Enterprise hidden danger-accident association early warning method based on text mining

Abstract

The invention discloses an enterprise hidden danger-accident association early warning method based on text mining, and relates to the technical field of association early warning and oil gas safety. According to the method, the high relevance between hidden dangers and accidents can be better revealed, the incidence relation between'accident cause segmented words' and'hidden dangers' is used as a data source of the model, and an association rule mining algorithm (Apriori algorithm) is used for finding out the hidden dangers; a potential association mode between the hidden danger and the accident is mined through a chain of hidden danger-keyword-accident. The method not only comprehensively reflects the relationship between the hidden danger and the accident from the perspective of data, but also reveals that the hidden danger is the key cause of the occurrence of some accidents, and the analysis method is obviously superior to the traditional hidden danger analysis means depending on subjective judgment or single dimension.

Description

Technical Field

[0001] The present invention relates to the technical field of correlation warning and oil and gas safety, and in particular to a text mining-based enterprise hidden danger-accident correlation warning method. Background Art

[0002] In the field of oil and gas safety, causal evolution theory is rarely used, and most research focuses on the occurrence of accidents themselves and the multiplication of related causal factors.

[0003] The evolution of oil and gas production accidents follows the chain process of "oil and gas production safety hazards-oil and gas production disasters and accidents", but there is still a lack of relevant research on the paths through which oil and gas production hazards achieve phased evolution and how to quantitatively analyze the key nodes of the oil and gas production safety hazard network from the perspective of accident inversion.

[0004] This is mainly manifested in the fact that research on the impact of hidden dangers on the occurrence of accidents is not sufficient. Most existing studies are qualitative, with few studies on the correlation between various influencing factors, and the analysis of the evolution mechanism of hidden dangers and accidents is still in a vague stage.

[0005] Therefore, analyzing and studying the relationship between "hidden dangers-accidents" has important theoretical significance and practical value for developing the connotation of risk evolution and expanding the research on safety risk prevention and control.

[0006] Therefore, a new solution to the above problems needs to be proposed. Summary of the invention

[0007] The purpose of the present invention is to provide an enterprise hidden danger-accident association early warning method based on text mining to solve the technical problems raised in the background technology.

[0008] To achieve the above object, the present invention provides the following technical solution: a method for early warning of enterprise hidden danger-accident association based on text mining, comprising at least the following steps:

[0009] S1: data preprocessing, which at least includes data cleaning, data integration and data missing marking;

[0010] S2: Performing text segmentation processing, wherein the text segmentation processing at least includes word library construction and text segmentation;

[0011] S3: extract keywords, and extract key words or phrases from each hidden danger segmentation record according to the results of the word segmentation processing in S2. These key words or phrases are information units that can reflect the core theme or essence of the hidden danger text;

[0012] S4: Analysis of hidden danger association rules based on Apriori algorithm. The hidden danger data set of the enterprise is processed and mined using Apriori algorithm for Boolean data in association rule mining. The association relationship between hidden dangers is mined using three indicators: support, confidence and lift.

[0013] S5: Build a hidden danger-accident association method. By analyzing the background, process and main causes of the accident, extract the causal factors related to the accident. After word segmentation, form a vocabulary set that can describe the cause of the accident, namely the accident causal word segmentation. After completing the extraction of hidden danger keywords and accident causal word segmentation, find the connection path between hidden dangers and accidents by associating and matching hidden danger keywords with accident causal word segmentation, form a "hidden danger-keyword-accident" chain, and link hidden dangers with historical accidents;

[0014] S6: Build a hidden danger-accident correlation model. Through in-depth analysis of the correlation between hidden dangers and accidents, use the correlation between "accident cause participles" and "hidden dangers and hidden dangers" as the core data source to build a "hidden danger-accident" correlation model. Furthermore, establish a causal relationship model through causal inference methods, clarify the correlation model between hidden dangers and accidents from the perspective of causal logic, and provide theoretical support and practical basis for scientific hidden danger management and accident prevention.

