 1031 Gaopu Road, Tianhe District, Guangzhou City, Guangdong Province, 510520 (office only) Applicant after: Ogilvy Technology Co.,Ltd. Address before: 2 / F, building C, 1029 Gaopu Road, Tianhe Software Park, Tianhe District, Guangzhou, Guangdong 510520 Applicant before: AUGUR INTELLIGENCE TECHNOLOGY (GUANGZHOU) Co.,Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |