A dual-drive prediction method for the spatiotemporal characteristics of typical coal mine disasters

Abstract

This application proposes a dual-driven prediction method for the spatiotemporal characteristics of typical coal mine disasters. The method includes: acquiring historically monitored first disaster data and a pre-constructed expert knowledge base; extracting the first temporal features corresponding to the first disaster data based on the corresponding time information; performing statistical analysis on the first disaster data to determine its spatial features; and generating a disaster prediction model based on the first temporal features, spatial features, and the expert knowledge base. This allows the model to achieve dual-driven prediction using both knowledge and data, improving the accuracy and reliability of disaster prediction. It enables timely identification and early warning of typical coal mine disaster risks at the production site, thereby improving the safety and production efficiency of coal mine scenarios.

Description

Technical Field

[0001] This application relates to the field of coal mine safety production technology, and in particular to a dual-drive prediction method for the spatiotemporal characteristics of typical coal mine disasters. Background Technology

[0002] As a critical energy industry, coal mine safety has always been a focus of industry attention. With the deepening of mining activities, especially the increase in mining depth, geological conditions become more complex, thus inducing a series of dynamic disasters, such as gas explosions, coal dust explosions, and roof collapses, making disaster prediction more difficult.

[0003] Traditional single-dimensional prediction methods struggle to capture the nonlinear evolution of time series and the heterogeneous distribution of spatial fields during disaster gestation, limiting prediction accuracy. Therefore, by integrating the temporal dynamics and spatial correlation characteristics of multi-source monitoring data, a spatiotemporal dual-driven model of the disaster gestation process can be achieved. This effectively reveals the intrinsic mechanisms and critical conditions of disaster occurrence, significantly improving the accuracy and timeliness of predictions for typical coal mine disasters, and providing a scientific basis for early warning of coal mine disasters. Summary of the Invention

[0004] This application aims to at least partially address one of the technical problems in the related art.

[0005] Therefore, this application proposes a dual-driven prediction method for the spatiotemporal characteristics of typical coal mine disasters, including:

[0006] Acquire historical data on the first disasters detected, as well as a pre-built expert knowledge base;

[0007] Based on the time information corresponding to the first disaster data, extract the first time feature corresponding to the first disaster data;

[0008] Statistical analysis is performed on the first disaster data to determine the spatial characteristics corresponding to the first disaster data;

[0009] A disaster prediction model is generated based on the first temporal feature, the spatial feature, and the expert knowledge base.

[0010] In some embodiments, extracting the first time feature corresponding to the first disaster data based on the time information corresponding to the first disaster data includes:

[0011] Based on the time information corresponding to the first disaster data, the first disaster data is clustered to determine at least one dataset;

[0012] Perform time series analysis on the second disaster data in each dataset to determine the second time characteristics corresponding to the dataset;

[0013] Based on the second time features corresponding to all the datasets, the first time features corresponding to the first disaster data are determined.

[0014] In some embodiments, performing time series analysis on the second disaster data in each dataset to determine the second temporal characteristics corresponding to the dataset includes:

[0015] Based on the dispersion of the second disaster data and the corresponding time information, determine the associated time period of the second disaster data and its corresponding association strength;

[0016] Calculate the local anomaly factor corresponding to each of the second disaster data to identify anomalies in the second disaster data;

[0017] The second temporal feature corresponding to the dataset is obtained from the associated time period, the associated strength corresponding to the associated time period, and the outlier.

[0018] In some embodiments, calculating the local anomaly factor corresponding to each of the second disaster data to identify anomalies in the second disaster data includes:

[0019] Determine the first distance between each of the second disaster data and other second disaster data;

[0020] Based on a preset value of k and all first distances corresponding to the second disaster data, the neighborhood and second distance corresponding to the second disaster data are determined, wherein the neighborhood contains at least k third disaster data, and k is a positive integer;

[0021] Based on the second distance and the first distances corresponding to all the third disaster data, the first local reachability density corresponding to the second disaster data is determined;

[0022] Based on the first local reachability density and the second local reachability density corresponding to all the third disaster data, determine the local anomaly factor corresponding to the second disaster data;

[0023] If the local anomaly factor is greater than the first threshold, the second disaster data is identified as an anomaly point, and the operation of determining the area and second distance corresponding to the next second disaster data is returned until all anomalies in the second disaster data are identified.