[0015] Furthermore, the data preprocessing in S1 at least includes the following steps:

[0016] First, data cleaning is performed. Since some irrelevant information in the original hidden danger text data does not provide important information for the construction of the analysis model, but instead creates additional burden during model training, it is necessary to remove it from the training set through data cleaning.

[0017] Secondly, data integration is performed. The original hidden danger text data still lacks the fields of "hidden danger occurrence process description" or "hidden danger occurrence cause description", so data integration is needed to improve the hidden danger description;

[0018] Finally, the missing data are marked. For potential data that are still missing information after data integration, they are marked as having unknown data sources and will not be included in subsequent analysis.

[0019] Furthermore, the word library construction includes at least the following steps:

[0020] First, in order to obtain better word segmentation effect, a third-party professional word library is introduced as a supplement. Jieba tool supports user-defined dictionaries to achieve better word segmentation effect. Therefore, the word library uses the safety production word library, petroleum and petrochemical professional terminology word library and petroleum professional word library in Sogou Cell Word Library as the basic word library, that is, the third-party library is used as the basic word library, and custom words are added according to the characteristics of enterprise hidden danger text to improve the custom dictionary;

[0021] Secondly, by supplementing the custom dictionary, specific words are added according to actual needs to better adapt to the characteristics of enterprise hidden danger texts and accident texts, which can enhance the accuracy of word segmentation of industry-specific texts;

[0022] Finally, stop words are processed to improve the accuracy of text processing. In practical applications, there are often some contents in the text that are not specifically helpful for analysis, namely stop words. The stop words include at least numbers, modal particles, conjunctions and symbols, etc. The stop words have no substantial contribution to text analysis and will affect subsequent text processing. Therefore, they need to be removed from the word segmentation results. The Harbin Institute of Technology stop word list and the custom stop word library are used. The Harbin Institute of Technology stop word list is a general Chinese stop word list, and the custom stop word library is based on the characteristics of the enterprise hidden danger text to remove some specific meaningless words. The specific meaningless words include at least names of people and places.

[0023] Furthermore, the purpose of the text segmentation is to segment each data risk into separate words, so as to facilitate better association in subsequent association rule mining;

[0024] The text segmentation uses Jieba segmentation tool to perform segmentation processing on accident potential hazard data, and obtains segmentation results efficiently and accurately by importing the constructed hidden danger professional word library and stop word library;

[0025] The word segmentation process is written in Python language.

[0026] Furthermore, the S4 at least includes the following steps:

[0027] Establish a hidden danger data set for the cleaned hidden danger data;

[0028] Using the Apriori algorithm, under the minimum support threshold of 0.002, the minimum confidence threshold of 0.3 and the minimum lift threshold of 1, the rules with significant association, namely association rules, are extracted from the frequent item sets;

[0029] The top-ranked association rules are analyzed with high support, high confidence, and high lift to study the complex characteristics of hidden dangers.

[0030] Furthermore, the S5 at least includes the following steps:

[0031] Pre-process the hidden danger data, perform word segmentation on it according to the hidden danger description, and extract hidden danger keywords based on the word segmentation results;

[0032] Then, the hidden danger association rules are used to find the hidden danger words in other frequent item sets associated with the current hidden danger word, and the hidden danger elements are associated with each other;

[0033] Accident data mining: classify accidents according to accident investigation reports, find the direct and indirect causes of the accidents, segment the causes of the accidents through Jieba word segmentation tools, and finally obtain the accident cause segmentation;

[0034] Based on the association rules between discovered hidden danger data and hidden dangers, the existing hidden danger data, possible chain hidden danger data and possible accident types are analyzed for association. Specifically, the association rule analysis from hidden danger information to accident cases is realized through the path of "hidden danger-keyword-accident".

