Workers' unsafe behavior trend prediction method and system based on dual indicator clustering

Abstract

The present invention relates to the field of trend prediction technology, and in particular, to a method and system for predicting the trend of unsafe behavior of workers based on dual-indicator clustering. The method comprises: stratified sampling of all coal miners; constructing basic risk indicators and unsafe behavior risk indicators; performing kmeans clustering on individuals, teams or district teams among all coal miners based on the basic risk indicators and unsafe behavior risk indicators; constructing a Bayesian discriminant model, and clustering the individuals, teams or district teams among all coal miners based on the Bayesian discriminant model; predicting the unsafe behavior of individuals, teams or district teams among all coal miners based on the results of the clustering, and formulating corresponding preventive safety education plans and preventive control measures. This solves the problem that existing prediction methods cannot be applied to both basic risk indicators and unsafe behavior risk indicators.

Description

Technical Field

[0001] The present invention relates to the technical field of trend prediction, and in particular to a method and system for predicting worker unsafe behavior trends based on dual indicator clustering. Background Art

[0002] Underground coal mine operations have always been high-risk. Statistics show that more than half of the time spent underground is due to workers' individual risk factors and unsafe behaviors. Safety education for coal miners and preventative measures against these behaviors have always been key tasks in underground operations. Therefore, predicting trends in workers' unsafe behaviors is particularly important.

[0003] In existing technologies, most predictions of coal miner behavior are based on clustering. This involves analyzing workers' multidimensional attributes to construct user profiles, treating each worker as a node and clustering them based on node vector similarity. However, this method cannot guarantee the applicability of clustering results to subsequent behavior. Furthermore, existing technologies for predicting unsafe behavior trends are limited to predicting the number of unsafe behaviors through other indicators, and do not include profiling and predicting unsafe behaviors at different levels of objects (such as individuals, teams, or teams).

[0004] In response to the above problems, the present invention proposes a method and system for predicting workers' unsafe behavior trends based on dual indicator clustering. Summary of the Invention

[0005] In order to solve the problem that existing prediction methods cannot be applied to both basic risk indicators and unsafe behavior risk indicators, the present invention provides a worker unsafe behavior trend prediction method and system based on dual indicator clustering.

[0006] In a first aspect, the present invention provides a method for predicting worker unsafe behavior trends based on dual-indicator clustering, comprising:

[0007] Stratified sampling was conducted on all coal miners;

[0008] Construct basic risk indicators based on the sampled workers' personal basic indicators, safety education indicators, work situation indicators and health and physiological indicators;

[0009] Constructing an unsafe behavior risk index based on individual unsafe behavior events of sampled workers, wherein the individual unsafe behavior events are obtained based on the accident-causing dimension, the awareness dimension, and the overall dimension;

[0010] Perform kmeans clustering on individuals, teams or district teams among all coal miners based on the basic risk indicators and unsafe behavior risk indicators;

[0011] Construct a Bayesian discriminant model and classify all coal miners into clusters based on the Bayesian discriminant model;

[0012] Based on the results of the cluster division, unsafe behaviors of individuals, teams or teams among all coal mine workers are predicted, and corresponding preventive safety education plans and preventive control measures are formulated.

[0013] In some embodiments, the personal basic indicators specifically include length of service, education level, and professional skill level;

[0014] The safety education indicators include safety education participation and safety education assessment scores;

[0015] The work situation indicators include attendance, social exchange relationship I and social exchange relationship O;

[0016] The health physiological indicators include cardiovascular and cerebrovascular health, basal metabolic health, fatigue level, heart rate abnormality and body temperature abnormality.

[0017] In some embodiments, the accident-causing dimensions include: operating errors / ignoring safety / ignoring warnings, causing safety device failure, using unsafe equipment, operating with hands instead of tools, improper storage of objects, venturing into dangerous places, climbing in unsafe positions, working / staying under the boom, refueling / repairing / inspecting the machine while it is running, distracting behavior, not using personal protective equipment / equipment correctly, unsafe clothing, and incorrect handling of flammable / explosive dangerous goods;

[0018] The awareness dimension includes conscious unsafe behaviors and unconscious unsafe behaviors;

[0019] The overall dimension is the number of unsafe behaviors.

[0020] In some embodiments, performing kmeans clustering on the basic risk indicators and unsafe behavior risk indicators of individuals, teams, or teams among all coal miners includes:

[0021] Quantify the basic risk indicators and unsafe behavior risk indicators of individuals, teams or teams among all coal mine workers;

[0022] Initialize the k value to 2 and cluster the quantified basic risk indicators and unsafe behavior risk indicators;

[0023] After completing the clustering of the initialized k value, increase the k value from the initial value and cluster the quantified basic risk indicators and unsafe behavior risk indicators again;

[0024] After each clustering is completed, the sum of squares of the data errors within each cluster is calculated;

[0025] Draw a line graph about k based on the sum of squared errors of the data for each cluster;

[0026] Determine the optimal k value using the elbow method according to the line graph;

[0027] The clustering result corresponding to the optimal k value is selected as the clustering result of kmeams clustering;

[0028] Here, k is the number of clusters.