[0024] In some embodiments, determining the neighborhood and second distance corresponding to the second disaster data based on a preset k value and all first distances corresponding to the second disaster data includes:

[0025] Sort all the first distances, and determine the kth first distance in the sorting results as the second distance;

[0026] All other second disaster data whose corresponding first distance is less than or equal to the second distance are identified as third disaster data;

[0027] Based on the third disaster data and the second distance, the neighborhood corresponding to the second disaster data is obtained.

[0028] In some embodiments, the step of performing statistical analysis on the first disaster data to determine the spatial characteristics corresponding to the first disaster data includes:

[0029] Based on the environmental information in the first disaster data, at least one region where the first disaster data is distributed and the relationship between the at least one region are determined;

[0030] Based on the size of the first disaster data corresponding to each region, the target region among the at least one region is determined;

[0031] The spatial characteristics corresponding to the first disaster data are determined based on the relationships between the at least one region and the target region.

[0032] In some embodiments, determining at least one region of the distribution of the historical disaster data and the relationship between the at least one region based on environmental information in the first sample disaster data includes:

[0033] The environmental information in the first disaster data is analyzed to determine at least one area where the first disaster data is distributed;

[0034] Based on the third distance between every two regions in the at least one region, a spatial weight matrix corresponding to the first disaster data is generated;

[0035] Based on the spatial weight matrix and the first disaster data corresponding to each region, the Moran index corresponding to the at least one region is obtained;

[0036] Based on the Moran index, the first test value is obtained;

[0037] If the first test value is greater than the second threshold, the relationship between the at least one region is determined based on the sign of the Moran index.

[0038] In some embodiments, determining the target area among the at least one area based on the size of the first disaster data corresponding to each area includes:

[0039] Based on the spatial weight matrix and the first disaster data corresponding to each region, determine the statistics corresponding to each region;

[0040] Based on the statistic, the second test value corresponding to the region is obtained;

[0041] If the second test value is greater than the third threshold, the region is identified as a target region.

[0042] In some embodiments, generating a disaster prediction model based on the first temporal feature, the spatial feature, and the expert knowledge base includes:

[0043] Based on the first temporal feature and the spatial feature, a data model is constructed;

[0044] Based on the aforementioned expert knowledge base, a knowledge model is constructed;

[0045] The data model and the knowledge model are fused to generate a disaster prediction model driven by both data and knowledge.

[0046] In some embodiments, after generating the disaster prediction model, the method further includes:

[0047] Retrieve data to be updated within a preset time interval;

[0048] The disaster prediction model is updated based on the data to be updated.

[0049] In some embodiments, after updating the disaster prediction model, the method further includes:

[0050] Determine the first score corresponding to the updated disaster prediction model;

[0051] If the difference between the first score and the second score corresponding to the disaster prediction model before the update is less than the fourth threshold, the parameters of the updated disaster prediction model are adjusted, and the first score is recalculated until the first score and the second score are greater than the fourth threshold, thus completing the update of the disaster prediction model.

[0052] The spatiotemporal characteristic dual-driven prediction method for typical coal mine disasters provided in this application first acquires historical disaster monitoring data and a pre-built expert knowledge base, extracts temporal and spatial features from the disaster data respectively, and then integrates the temporal features, spatial features and expert knowledge base to generate a disaster prediction model for predicting typical disasters in coal mine scenarios. The generated model can achieve dual-driven prediction of disasters using knowledge and data, improving the accuracy and reliability of disaster prediction, ensuring the timeliness of disaster risk warning, and contributing to improving the safety and production efficiency of coal mine scenarios.

[0053] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0054] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0055] Figure 1 A flowchart illustrating a spatiotemporal feature dual-drive prediction method provided in an embodiment of this application;

[0056] Figure 2 A flowchart illustrating another spatiotemporal feature dual-drive prediction method provided in an embodiment of this application;

[0057] Figure 3 A flowchart illustrating another spatiotemporal feature dual-drive prediction method provided in an embodiment of this application;

[0058] Figure 4 This is a flowchart illustrating another spatiotemporal feature dual-drive prediction method provided in an embodiment of this application. Detailed Implementation

[0059] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0060] The following describes a dual-drive prediction method for the spatiotemporal characteristics of typical coal mine disasters according to embodiments of this application, with reference to the accompanying drawings.