[0035] Further, the S6 at least includes the following steps:

[0036] Collection and organization of accident data: Collect and organize accident investigation reports to extract the direct and indirect causes of each accident;

[0037] Accident data segmentation processing: Import the cause of the accident into the accident cause database, and use the Jieba segmentation tool to segment the cause of the accident to obtain the accident cause segmentation;

[0038] Analysis of the relationship between hidden dangers and accidents: Through the chain of "hidden dangers - keywords - accidents", hidden dangers are linked to historical accidents, thus obtaining the causal chain of hidden dangers - accidents;

[0039] Model construction: Establish a causal relationship model through causal inference methods to clarify the causal relationship between hidden dangers and accidents.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] 1. The present invention can better reveal the high correlation between hidden dangers and accidents. It uses the correlation between "accident cause segmentation" and "hidden danger and hidden danger" as the data source of the model, and uses the association rule mining algorithm (Apriori algorithm) to find out the relationship between hidden dangers and hidden dangers. Through the chain of "hidden danger-keyword-accident", the potential correlation pattern between hidden dangers and accidents is mined. This method not only fully reflects the relationship between hidden dangers and accidents from a data perspective, but also reveals that hidden dangers are the key cause of certain accidents. This analysis method is significantly better than traditional hidden danger analysis methods that rely on subjective judgment or a single dimension;

[0042] 2. The present invention introduces a causal inference method to construct a "hidden danger-accident" association model. The causal inference method is used to associate and match hidden danger keywords with accident cause segmentation words to construct a "hidden danger-accident" association model, which further verifies the causal relationship between hidden dangers and accidents. Compared with the prior art, which generally lacks in-depth analysis of the causal relationship between hidden dangers and accidents, the technical solution of the present invention is more scientific and convincing, and provides strong theoretical support for accident prevention and hidden danger management. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying any creative work.

[0044] Figure 1 It is a schematic diagram of the overall process of the present invention;

[0045] Figure 2 This is an example diagram of the "hidden danger-accident" association model of the present invention. DETAILED DESCRIPTION

[0046] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0047] See also Figure 1 and Figure 2 ,The enterprise hidden danger-accident association early warning method based on text mining at least includes the following steps:

[0048] S1: Data preprocessing, which includes at least data cleaning, data integration and data missing marking;

[0049] S2: Perform text segmentation processing, which at least includes word library construction and text segmentation;

[0050] S3: extract keywords, and extract key words or phrases from each hidden danger segmentation record according to the results of the word segmentation processing in S2. These key words or phrases are information units that can reflect the core theme or essence of the hidden danger text;

[0051] S4: Analysis of hidden danger association rules based on Apriori algorithm. The hidden danger data set of the enterprise is processed and mined using Apriori algorithm for Boolean data in association rule mining. The association relationship between hidden dangers is mined using three indicators: support, confidence and lift.

[0052] S5: Build a hidden danger-accident association method. By analyzing the background, process and main causes of the accident, extract the causal factors related to the accident. After word segmentation, form a vocabulary set that can describe the cause of the accident, namely the accident causal word segmentation. After completing the extraction of hidden danger keywords and accident causal word segmentation, find the connection path between hidden dangers and accidents by associating and matching hidden danger keywords with accident causal word segmentation, form a "hidden danger-keyword-accident" chain, and link hidden dangers with historical accidents;

[0053] S6: Build a hidden danger-accident correlation model. Through in-depth analysis of the correlation between hidden dangers and accidents, use the correlation between "accident cause participles" and "hidden dangers and hidden dangers" as the core data source to construct a "hidden danger-accident" correlation model. Use causal inference methods to establish a causal relationship model, clarify the correlation model between hidden dangers and accidents from the perspective of causal logic, and provide theoretical support and practical basis for scientific hidden danger management and accident prevention.

[0054] The data preprocessing in S1 includes at least the following steps:

[0055] First, data cleaning is performed. Since some irrelevant information in the original hidden danger text data does not provide important information for the construction of the analysis model, but instead creates additional burden during model training, it is necessary to remove it from the training set through data cleaning.