[0029] In some embodiments, the formula for calculating the sum of squared errors (SSE) of the data within each cluster is:

[0030] ;

[0031] Where SSE is the sum of squared errors, m is the total number of samples, is the i-th sample, is the center point of the j-th cluster, is the indicator function, It is a sample With cluster center The square of the Euclidean distance between belong hour, The value is 1, otherwise it is 0.

[0032] In some embodiments, the constructing of a Bayesian discriminant model and clustering of individuals, teams, or teams among all coal miners based on the Bayesian discriminant model includes:

[0033] Based on each cluster of the clustering results, calculate the prior probability of each cluster;

[0034] Based on each cluster of the clustering results, estimate the conditional probability density function of each cluster;

[0035] According to Bayes' theorem, the posterior probability of an individual, team, or district team among all coal miners belonging to each cluster in the clustering result is calculated as follows:

[0036] ;

[0037] in, is the posterior probability, is the prior probability, is the conditional probability density function, is the marginal probability of sample x;

[0038] All individuals, teams or district teams among all coal miners are assigned to the cluster with the largest posterior probability in the clustering results.

[0039] In some embodiments, the average value of the unsafe behavior risk index in the cluster to which the individuals, teams, or teams of all coal miners belong is used as the first unsafe behavior profile;

[0040] Based on the first unsafe behavior portrait, safety education plans and control measures are formulated for corresponding individuals, teams or district teams.

[0041] In some embodiments, the unsafe behaviors of individuals, teams, or divisions among all coal miners are predicted based on the clustering results, and corresponding preventive safety education plans and preventive control measures are formulated, including:

[0042] The basic risk indicators and unsafe behavior risk indicators of individuals, teams or districts in the T-1 period of all coal miners are input into the Bayesian discriminant model to obtain the first type of clusters of individuals, teams or districts in the T-1 period of all coal miners;

[0043] The basic risk index of period T and the unsafe behavior risk index of period T-1 of all coal miners are input into the Bayesian discriminant model to obtain the second type of cluster of period T of all coal miners, groups or teams;

[0044] In response to the fact that the first cluster and the second cluster are not the same cluster, the mean of the unsafe behavior risk indicators of the individuals, teams or districts in the second cluster during period T is used as the second unsafe behavior profile;

[0045] Based on the second unsafe behavior portrait, formulate corresponding preventive safety education plans and preventive control measures.

[0046] In some embodiments, in response to the first cluster and the second cluster being the same cluster, individuals, teams or divisions of all coal miners continue to use the safety education plan and control measures.

[0047] In the second aspect, the present invention proposes a worker unsafe behavior trend prediction system based on dual indicator clustering, which includes the following modules:

[0048] Sampling module: used to implement stratified sampling of all coal miners;

[0049] Quantification module: used to construct basic risk indicators based on the individual basic indicators, safety education indicators, work situation indicators and health and physiological indicators of the sampled workers, and to construct unsafe behavior risk indicators based on the individual unsafe behavior events of the sampled workers. The individual unsafe behavior events are obtained based on the accident-causing dimension, awareness dimension and overall dimension;

[0050] Clustering module: used to implement kmeans clustering of individuals, teams or district teams among all coal miners based on the basic risk indicators and unsafe behavior risk indicators;

[0051] Discrimination module: used to build a Bayesian discriminant model, and cluster the individuals, teams or districts among all coal mine workers based on the Bayesian discriminant model, predict the unsafe behaviors of individuals, teams or districts among all coal mine workers based on the results of the cluster division, and formulate corresponding preventive safety education plans and preventive control measures.

[0052] To solve the problem that existing prediction methods cannot be applied to both basic risk indicators and unsafe behavior risk indicators, the present invention has the following advantages:

[0053] The technical solution of this invention can predict unsafe behavior trends for individuals, teams, or teams based on coal miners' basic risk indicators and unsafe behavior risk indicators. Appropriate safety education programs and control measures can be selected based on the trend prediction results, thereby reducing underground operation risks caused by coal miners' own factors. Targeted education programs and control measures can also significantly save manpower and material costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 A flowchart of a method for predicting worker unsafe behavior trends based on dual indicator clustering is shown;

[0055] Figure 2 The structural block diagram of the worker unsafe behavior trend prediction system based on dual indicator clustering is shown. DETAILED DESCRIPTION

[0056] The present disclosure will now be discussed with reference to several exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the present disclosure, rather than to imply any limitation on the scope of the present disclosure.

[0057] As used herein, the term "including" and its variations are to be interpreted as open-ended terms meaning "including, but not limited to." The term "based on" is to be interpreted as "based, at least in part, on." The terms "one embodiment" and "an embodiment" are to be interpreted as "at least one embodiment." The term "another embodiment" is to be interpreted as "at least one other embodiment." Terms such as "upper," "lower," "left," "right," "front," "back," "top," "bottom," "inner," "outer," "vertical," "horizontal," "transverse," and "longitudinal" indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings. These terms are primarily intended to better describe the present application and its embodiments and are not intended to limit the devices, elements, or components indicated to having a specific orientation, or to being constructed and operated in a specific orientation. Furthermore, some of the above terms may be used to indicate other meanings besides orientation or positional relationships. For example, the term "on" may, in certain circumstances, be used to indicate a dependency or connection relationship. Those skilled in the art will understand the specific meanings of these terms in this application based on the specific circumstances. Furthermore, the terms "installed," "disposed," "provided with," "connected," and "connected" are to be interpreted broadly. For example, it can be a fixed connection, a detachable connection, or an integral structure; it can be a mechanical connection or an electrical connection; it can be a direct connection, an indirect connection through an intermediate medium, or an internal connection between two devices, elements, or components. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to the specific circumstances. In addition, the terms "first", "second", etc. are mainly used to distinguish different devices, elements, or components (the specific types and structures may be the same or different), and are not used to indicate or imply the relative importance or quantity of the indicated devices, elements, or components. Unless otherwise specified, "plurality" means two or more.

[0058] In the first aspect, this embodiment discloses a method for predicting worker unsafe behavior trends based on dual indicator clustering, such as Figure 1 As shown, including:

[0059] Stratified sampling was conducted on all coal miners;

[0060] Construct basic risk indicators based on the sampled workers' personal basic indicators, safety education indicators, work situation indicators and health and physiological indicators;

[0061] Constructing an unsafe behavior risk index based on individual unsafe behavior events of sampled workers, wherein the individual unsafe behavior events are obtained based on the accident-causing dimension, the awareness dimension, and the overall dimension;

[0062] Perform kmeans clustering on individuals, teams or district teams among all coal miners based on the basic risk indicators and unsafe behavior risk indicators;

[0063] Construct a Bayesian discriminant model and classify all coal miners into clusters based on the Bayesian discriminant model;

[0064] Based on the results of the cluster division, unsafe behaviors of individuals, teams or teams among all coal mine workers are predicted, and corresponding preventive safety education plans and preventive control measures are formulated.

[0065] In this embodiment, a method for predicting the trend of unsafe behavior of workers based on dual indicator clustering is proposed. First, all coal miners are stratified and sampled, and basic risk indicators are constructed based on the personal basic indicators, safety education indicators, work situation indicators and health and physiological indicators of the sampled workers. Unsafe behavior risk indicators are constructed based on the personal unsafe behavior events of the sampled workers in the accident-causing dimension, awareness dimension and overall dimension. Secondly, kmeans clustering is performed on individuals, teams or teams in all coal miners based on the basic risk indicators and unsafe behavior indicators. Thirdly, a Bayesian discriminant model is constructed, and the individuals, teams or teams in all coal miners are divided into the clustering results of kmeans clustering through the Bayesian discriminant model. Finally, the unsafe behaviors of individuals, teams or teams in all coal miners are predicted based on the division results of the Bayesian discriminant model, and corresponding preventive safety education plans and preventive control measures are formulated.

[0066] Specifically, stratified sampling was conducted by age group, including under 30 years old, 30-40 years old (excluding 40), 40-50 years old (excluding 50), and 50 years old and above. Simple random sampling was performed on the stratified results, and 20% of the workers in each stratum were sampled as the research sample.

[0067] Specifically, basic risk indicators are basic information about coal miners, including length of service, education, attendance, social relationships, and personal health. Unsafe behavior risk indicators refer to the types and number of accidents, the number of intentional unsafe behaviors, and the overall number of unsafe behaviors.

[0068] Specifically, K-Means clustering is a widely used unsupervised learning algorithm for dividing a dataset into K clusters such that the data points within each cluster are as similar as possible to each other, while the data points in different clusters are as different as possible.

[0069] Specifically, during clustering, all subcategories contained in the basic risk indicators and unsafe behavior risk indicators are clustered simultaneously, and the final clustering results are divided into individual, team and district categories. This makes the clustering results applicable to both basic risk indicators and unsafe behavior risk indicators. At the same time, for individuals or groups at different levels, individual or collective unsafe behavior portraits can also be formed based on the clustering results.

[0070] Specifically, the Bayesian Classifier is a statistical classification method based on Bayes' theorem that aims to predict the category of a data point based on known features. It calculates the posterior probability of each category and selects the category with the highest posterior probability as the prediction result.

[0071] Specifically, after obtaining a classification of workers through stratified sampling, the established Bayesian discriminant model is used to divide all coal miners into different clusters based on the kmeans clustering results. (If the clustering targets are work groups or teams, the work groups or teams are divided into the corresponding clusters in the clustering results.) Then, based on the changes in the current cluster and the previous cluster, future unsafe behaviors of workers, work groups, or teams are predicted. Based on the predicted results, corresponding preventive safety education plans and preventive control measures are formulated.

[0072] Specifically, the above method clusters both the basic risk indicator and the unsafe behavior risk indicator, allowing the clustering results to apply to both indicators simultaneously. Using a Bayesian discriminant model, all individuals, teams, or districts are grouped into clusters to predict unsafe behavior trends. This effectively captures behavioral changes in individuals, teams, or districts, allowing for predictions of their subsequent behavior and targeted safety education and preventive measures. This can significantly reduce the incidence of accidents caused by workers in underground operations, and targeted safety education and preventive measures can also significantly reduce human and material costs.

[0073] In some embodiments, the personal basic indicators specifically include length of service, education level, and professional skill level;

[0074] The safety education indicators include safety education participation and safety education assessment scores;

[0075] The work situation indicators include attendance, social exchange relationship I and social exchange relationship O;

[0076] The health physiological indicators include cardiovascular and cerebrovascular health, basal metabolic health, fatigue level, heart rate abnormality and body temperature abnormality.

[0077] Table 1 Basic risk indicators

[0078]

[0079] In this embodiment, as shown in Table 1, a detailed division of basic risk indicators into personal basic indicators, safety education indicators, work situation indicators, and health and physiological indicators is introduced.

[0080] Specifically, personal basic indicators include length of service, education level, and professional skill level. By normalizing the length of service of the sampled workers, a dimensionless indicator value is obtained. The normalization formula is:

[0081] ;

[0082] Among them, is the normalized value of worker i’s seniority index, is the original length of service of worker i, Take 0, Take 37 (according to the retirement age of coal miners at 55, the maximum length of service is 37 years). The simplified formula is:

[0083] ;

[0084] The quantitative method of academic qualifications is:

[0085] ;

[0086] The quantification method of professional skill level is:

[0087] .

[0088] Specifically, safety education indicators include safety education participation and safety education assessment scores. Safety education participation is quantified as follows:

[0089] ;

[0090] The quantification method of safety education assessment scores is:

[0091] ;

[0092] in, is the quantitative result of the safety education assessment score indicator, M is the number of times of participating in the assessment, is the score of worker i in the mth assessment, is the total score of the mth assessment.

[0093] Specifically, the work performance indicators include attendance, social exchange relationship I, and social exchange relationship O. The attendance performance indicator does not take leave factors into account, and only counts the expected attendance and actual attendance. The attendance performance indicator is quantified as follows:

[0094] ;

[0095] The data on social exchange relationships come from the mutual evaluation form of workers in the team / district. The mutual evaluation form is an evaluation of the work performance of each worker in the team / district in the coal mine, and the evaluation results are divided into three options: excellent, medium, and poor. is worker i's evaluation of worker j. More specifically, social exchange relationship I represents the evaluation of worker i by all other workers, highlighting the performance of a particular worker in the eyes of other workers. To highlight risk trends, excellent and average are grouped together, and the quantification method is the proportion of negative reviews. For example, the social exchange relationship I of worker i is:

[0096] ;

[0097] The social exchange relationship O represents the evaluation of a worker i on all other workers. Here, excellent and average are also considered as one category, and the quantification method is the proportion of negative reviews. For example, the social exchange relationship O of worker i is:

[0098] .

[0099] Specifically, health physiological indicators include cardiovascular health, basal metabolic health, fatigue level, heart rate abnormality, and body temperature abnormality. More specifically, cardiovascular health focuses on whether workers have cardiovascular diseases, including hypertension, hyperlipidemia, atherosclerosis, etc., and its quantification method is:

[0100] ;

[0101] Basic metabolic health focuses on whether workers have basic metabolic diseases, including diabetes, thyroid dysfunction, high uric acid, and liver dysfunction. Its quantification method is:

[0102] ;

[0103] The fatigue index data comes from fatigue warning events of smart helmets worn by coal miners. Taking the number of fatigue warning events x in the past month, the fatigue level is quantified as follows:

[0104] ;

[0105] The heart rate abnormality indicator data comes from the heart rate abnormality warning events of smart bracelets worn by coal miners. Taking the number of heart rate abnormality warning events in the past month as x, the heart rate abnormality is quantified as follows:

[0106] ;

[0107] The body temperature abnormality index data comes from the abnormal temperature warning events of smart bracelets worn by coal miners. Taking the number of abnormal temperature warning events in the past month as x, the body temperature abnormality is quantified as follows:

[0108] .