[0061] In existing technologies, the prediction of typical coal mine disasters mainly employs three strategies: traditional methods based on expert experience, statistical methods based on historical data, and methods utilizing big data technology. Each of these strategies has its shortcomings. Traditional methods based on expert experience are highly subjective and limited by individual experience and knowledge. Statistical methods based on historical data can only provide probabilistic predictions and cannot offer definitive conclusions, requiring high-quality and high-quantity data. Methods utilizing big data technology and algorithms can rapidly process and analyze large amounts of data, uncovering potential patterns in disaster occurrence; however, their effectiveness depends heavily on data quality and algorithm accuracy, and the algorithms have relatively weak interpretability. Furthermore, all three strategies focus on single factors, lacking comprehensive analysis and early warning capabilities for multiple factors, making it difficult to accurately capture the spatiotemporal evolution characteristics of disasters, resulting in insufficient accuracy and timeliness of early warning information.

[0062] To address this issue, this application provides a dual-driven prediction method for the spatiotemporal characteristics of typical coal mine disasters. This method combines expert experience with big data algorithms to achieve dual-driven prediction based on knowledge and data, resulting in a more comprehensive and accurate disaster prediction model and improving the accuracy and reliability of disaster prediction.

[0063] In this embodiment of the application, typical coal mine disasters can refer to geological disasters such as surface subsidence and mine water inrush, or to machine or man-made disasters such as equipment failure, etc. Since the causes of disasters are usually different, it is necessary to analyze the temporal and spatial characteristics of the disaster based on multi-source historical data monitored before and after the disaster.

[0064] Figure 1 This is a flowchart illustrating a dual-drive prediction method for the spatiotemporal characteristics of typical coal mine disasters provided in an embodiment of this application.

[0065] like Figure 1 As shown, the spatiotemporal characteristic dual-drive prediction method for typical coal mine disasters may include the following steps:

[0066] Step 101: Obtain historically monitored first disaster data and a pre-built expert knowledge base.

[0067] In this embodiment, the first disaster data may include data obtained from monitoring typical disasters in multiple coal mine scenarios. The first disaster data obtained for each disaster may be an integration of multi-source data, which may include at least one of the following: geological data such as coal seam thickness, dip angle, and roof lithology at the disaster location; environmental monitoring data such as gas concentration, humidity, and temperature; operating status data such as speed, current, and voltage of key equipment such as coal mining machines and conveyor belts; and real-time status information of the internal environment such as microseismic monitoring, stress-strain monitoring, and fiber optic temperature measurement.

[0068] In this embodiment, if the amount of data corresponding to the first disaster data is large enough, then by analyzing the first disaster data, the spatiotemporal characteristics of different disasters can be summarized, and the occurrence patterns of disasters can be obtained for disaster prediction. The first disaster data can be selected according to the actual disaster prediction needs, and the corresponding collection time period and geographical location can be selected. This application does not limit this. For example, if it is necessary to predict the disaster situation in a certain region in summer, when obtaining the first disaster data, only the data corresponding to disasters that occurred in the historical summers of that region can be obtained.

[0069] In this embodiment, expert knowledge and experience related to disasters in the currently predicted business scenario (such as equipment failure) can be collected and organized in advance to construct an expert knowledge base. For example, the constructed expert knowledge base can cover knowledge data such as equipment operation and maintenance records, historical coal mine accident records, disaster formation mechanisms, prevention measures, risk assessment standards, emergency strategies, as well as industry standards, specifications, and empirical formulas.

[0070] Step 102: Extract the first time feature corresponding to the first disaster data based on the time information corresponding to the first disaster data.

[0071] The time information corresponding to the first disaster data refers to the time point when the historical disaster occurred. In this application, the time point can also be converted into a suitable numerical representation as needed, such as the number of days between the time of the disaster and the reference time. The reference time can be determined according to specific forecasting needs.

[0072] The first-time characteristics may include at least one of the following: the periodicity of disaster occurrence, the trend of disaster frequency and intensity, and the abnormal patterns or abrupt changes in disaster occurrence.

[0073] It should be noted that anomaly patterns refer to the phenomenon where a group of data points deviate from the normal distribution or pattern. Outliers are individual data points that are significantly different from other observations in the dataset. Identifying anomalous patterns or outliers in disaster data can reflect potential disaster mechanisms. By using them as temporal features of disaster data, the extracted temporal features can be used in model building to more accurately capture precursory signals of disasters and improve the accuracy of prediction models.

[0074] In this embodiment, time series analysis can be performed on the first disaster data using the time information corresponding to the first disaster data to identify abnormal patterns or abrupt changes in the data. This allows the model to learn the potential patterns and characteristics of disaster occurrence, thereby improving the accuracy of predicting future disaster events. Furthermore, machine learning algorithms can be applied to perform periodic analysis on the time information corresponding to the first disaster data to predict the frequency and intensity trends of disaster occurrences. By combining the frequency and intensity trends of disaster occurrences with the identified abnormal patterns or abrupt changes, the first time characteristics corresponding to the first disaster data can be obtained, reflecting the temporal patterns of typical coal mine disasters corresponding to the first disaster data.