[0056] Secondly, data integration is performed. The original hidden danger text data still lacks the fields of "hidden danger occurrence process description" or "hidden danger occurrence cause description", so data integration is needed to improve the hidden danger description;

[0057] Finally, the missing data are marked. For potential data that are still missing information after data integration, they are marked as having unknown data sources and will not be included in subsequent analysis.

[0058] The construction of the vocabulary library includes at least the following steps:

[0059] First, in order to obtain better word segmentation effect, a third-party professional word library is introduced as a supplement. Jieba tool supports user-defined dictionaries to achieve better word segmentation effect. Therefore, the word library uses the safety production word library, petroleum and petrochemical professional terminology word library and petroleum professional word library in Sogou Cell Word Library as the basic word library, that is, the third-party library is used as the basic word library, and custom words are added according to the characteristics of enterprise hidden danger text to improve the custom dictionary;

[0060] Secondly, by supplementing the custom dictionary, specific words are added according to actual needs to better adapt to the characteristics of enterprise hidden danger texts and accident texts, which can enhance the accuracy of word segmentation of industry-specific texts;

[0061] Finally, stop words are processed to improve the accuracy of text processing. In practical applications, there are often some contents in the text that are not specifically helpful for analysis, namely stop words. Stop words include at least numbers, modal particles, conjunctions and symbols. Stop words have no substantial contribution to text analysis and will affect subsequent text processing. Therefore, they need to be removed from the word segmentation results. The Harbin Institute of Technology stop word list and custom stop word library are used. The Harbin Institute of Technology stop word list is a general Chinese stop word list. The custom stop word library is based on the characteristics of the enterprise hidden danger text to remove some specific meaningless words. Specific meaningless words include at least names of people and places.

[0062] The purpose of text segmentation is to divide each data risk into separate words, so as to better associate them in the subsequent association rule mining;

[0063] For text segmentation, Jieba word segmentation tool is selected to perform word segmentation on accident potential hazard data. By importing the constructed professional word library and stop word library, the word segmentation results can be obtained efficiently and accurately.

[0064] The word segmentation processing is written in Python.

[0065] S4 at least includes the following steps:

[0066] Establish a hidden danger data set for the cleaned hidden danger data;

[0067] Using the Apriori algorithm, under the minimum support threshold of 0.002, the minimum confidence threshold of 0.3 and the minimum lift threshold of 1, the rules with significant association, namely association rules, are extracted from the frequent item sets;

[0068] The top-ranked association rules are analyzed with high support, high confidence, and high lift to study the complex characteristics of hidden dangers.

[0069] S5 at least includes the following steps:

[0070] Pre-process the hidden danger data, perform word segmentation on it according to the hidden danger description, and extract hidden danger keywords based on the word segmentation results;

[0071] Then, the hidden danger association rules are used to find the hidden danger words in other frequent item sets associated with the current hidden danger word, and the hidden danger elements are associated with each other;

[0072] Accident data mining: classify accidents according to accident investigation reports, find the direct and indirect causes of the accidents, segment the causes of the accidents through Jieba word segmentation tools, and finally obtain the accident cause segmentation;

[0073] Based on the association rules between discovered hidden danger data and hidden dangers, the existing hidden danger data, possible chain hidden danger data and possible accident types are analyzed for association. Specifically, the association rule analysis from hidden danger information to accident cases is realized through the path of "hidden danger-keyword-accident".

[0074] S6 at least comprises the following steps:

[0075] Collection and organization of accident data: Collect and organize accident investigation reports to extract the direct and indirect causes of each accident;

[0076] Accident data segmentation processing: Import the cause of the accident into the accident cause database, and use the Jieba segmentation tool to segment the cause of the accident to obtain the accident cause segmentation;

[0077] Analysis of the relationship between hidden dangers and accidents: Through the chain of "hidden dangers - keywords - accidents", hidden dangers are linked to historical accidents, thus obtaining the causal chain of hidden dangers - accidents;

[0078] Model construction: Establish a causal relationship model through causal inference methods to clarify the causal relationship between hidden dangers and accidents.