[0109] In some embodiments, the accident-causing dimensions include: operating errors / ignoring safety / ignoring warnings, causing safety device failure, using unsafe equipment, operating with hands instead of tools, improper storage of objects, venturing into dangerous places, climbing in unsafe positions, working / staying under the boom, refueling / repairing / inspecting the machine while it is running, distracting behavior, not using personal protective equipment / equipment correctly, unsafe clothing, and incorrect handling of flammable / explosive dangerous goods;

[0110] The awareness dimension includes conscious unsafe behaviors and unconscious unsafe behaviors;

[0111] The overall dimension is the number of unsafe behaviors.

[0112] In this embodiment, as shown in Table 2, a detailed division of basic risk indicators into personal basic indicators, safety education indicators, work situation indicators, and health and physiological indicators is introduced.

[0113] Specifically, as shown in Table 2, the quantification method of indicators R01-R15 is the same, that is, the ratio of the number of unsafe behaviors corresponding to the indicator to the total number of unsafe behaviors of an employee in the past month:

[0114] ;

[0115] in, is the quantitative value of the indicator m, The number of unsafe behaviors of worker i in the past month is m indicators, The total number of unsafe behaviors committed by worker i in the past month.

[0116] Specifically, the R16 indicator is quantified by taking the number of unsafe behaviors x committed by a worker in the past month and calculating:

[0117] .

[0118] surface Unsafe behavior risk indicators

[0119]

[0120] In some embodiments, performing kmeans clustering on the basic risk indicators and unsafe behavior risk indicators of individuals, teams, or teams among all coal miners includes:

[0121] Quantify the basic risk indicators and unsafe behavior risk indicators of individuals, teams or teams among all coal mine workers;

[0122] Initialize the k value to 2 and cluster the quantified basic risk indicators and unsafe behavior risk indicators;

[0123] After completing the clustering of the initialized k value, increase the k value from the initial value and cluster the quantified basic risk indicators and unsafe behavior risk indicators again;

[0124] After each clustering is completed, the sum of squares of the data errors within each cluster is calculated;

[0125] Draw a line graph about k based on the sum of squared errors of the data for each cluster;

[0126] Determine the optimal k value using the elbow method according to the line graph;

[0127] The clustering result corresponding to the optimal k value is selected as the clustering result of kmeans clustering;

[0128] Here, k is the number of clusters.

[0129] In some embodiments, the formula for calculating the sum of square errors of the data within each cluster is:

[0130] ;

[0131] Among them, SEE is the sum of squared errors of the data, m is the total number of samples, is the i-th sample, is the center point of the j-th cluster, is the indicator function, It is a sample With cluster center The square of the Euclidean distance between belong hour, The value is 1, otherwise it is 0.

[0132] This example describes the specific process of kmeans clustering. First, the basic risk indicators and unsafe behavior risk indicators for individuals, teams, or divisions within a sample of coal miners are quantified. Then, the k value is initialized to 2, and clustering is performed simultaneously on each of the quantified indicators. Finally, clustering is performed separately with increasing k values, and the elbow method is used to select the optimal k value. The clustering result corresponding to the optimal k value is obtained as the final clustering result.

[0133] Specifically, the k value is initialized to 2, and k centroids are randomly selected in the sample space. Each sample (coal miner) is assigned to the nearest cluster center (centroid). At this time, the distance is calculated using the Euclidean distance. For the sample and cluster centers , the distance calculation formula is:

[0134] ;

[0135] in, yes The mth characteristic (index) value of is the cluster center The mth eigenvalue of .

[0136] Recalculate the position of each cluster center. The new cluster center is the mean of all data points assigned to the cluster. For cluster j, the new cluster center is The calculation formula is:

[0137] ;

[0138] in, is the set of all data points assigned to cluster j, yes Repeat this step until the cluster of sample points no longer changes, and the clustering calculation is completed.

[0139] Specifically, when the elbow method is used to select the optimal k value, starting from the initial k value, after each clustering, the sum of squared errors (SSE) of the data within each cluster is calculated. The formula is:

[0140] ;

[0141] Among them, m is the total number of samples, k is the number of clusters, is the i-th sample, is the center point of the j-th cluster, is the indicator function, It is a sample With cluster center The square of the Euclidean distance between belong hour, The value is 1, otherwise it is 0.