[0075] In this embodiment, since different types of time features may vary significantly, in order to improve the accuracy and efficiency of time feature extraction and avoid excessively high feature dimensionality and noise interference in the feature extraction results, the first disaster data can be classified first, grouping disasters with similar temporal characteristics and similar occurrence patterns into the same category. Then, for different types of disasters, their patterns in the time dimension are identified separately. Finally, the set of time features corresponding to all types of disasters is determined as the first time feature corresponding to the first disaster data.

[0076] Step 103: Perform statistical analysis on the first disaster data to determine the spatial characteristics corresponding to the first disaster data.

[0077] Spatial characteristics may include at least one of the following: the distribution characteristics of disasters in different regions, the interaction between disasters in different regions, and regions with high frequency of disaster occurrence.

[0078] In this embodiment, the geological data and environmental monitoring data included in the first disaster data can be used to determine the characteristics of the disaster-occurring area. Therefore, statistical analysis of various monitoring data in the first disaster data can be performed first to summarize the distribution characteristics of the disaster in different regions. Then, Moran's I index is used to calculate the spatial autocorrelation of the first disaster data, i.e., to determine the interaction between disasters in different regions. Furthermore, a hotspot area identification algorithm can be used to identify areas with abnormally high disaster occurrence frequencies. Subsequently, the spatial characteristics corresponding to the first disaster data can be obtained from one or more of the following: the distribution characteristics of the disaster in different regions, the interaction between disasters in different regions, and areas with high disaster occurrence frequencies.

[0079] Step 104: Generate a disaster prediction model based on the first time features, spatial features, and expert knowledge base.

[0080] In this embodiment, a disaster prediction data model can be constructed based on the extracted temporal and spatial features to predict the frequency and intensity of disasters. Furthermore, a knowledge model can be constructed by combining knowledge and experience from an expert knowledge base. The knowledge model can correct and optimize the results of the data model. For example, when the data model predicts a risk of mining-induced disasters in a certain area, relevant information such as the geological structure and mining activities of that area can be retrieved from the expert knowledge base. A deeper understanding of the disaster's mechanism and influencing factors provides strong support for disaster prediction. The data model and the knowledge model can be combined to generate a disaster prediction model driven by both data and knowledge.

[0081] In this embodiment, the data model provides quantitative predictions of disaster frequency and intensity, while the knowledge model provides qualitative explanations and analyses of disaster occurrence mechanisms and influencing factors. When the disaster prediction model predicts that the probability of a disaster is greater than a certain value, it can issue an early warning signal in a timely manner to notify relevant personnel to take preventive measures, thereby improving the accuracy and reliability of disaster prediction.

[0082] In this embodiment, by first acquiring data from historical disaster monitoring and a pre-built expert knowledge base, time and spatial features are extracted from the disaster data, and then the time features, spatial features, and expert knowledge base are integrated to generate a disaster prediction model for predicting typical disasters in coal mine scenarios. The generated model can achieve dual-driven disaster prediction based on knowledge and data, improving the accuracy and reliability of disaster prediction, ensuring the timeliness of disaster risk warning, and helping to improve the safety and production efficiency of coal mine scenarios.

[0083] Because this application employs clustering to divide the first disaster data into different datasets, and then analyzes the temporal patterns of disaster data in each dataset to improve the reliability of temporal feature extraction, this embodiment also provides another method for predicting typical coal mine disasters using a dual-drive approach of spatiotemporal features. Figure 2 This is a flowchart illustrating another method for predicting the spatiotemporal characteristics of typical coal mine disasters, provided in an embodiment of this application.

[0084] like Figure 2 As shown, the spatiotemporal characteristics dual-driven prediction method for typical coal mine disasters can include the following steps:

[0085] Step 201: Obtain historically monitored first disaster data and a pre-built expert knowledge base.

[0086] For a detailed description of step 201 above, please refer to the description in other embodiments of this application, which will not be repeated here.

[0087] Step 202: Based on the time information corresponding to the first disaster data, cluster the first disaster data to determine at least one dataset.

[0088] In this embodiment of the application, the time information corresponding to the first disaster data can be used as a data sample, and then a clustering algorithm (such as K-Means clustering algorithm) can be used to cluster these data samples, so as to classify disasters with similar characteristics and similar occurrence patterns in time into the same cluster, and each cluster is a dataset.