[0079] The above process can be simplified as follows:

[0080] 1. “Hazard-Accident” Correlation Pattern and Evolution Analysis

[0081] Based on historical data, hidden dangers are linked to historical accidents through the chain of "hidden dangers-keywords-accidents", the "hidden dangers-accidents" evolution pattern is extracted and summarized, the evolutionary relationship between hidden dangers and accidents on the timeline is analyzed, and the path of hidden dangers triggering accidents is revealed.

[0082] 2. Establish a “Hazard-Accident” Model

[0083] The causal inference method is used to analyze the causal relationship between hidden dangers and accidents, and a "hidden danger-accident" association model is constructed to clarify how specific hidden dangers affect or cause specific types of accidents.

[0084] 3. Analysis of hidden danger association rules

[0085] Based on the Apriori algorithm, combined with rules with high support, high confidence and high lift, the correlation between hidden dangers is explored to reveal the potential mutual influence of hidden dangers.

[0086] In summary:

[0087] The hidden danger data is preliminarily processed through text mining technology, including preprocessing operations such as data cleaning, data integration, and missing data marking. The cleaned text data is segmented, the word library is constructed, and keywords are extracted. Then, the hidden danger association analysis is performed based on the Apriori algorithm, and the association relationship between hidden dangers is mined from the rules with high support, high confidence, and high lift. On this basis, the "hidden danger-keyword-accident" path is used to construct the association network between hidden dangers and accidents. By analyzing the background, process, and main causes of the accident, a set of keywords related to the accident is extracted, and the causal inference method is used to match the hidden danger keywords with the accident cause keywords to form a "hidden danger-keyword-accident" chain, effectively linking hidden dangers with historical accidents, and finally forming a "hidden danger-accident" association model.

[0088] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. Enterprise hidden danger-accident correlation early warning method based on text mining, characterized by: At least the following steps are included: S1: data preprocessing, which at least includes data cleaning, data integration and data missing marking; S2: Performing text segmentation processing, wherein the text segmentation processing at least includes word library construction and text segmentation; S3: extract keywords, and extract key words or phrases from each hidden danger segmentation record according to the results of the word segmentation processing in S2. These key words or phrases are information units that can reflect the core theme or essence of the hidden danger text; S4: Analysis of hidden danger association rules based on Apriori algorithm. The hidden danger data set of the enterprise is processed and mined using Apriori algorithm for Boolean data in association rule mining. The association relationship between hidden dangers is mined using three indicators: support, confidence and lift. S5: Build a hidden danger-accident association method. By analyzing the background, process and main causes of the accident, extract the causal factors related to the accident. After word segmentation, form a vocabulary set that can describe the cause of the accident, namely the accident causal word segmentation. After completing the extraction of hidden danger keywords and accident causal word segmentation, find the connection path between hidden dangers and accidents by associating and matching hidden danger keywords with accident causal word segmentation, form a "hidden danger-keyword-accident" chain, and link hidden dangers with historical accidents; S6: Build a hidden danger-accident correlation model. Through in-depth analysis of the correlation between hidden dangers and accidents, take the correlation between "accident cause participles" and "hidden dangers and hidden dangers" as the core data source, and build a "hidden danger-accident" correlation model. Furthermore, establish a causal relationship model through causal inference methods, clarify the correlation model between hidden dangers and accidents from the perspective of causal logic, and provide theoretical support and practical basis for scientific hidden danger management and accident prevention.

2. The enterprise hidden danger-accident association early warning method based on text mining according to claim 1 is characterized by: The data preprocessing in S1 at least includes the following steps: First, data cleaning is performed. Since some irrelevant information in the original hidden danger text data does not provide important information for the construction of the analysis model, but instead creates additional burden during model training, it is necessary to remove it from the training set through data cleaning. Secondly, data integration is performed. The original hidden danger text data still lacks the fields of "hidden danger occurrence process description" or "hidden danger occurrence cause description", so data integration is needed to improve the hidden danger description; Finally, the missing data are marked. For potential data that are still missing information after data integration, they are marked as having unknown data sources and will not be included in subsequent analysis.