[0142] Then, a line graph about k is drawn according to the SEE results of each clustering, the k value corresponding to the elbow position is selected as the optimal k value, and the clustering result corresponding to the optimal k value is selected as the final result.

[0143] Specifically, the elbow method is a heuristic method for determining the optimal number of clusters k in K-means clustering. It helps select the optimal number of clusters by analyzing the changing trend of the sum of squares within clusters under different values ​​of k.

[0144] In some embodiments, the constructing of a Bayesian discriminant model and clustering of individuals, teams, or teams among all coal miners based on the Bayesian discriminant model includes:

[0145] Based on each cluster of the clustering results, calculate the prior probability of each cluster;

[0146] Based on each cluster of the clustering results, estimate the conditional probability density function of each cluster;

[0147] According to Bayes' theorem, the posterior probability of an individual, team, or district team among all coal miners belonging to each cluster in the clustering result is calculated as follows:

[0148] ;

[0149] in, is the posterior probability, is the prior probability, is the conditional probability density function, is the marginal probability of sample x;

[0150] All individuals, teams or district teams among all coal miners are assigned to the cluster with the largest posterior probability in the clustering results.

[0151] In this embodiment, it is described how to construct a Bayesian discriminant model and use the model to assign individuals, teams, or districts among all coal miners to corresponding clusters. First, for each cluster in the clustering result, the prior probability is calculated, that is, the proportion of all samples belonging to this cluster. Second, for each cluster, its conditional probability density is estimated, assuming that each cluster obeys a normal distribution. Finally, according to Bayes' theorem, the posterior probability of each individual, team, or district among all coal miners belonging to each cluster is calculated, and they are divided into the cluster with the largest posterior probability.

[0152] Specifically, the Bayesian Classifier is a statistical classification method based on Bayes' theorem. It determines the most likely category for a new data point by calculating the posterior probability of each category. This method not only considers the data distribution but also leverages prior knowledge, thus providing relatively accurate classification results in many cases.

[0153] Specifically, by constructing a Bayesian discriminant model, the corresponding individuals, teams or district teams are divided into the clusters in the clustering results to which they belong, so as to facilitate the subsequent acquisition of unsafe behavior portraits of all coal miners.

[0154] In some embodiments, the average value of the unsafe behavior risk index in the cluster to which the individuals, teams, or teams of all coal miners belong is used as the first unsafe behavior profile;

[0155] Based on the first unsafe behavior portrait, safety education plans and control measures are formulated for corresponding individuals, teams or district teams.

[0156] In this embodiment, after dividing all individuals, teams or districts among all coal miners into their corresponding clusters through the Bayesian discriminant model, the average of the accumulated unsafe behavior risk indicators of individuals, teams or districts is used as the first unsafe behavior portrait. Through the first unsafe behavior portrait, safety education plans and control measures can be formulated for the corresponding individuals, teams or districts for the unsafe behaviors that have occurred.

[0157] In some embodiments, the unsafe behaviors of individuals, teams, or divisions among all coal miners are predicted based on the clustering results, and corresponding preventive safety education plans and preventive control measures are formulated, including:

[0158] The basic risk indicators and unsafe behavior risk indicators of individuals, teams or districts in the T-1 period of all coal miners are input into the Bayesian discriminant model to obtain the first type of clusters of individuals, teams or districts in the T-1 period of all coal miners;

[0159] The basic risk index of period T and the unsafe behavior risk index of period T-1 of all coal miners are input into the Bayesian discriminant model to obtain the second type of cluster of period T of all coal miners, groups or teams;

[0160] In response to the fact that the first cluster and the second cluster are not the same cluster, the mean of the unsafe behavior risk indicators of the individuals, teams or districts in the second cluster during period T is used as the second unsafe behavior profile;

[0161] Based on the second unsafe behavior portrait, formulate corresponding preventive safety education plans and preventive control measures.

[0162] This embodiment describes how to predict the unsafe behavior trends of coal miners using a Bayesian discriminant model. First, the basic risk indicators and unsafe behavior risk indicators for the individual, team, or team during period T-1 (i.e., the previous period) are introduced into the Bayesian discriminant model to obtain the first cluster to which they belong during period T-1. Then, the basic risk indicators and unsafe behavior risk indicators for the individual, team, or team during period T (i.e., the current period) are introduced into the Bayesian discriminant model to obtain the second cluster to which they belong during period T. Finally, the first cluster and the second cluster are compared. If the first and second clusters are not the same cluster, the mean of the unsafe behavior risk indicators for period T in the second cluster is used as the second unsafe behavior profile of the individual, team, or team. Based on this second unsafe behavior profile, a corresponding preventive safety education plan and preventive control measures are formulated.