[0089] For example, taking the K-Means algorithm as an example, the objective function for determining the clustering result is shown in the following formula (1). By iteratively optimizing, the J value obtained by formula (1) can be minimized, and the final clustering result can be determined, thus obtaining the data set corresponding to each cluster.

[0090]

[0091] In formula (1), K is the number of clusters set according to the actual situation, and C k Let x represent the first disaster data set belonging to the k-th cluster. i It is the time point corresponding to the i-th first disaster data in the k-th cluster. μ k It is the centroid of the k-th cluster, that is, the mean time point corresponding to all the first disaster data in this cluster.

[0092] Step 203: Perform time series analysis on the second disaster data in each dataset to determine the second time characteristics corresponding to the dataset.

[0093] The second disaster data refers to the same type of monitoring data extracted from all the first disaster data belonging to the same dataset, such as vertical stress or horizontal stress. In this application, time series analysis can be performed on various types of second disaster data separately, thereby determining the second time characteristics from multiple perspectives and improving the reliability of the time characteristics.

[0094] The second time feature refers to the temporal pattern that a type of disaster corresponds to in each dataset. It may include at least one of the following: the periodicity of disaster occurrence, the trend of disaster frequency and intensity, and the identified abnormal patterns or abrupt changes in disaster occurrence. The first time feature is a set of the second time features.

[0095] In this embodiment of the application, after determining at least one dataset in the clustering results, second disaster data of the same type, such as all vertical stress monitoring data, can be obtained from all first disaster data corresponding to each dataset. Then, machine learning algorithms can be applied to perform time series analysis on the obtained second disaster data to reveal whether disaster events follow a certain periodic pattern, capture the long-term trend of the frequency or intensity of disasters, and identify abnormal patterns of disasters, thereby obtaining the second time features corresponding to the data.

[0096] For example, taking mine pressure data as the second hazard data, periodic analysis of the mine pressure data can reveal significant fluctuations in mine pressure during specific time periods (such as when a coal face advances to the vicinity of a fault or during blasting operations), indicating that the magnitude of mine pressure is affected by factors such as geological structural changes and mining activities. Then, by calculating the variance or standard deviation of the second hazard data, the correlation strength between mine pressure and specific time periods can be analyzed, and the analysis results can be stored in an expert knowledge base. Furthermore, trend analysis can be used to observe the changing trend of mine pressure over time. If an upward trend is observed, it indicates a potential mine pressure risk, requiring further analysis of the rate and intensity of the increase. Anomalies in mine pressure can also be monitored using a local anomaly factor algorithm. By comparing the density differences between a data point in the second hazard data and other data points in its neighborhood, anomalies can be determined. Finally, the specific time periods showing significant fluctuations in mine pressure, the correlation strength between mine pressure and specific time periods, and the anomalies in mine pressure can be identified as the second temporal characteristics corresponding to the mine pressure data.

[0097] In some possible embodiments, the associated time period and its corresponding association strength of the second disaster data can be determined first based on the degree of dispersion of the second disaster data and the corresponding time information.

[0098] In this embodiment, the second disaster data can be arranged in chronological order, and trend analysis can be used to observe the changing trend of the second disaster data over time. It can be determined that the value of the second disaster data fluctuates significantly or has a high rate of increase during certain time periods. This indicates that typical coal mine disasters are likely to occur during these time periods, and these time periods can be identified as the associated time periods of the second disaster data.

[0099] It should be noted that the associated time period for the second disaster data may be a specific time interval, or it may be during the process of certain actions, such as the coal mining face advancing to the vicinity of the fault or blasting operations.

[0100] In this embodiment of the application, the dispersion of the second disaster data can be determined by calculating the variance or standard deviation of the second disaster data. The higher the dispersion, the lower the correlation between the data and time. For example, the variance calculation formula used to measure the dispersion of the second disaster data can be shown in the following formula (2).

[0101]

[0102] Where, x i is the i-th second disaster data point in the dataset, μ is the mean of all second disaster data points in the dataset, and N is the size of the dataset, i.e., the number of data points for the second disaster data.

[0103] Alternatively, the average root can be calculated from the result obtained by formula (2) to obtain the standard deviation, which can be used to describe the dispersion of the second disaster data distribution.

[0104] Then, the local anomaly factor corresponding to each second disaster data point can be calculated to identify outliers in the second disaster data.

[0105] Among them, the local anomaly factor is used to describe the difference between the data density around a certain data point (i.e., the second disaster data) and the data density of other neighboring data points.