3. The enterprise hidden danger-accident association early warning method based on text mining according to claim 1 is characterized by: The word library construction includes at least the following steps: First, in order to obtain better word segmentation effect, a third-party professional word library is introduced as a supplement. Jieba tool supports user-defined dictionaries to achieve better word segmentation effect. Therefore, the word library uses the safety production word library, petroleum and petrochemical professional terminology word library and petroleum professional word library in Sogou Cell Word Library as the basic word library, that is, the third-party library is used as the basic word library, and custom words are added according to the characteristics of enterprise hidden danger text to improve the custom dictionary; Secondly, by supplementing the custom dictionary, specific words are added according to actual needs to better adapt to the characteristics of enterprise hidden danger texts and accident texts, which can enhance the accuracy of word segmentation of industry-specific texts; Finally, stop words are processed to improve the accuracy of text processing. In practical applications, there are often some contents in the text that are not specifically helpful for analysis, namely stop words. The stop words include at least numbers, modal particles, conjunctions and symbols, etc. The stop words have no substantial contribution to text analysis and will affect subsequent text processing. Therefore, they need to be removed from the word segmentation results. The Harbin Institute of Technology stop word list and the custom stop word library are used. The Harbin Institute of Technology stop word list is a general Chinese stop word list, and the custom stop word library is based on the characteristics of the enterprise hidden danger text to remove some specific meaningless words. The specific meaningless words include at least names of people and places.

4. The enterprise hidden danger-accident association early warning method based on text mining according to claim 1 is characterized by: The purpose of the text segmentation is to segment each data risk into individual words, so as to facilitate better association in subsequent association rule mining; The text segmentation uses Jieba segmentation tool to perform segmentation processing on accident potential hazard data, and obtains segmentation results efficiently and accurately by importing the constructed hidden danger professional word library and stop word library; The word segmentation process is written in Python language.

5. The enterprise hidden danger-accident association early warning method based on text mining according to claim 1 is characterized by: The S4 at least comprises the following steps: Establish a hidden danger data set for the cleaned hidden danger data; Using the Apriori algorithm, under the minimum support threshold of 0.002, the minimum confidence threshold of 0.3 and the minimum lift threshold of 1, the rules with significant association, namely association rules, are extracted from the frequent item sets; The top-ranked association rules are analyzed with high support, high confidence, and high lift to study the complex characteristics of hidden dangers.

6. The enterprise hidden danger-accident association early warning method based on text mining according to claim 1 is characterized by: The S5 at least comprises the following steps: Pre-process the hidden danger data, perform word segmentation on it according to the hidden danger description, and extract hidden danger keywords based on the word segmentation results; Then, the hidden danger association rules are used to find the hidden danger words in other frequent item sets associated with the current hidden danger word, and the hidden danger elements are associated with each other; Accident data mining: classify accidents according to accident investigation reports, find the direct and indirect causes of the accidents, segment the causes of the accidents through Jieba word segmentation tools, and finally obtain the accident cause segmentation; Based on the association rules between the discovered hidden danger data and the hidden dangers, the existing hidden danger data, the potential chain hidden danger data and the possible accident types are analyzed for association. Specifically, the association rule analysis from hidden danger information to accident cases is realized through the path of "hidden danger-keyword-accident".

7. The enterprise hidden danger-accident association early warning method based on text mining according to claim 1 is characterized by: The S6 at least comprises the following steps: Collection and organization of accident data: Collect and organize accident investigation reports to extract the direct and indirect causes of each accident; Accident data segmentation processing: Import the cause of the accident into the accident cause database, and use the Jieba segmentation tool to segment the cause of the accident to obtain the accident cause segmentation; Analysis of the relationship between hidden dangers and accidents: Through the chain of "hidden dangers - keywords - accidents", hidden dangers are linked with historical accidents, thus obtaining the causal chain of hidden dangers - accidents; Model construction: Establish a causal relationship model through causal inference methods to clarify the causal relationship between hidden dangers and accidents.

Images