[0163] Specifically, when the first cluster and the second cluster are not the same cluster, it means that the unsafe behavior risk indicators of the individual, team or district team will change because of the change in their basic risk indicators. Therefore, the average of their unsafe behavior indicators in period T is taken as the second unsafe behavior portrait, as well as their unsafe behavior portrait in the future, and targeted preventive safety education plans and preventive control measures are formulated.

[0164] Specifically, by using the above method to predict the unsafe behavior trends of individuals, work groups or teams, targeted education plans and control measures can be formulated based on the prediction results, thereby reducing the accident rate of underground coal mine operations. At the same time, targeted education plans and control measures can also save a lot of manpower and material resources.

[0165] In some embodiments, in response to the first cluster and the second cluster being the same cluster, individuals, teams or divisions of all coal miners continue to use the safety education plan and control measures.

[0166] In this embodiment, if the first cluster and the second cluster are the same cluster, it means that the unsafe behavior profile of the individual, team or district will not change significantly at time T, and the existing safety education plan and control measures can be continued.

[0167] Specifically, by continuing to use the previous safety education plans and control measures, a large amount of human and material resources can be saved while ensuring a low accident rate.

[0168] In the second aspect, the present invention proposes a worker unsafe behavior trend prediction system based on dual indicator clustering, which includes the following modules:

[0169] Sampling module: used to implement stratified sampling of all coal miners;

[0170] Quantification module: used to construct basic risk indicators based on the individual basic indicators, safety education indicators, work situation indicators and health and physiological indicators of the sampled workers, and to construct unsafe behavior risk indicators based on the individual unsafe behavior events of the sampled workers. The individual unsafe behavior events are obtained based on the accident-causing dimension, awareness dimension and overall dimension;

[0171] Clustering module: used to implement kmeans clustering of individuals, teams or district teams among all coal miners based on the basic risk indicators and unsafe behavior risk indicators;

[0172] Discrimination module: used to build a Bayesian discriminant model, and cluster the individuals, teams or districts among all coal mine workers based on the Bayesian discriminant model, predict the unsafe behaviors of individuals, teams or districts among all coal mine workers based on the results of the cluster division, and formulate corresponding preventive safety education plans and preventive control measures.

[0173] In summary, the technical solution of the present invention can predict unsafe behavior trends for individuals, teams, or teams based on coal miners' basic risk indicators and unsafe behavior risk indicators. Appropriate safety education programs and control measures can be selected based on the trend prediction results, thereby reducing underground operation risks caused by coal miners' own factors. Targeted education programs and control measures can also significantly save manpower and material costs.

[0174] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations that come within the meaning and range of equivalents of the claims be embraced therein.

[0175] In addition, it should be understood that although this specification is described according to embodiments, not every embodiment contains only one independent technical solution. This narrative method of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.

Claims

1. A worker unsafe behavior trend prediction method based on dual indicator clustering, characterized by: include: Stratified sampling was conducted on all coal miners; Construct basic risk indicators based on the sampled workers' personal basic indicators, safety education indicators, work situation indicators and health and physiological indicators; Constructing an unsafe behavior risk index based on individual unsafe behavior events of sampled workers, wherein the individual unsafe behavior events are obtained based on the accident-causing dimension, the awareness dimension, and the overall dimension; Perform kmeans clustering on individuals, teams or district teams among all coal miners based on the basic risk indicators and unsafe behavior risk indicators; Construct a Bayesian discriminant model and classify all coal miners into clusters based on the Bayesian discriminant model; Based on the clustering results, unsafe behaviors of individuals, teams, or divisions among all coal mine workers are predicted, and corresponding preventive safety education plans and preventive control measures are formulated; The kmeans clustering of the basic risk indicators and unsafe behavior risk indicators of individuals, teams or teams among all coal miners includes: Quantify the basic risk indicators and unsafe behavior risk indicators of individuals, teams or teams among all coal mine workers; Initialize the k value to 2 and cluster the quantified basic risk indicators and unsafe behavior risk indicators; After completing the clustering of the initialized k value, increase the k value from the initial value and cluster the quantified basic risk indicators and unsafe behavior risk indicators again; After each clustering is completed, the sum of squares of the data errors within each cluster is calculated; Draw a line graph about k based on the sum of squared errors of the data for each cluster; Determine the optimal k value using the elbow method according to the line graph; The clustering result corresponding to the optimal k value is selected as the clustering result of kmeans clustering; Where k is the number of clusters; The formula for calculating the sum of square errors of data within each cluster is: ； Where SSE is the sum of squared errors, m is the total number of samples, is the i-th sample, is the center point of the j-th cluster, is the indicator function, It is a sample With cluster center The square of the Euclidean distance between belong hour, The value of is 1, otherwise it is 0; The Bayesian discriminant model is constructed, and clustering of individuals, teams, or teams among all coal miners is performed based on the Bayesian discriminant model, including: Based on each cluster of the clustering results, calculate the prior probability of each cluster; Based on each cluster of the clustering results, estimate the conditional probability density function of each cluster; According to Bayes' theorem, the posterior probability of an individual, team, or district team among all coal miners belonging to each cluster in the clustering result is calculated as follows: ； in, is the posterior probability, is the prior probability, is the conditional probability density function, is the marginal probability of sample x; All individuals, teams or district teams among all coal miners are assigned to the cluster with the largest posterior probability in the clustering results.