[0106] In this embodiment, a local anomaly factor algorithm can be used to sequentially calculate the k-nearest neighbor distance, k-distance neighborhood, reachability distance, and local reachability density of each second disaster data, thereby obtaining the local anomaly factor of the second disaster data. Then, based on the magnitude of the local anomaly factor corresponding to each second disaster data, it can be determined which second disaster data are anomalies. For example, all second disaster data with a local anomaly factor greater than a certain value can be identified as anomalies.

[0107] Optionally, a first distance can be determined between each second hazard data and other second hazard data.

[0108] For example, a database contains three data points related to a second disaster, labeled as Second Disaster Data 1, Second Disaster Data 2, and Second Disaster Data 3, with values ​​of 'a' for Second Disaster Data 1, 'b' for Second Disaster Data 2, and 'c' for Second Disaster Data 3. Then, we need to calculate the first distance between Second Disaster Data 1 and Second Disaster Data 2 (the absolute value of a minus b), the first distance between Second Disaster Data 1 and Second Disaster Data 3 (the absolute value of a minus c), and the first distance between Second Disaster Data 2 and Second Disaster Data 3 (the absolute value of b minus c).

[0109] Then, based on a preset k value and all first distances corresponding to a second disaster data, the neighborhood and second distance corresponding to the second disaster data can be determined.

[0110] Where k is a positive integer, and the value of k can be configured reasonably as needed. Generally speaking, the larger the value of k, the more stable the algorithm's estimation of local density, but it may ignore some small local outliers; the smaller the value of k, the more sensitive the algorithm is to changes in local density, but it may be affected by noise.

[0111] The second distance is the k-distance of a second disaster data point, which is the radius of the neighborhood centered on the second disaster data point.

[0112] In this embodiment, all first distances corresponding to a second disaster data point can be sorted. The k-th first distance from the smallest to the largest in the sorting result is determined as the second distance, which is the neighborhood radius and k-nearest neighbor distance of the second disaster data point. This second disaster data point is the neighborhood center. Then, all other second disaster data points, excluding the neighborhood center, whose corresponding first distances are less than or equal to the second distances, can be determined as third disaster data points. Based on the determined neighborhood center, second distance (neighborhood radius), and third disaster data points, the neighborhood corresponding to the second disaster data point can be determined.

[0113] It is understandable that the second distance is the minimum distance value that satisfies the following two conditions: (The condition is missing from the original text, so the translation is incomplete.) i Centered on a radius of d k (x i The neighborhood of ) contains at least k third-hazard data (excluding x). i (itself); and, for any less than d k (x i The distance value r of x i The number of third disaster data points contained within a neighborhood of radius r centered at k is less than k.

[0114] Then, based on the second distance and the first distance corresponding to all third hazard data respectively, the first local reachability density corresponding to the second hazard data can be determined.

[0115] In this embodiment of the application, the first distances corresponding to the second distance and all third disaster data can be substituted into the following formula (3) to calculate the reachable distance of each third disaster data relative to the second disaster data of the neighborhood center.

[0116] reach k (x i ,x j )=max{d k (x i ),||x i -x j ||} (3)

[0117] In formula (3), x i This refers to the secondary disaster data, d, which serves as the center of the neighborhood. k (x i This represents the second distance (territory radius) corresponding to the second disaster data, x. j For the j-th third disaster data, ||x i -x j || is represented as x i With x j The first distance between them. Then, according to formula (3), xj Relative to x i The reachable distance is the sum of the radius of the domain and x. i With x j The maximum value of the first distance between.

[0118] In this embodiment of the application, after determining the reachability distance between each third disaster data and the second disaster data, the second disaster data x can be calculated using the following formula (4). i The corresponding first locally reachable density.

[0119]

[0120] In formula (4), N k (x i ) indicates that the second disaster data x i The k-neighborhood of the center, x j For the j-th third hazard data in the k-neighborhood, LRD k (x i This refers to the second disaster data x. i The corresponding first locally reachable density.

[0121] It should be noted that the above operations of determining the neighborhood, domain radius, reachability distance, and local reachability density can be repeated to calculate the local reachability density corresponding to each third hazard data point within the neighborhood, with the neighborhood center as the corresponding value.

[0122] Then, based on the first local reachability density and the second local reachability density corresponding to all third hazard data, the local anomaly factor corresponding to the second hazard data can be determined.

[0123] The second local reachability density refers to the value obtained by repeating the above operations of determining the neighborhood, domain radius, reachability distance, and local reachability density when the third disaster data in each neighborhood is used as the neighborhood center.