2. The method for predicting worker unsafe behavior trends based on dual indicator clustering according to claim 1, characterized in that: The personal basic indicators specifically include length of service, educational background and professional skill level; The safety education indicators include safety education participation and safety education assessment scores; The work situation indicators include attendance, social exchange relationship I and social exchange relationship O; The health physiological indicators include cardiovascular health, basal metabolic health, fatigue level, heart rate abnormality and body temperature abnormality; The social exchange relationship I represents the evaluation results of all other workers on a certain worker i; The social exchange relationship O represents the evaluation result of a certain worker i on all other workers.

3. The method for predicting worker unsafe behavior trends based on dual indicator clustering according to claim 1, characterized in that: The accident-causing dimensions include: operating errors / ignoring safety / ignoring warnings, causing safety device failure, using unsafe equipment, operating with hands instead of tools, improper storage of objects, venturing into dangerous places, climbing in unsafe positions, working / staying under the boom, refueling / repairing / inspecting the machine while it is running, distracting behavior, not using personal protective equipment / equipment correctly, unsafe clothing, and incorrect handling of flammable / explosive dangerous goods; The awareness dimension includes conscious unsafe behaviors and unconscious unsafe behaviors; The overall dimension is the number of unsafe behaviors.

4. The method for predicting worker unsafe behavior trends based on dual indicator clustering according to claim 1, characterized in that: The mean of the unsafe behavior risk index of the cluster to which individuals, teams or teams belong among all coal miners is used as the first unsafe behavior profile; Based on the first unsafe behavior portrait, safety education plans and control measures are formulated for corresponding individuals, teams or district teams.

5. The method for predicting worker unsafe behavior trends based on dual indicator clustering according to claim 1, characterized in that: The unsafe behaviors of individuals, teams, or divisions among all coal miners are predicted based on the cluster division results, and corresponding preventive safety education plans and preventive control measures are formulated, including: The basic risk indicators and unsafe behavior risk indicators of individuals, teams or districts in the T-1 period of all coal miners are input into the Bayesian discriminant model to obtain the first type of clusters of individuals, teams or districts in the T-1 period of all coal miners; The basic risk index of period T and the unsafe behavior risk index of period T-1 of all coal miners are input into the Bayesian discriminant model to obtain the second type of cluster of period T of all coal miners, groups or teams; In response to the fact that the first cluster and the second cluster are not the same cluster, the mean of the unsafe behavior risk indicators of the individuals, teams or districts in the second cluster during period T is used as the second unsafe behavior profile; Based on the second unsafe behavior portrait, formulate corresponding preventive safety education plans and preventive control measures.

6. The method for predicting worker unsafe behavior trends based on dual index clustering according to claim 5, characterized in that: In response to the first cluster and the second cluster being the same cluster, the individuals, teams or district teams of all coal mine workers continue to use the safety education plan and control measures.

7. The worker unsafe behavior trend prediction system based on dual index clustering is characterized by: The system applies the worker unsafe behavior trend prediction method based on dual indicator clustering as described in any one of claims 1 to 6, and includes the following modules: Sampling module: used to implement stratified sampling of all coal miners; Quantification module: used to construct basic risk indicators based on the individual basic indicators, safety education indicators, work situation indicators and health and physiological indicators of the sampled workers, and to construct unsafe behavior risk indicators based on the individual unsafe behavior events of the sampled workers. The individual unsafe behavior events are obtained based on the accident-causing dimension, awareness dimension and overall dimension; Clustering module: used to implement kmeans clustering of individuals, teams or district teams among all coal miners based on the basic risk indicators and unsafe behavior risk indicators; Discrimination module: used to build a Bayesian discriminant model, and cluster the individuals, teams or districts among all coal mine workers based on the Bayesian discriminant model, predict the unsafe behaviors of individuals, teams or districts among all coal mine workers based on the results of the cluster division, and formulate corresponding preventive safety education plans and preventive control measures.

Images