[0124] In this embodiment of the application, the formula for calculating the local anomaly factor of the second disaster data can be shown in the following formula (5):

[0125]

[0126] In formula (5), LRD k (x j ) represents the j-th third disaster data point x within the neighborhood. j The second locally reachable density, LOF k (x i ) represents the second disaster data x i The corresponding local anomaly factor is the ratio of the local reachability density of a point to the local reachability density of other data points in its neighborhood.

[0127] Finally, if the local anomaly factor is greater than the first threshold, the second disaster data can be identified as an anomaly point, and the operation of determining the corresponding domain and second distance of the next second disaster data can be returned until all anomalies in the second disaster data are identified.

[0128] The first threshold is a critical value used to describe a large density difference between a data point in the neighborhood and other data points in the neighborhood. It can be a reasonable threshold set according to the actual data, for example, it can be 1.

[0129] In this embodiment, if the local anomaly factor of any second disaster data exceeds a first threshold, it can be determined that the data density around the second disaster data is significantly lower than the data density of other neighboring second disaster data, thus identifying the second disaster data as an anomaly. The process then returns to determining the next second disaster data's corresponding neighborhood and second distance, continuing until all anomalies in the second disaster data are identified.

[0130] Then, the associated time period, the associated strength of the associated time period, and the outliers can be used as the second time features of the dataset.

[0131] Step 204: Based on the second time features corresponding to all datasets, determine the first time features corresponding to the first disaster data.

[0132] In this embodiment, the disaster type corresponding to each dataset can be associated with its corresponding second time feature to determine the first time feature corresponding to the first disaster data. Therefore, by using clustering to divide the first disaster data into different disaster types, and then extracting time features for each disaster type, the accuracy and reliability of time feature extraction can be improved, providing reliable data conditions for the accurate prediction of typical coal mine disasters.

[0133] Step 205: Perform statistical analysis on the first disaster data to determine the spatial characteristics corresponding to the first disaster data.

[0134] Step 206: Generate a disaster prediction model based on the first time features, spatial features, and expert knowledge base.

[0135] In this embodiment of the application, after generating the disaster prediction model, the disaster prediction model can identify whether the data points monitored in real time in the coal mine environment are abnormal points. When it is determined that the data point monitored at a certain moment is an abnormal point, there is a high probability that a disaster will occur. At this time, the disaster prediction model can obtain relevant knowledge from the expert knowledge base, deeply analyze the cause of the anomaly, and formulate an effective response plan.

[0136] For a detailed description of steps 205 to 206 above, please refer to the description in other embodiments of this application, which will not be repeated here.

[0137] It should be noted that, in this application, the spatial features extracted from historical disaster data may include at least one of the following: the distribution characteristics of disasters in different regions, the interaction between disasters in different regions, and regions with high frequency of disaster occurrence. Therefore, different spatial features can be extracted sequentially from the first disaster data to ensure the accuracy of the spatial feature extraction results. Thus, this embodiment also provides another method for predicting typical coal mine disasters using a dual-drive method of spatiotemporal features. Figure 3 This is a flowchart illustrating another method for predicting the spatiotemporal characteristics of typical coal mine disasters, provided in an embodiment of this application.

[0138] like Figure 3 As shown, the method for generating this disaster prediction model may include the following steps:

[0139] Step 301: Obtain historically monitored first disaster data and a pre-built expert knowledge base.

[0140] Step 302: Extract the first time feature corresponding to the first disaster data based on the time information corresponding to the first disaster data.

[0141] For a detailed description of steps 301 and 302 above, please refer to the description in other embodiments of this application, which will not be repeated here.

[0142] Step 303: Based on the environmental information in the first disaster data, determine at least one region where the first disaster data is distributed and the relationship between at least one region.

[0143] Among them, environmental information refers to the information contained in the first disaster data that can describe the geographical location and environment of the disaster, such as geological data such as coal seam thickness, dip angle, and roof lithology, or environmental monitoring data such as temperature and humidity.

[0144] In this embodiment, due to differences in environmental conditions such as coal seam thickness, dip angle, roof lithology, temperature, and humidity, different areas are affected to varying degrees by factors such as machine operations, resulting in different probabilities of disasters occurring in those areas. Therefore, environmental information from historically monitored first disaster data can be analyzed to determine at least one area where historical disasters occurred and were distributed. Different areas can be divided based on criteria such as the range of coal seam thickness, or geographical location. After determining at least one area where the first disaster data is distributed, the relationship between areas can be determined based on the distance between any two areas; the closer the distance, the greater the impact of disasters between the areas.

[0145] Optionally, the environmental information in the first disaster data can be analyzed first to determine at least one region where the first disaster data is distributed. Then, a spatial weight matrix corresponding to the first disaster data can be generated based on the third distance between every two regions in the at least one region.

[0146] The spatial weight matrix describes the spatial adjacency relationship or degree of mutual influence between various spatial regions.

[0147] In this embodiment of the application, the spatial weight between any two regions can be represented by the following equation (6).

[0148]

[0149] In formula (6), d ij w represents the third distance between regions i and j. ij This represents the spatial weight between region i and region j.

[0150] In this embodiment of the application, after determining the spatial weights between all regions, a spatial weight matrix with a dimension of n×n can be constructed, where n is the number of regions. The larger the value of each element in the matrix, the closer the distance between the two regions represented by the corresponding row and column, and the higher the degree of mutual influence.

[0151] Then, based on the spatial weight matrix and the first disaster data corresponding to each region, at least one Moran index corresponding to a region can be obtained.

[0152] In this embodiment of the application, the formula for calculating the Moran index can be shown in the following formula (7).

[0153]

[0154] In formula (7), n is the number of spatial regions; x i and x j These are the first hazard data (such as mine pressure observation values) for regions i and j, respectively; It is the mean of the first disaster data across all regions, i.e. S 2 It is the variance of the first disaster data, calculated using the following formula:

[0155] Then, based on the Moran index, a first test value can be obtained. If the first test value is greater than the second threshold, at least one relationship between regions is determined according to the sign of the Moran index.

[0156] In this embodiment of the application, the first detection value is 3) statistical significance test, which can be the Z test based on the assumption of normal distribution. The calculation formula of the first detection value is shown in the following formula (8).

[0157]

[0158] In formula (8), Z is the first detection value, E(I) is the expected value of the Moran index obtained by formula (7) above, and VAR(I) is the variance of the Moran index.

[0159] In this embodiment, after obtaining the first test value, a critical value under the standard normal distribution can be determined as the second threshold based on a set significance level, typically 0.05. Then, the first test value is compared with the second threshold. If the first test value is greater than the second threshold, it indicates that Moran's I value is significant, and spatial autocorrelation does exist. Therefore, the spatial correlation characteristics of the first disaster data between regions can be determined based on the sign of Moran's I.

[0160] For example, taking mine pressure as the primary hazard data, if Morans's I index is positive, it indicates positive spatial autocorrelation, meaning that adjacent areas share commonalities in geological structure or mining activities. Mine pressure hazards in one area may affect other areas, so it is necessary to strengthen the monitoring and analysis of adjacent areas to prevent the spread of mine pressure hazards. If Morans's I index is negative, it indicates negative spatial autocorrelation, meaning that the mine pressure values ​​of adjacent areas differ significantly. Alternatively, if Morans's I index is 0, it indicates that there is no obvious spatial autocorrelation, and the mine pressures in each area are relatively independent.

[0161] Step 304: Determine the target area in at least one region based on the size of the first disaster data corresponding to each region.

[0162] The target area, also known as a hotspot or high-value cluster area, refers to an area where disasters occur frequently. The target area can be one or more of the primary disaster data distribution areas.

[0163] In this embodiment, a hotspot area identification algorithm can be used to determine areas with abnormally high levels of first-hazard data (such as mine pressure). The Getis-Ord Gi* statistic aims to identify hotspot areas (high-value clusters) and coldspot areas (low-value clusters) in spatial data. When the first-hazard data is distributed across different regions, the Gi* statistic for each region is calculated and compared with a critical value at a certain significance level to determine which regions have abnormally high (hotspot) or abnormally low (coldspot) levels of first-hazard data. For example, when extracting spatial features from mine pressure data, hotspot areas indicate a relatively high probability of mine pressure disasters and are key areas requiring focused attention and preventative measures.

[0164] In this embodiment of the application, the generated disaster prediction model performs in-depth analysis of the identified hotspot areas and can retrieve the specific causes from the expert knowledge base in order to take targeted preventive measures.

[0165] Optionally, the statistics for each region can be determined first based on the spatial weight matrix and the first disaster data for each region.

[0166] In this embodiment of the application, the statistics corresponding to each region can be obtained by the following formula (9).

[0167]

[0168] In formula (9), x is the statistic corresponding to the i-th region. j This is the first disaster data for region j, w ij These are the spatial weight matrix elements corresponding to regions i and j.

[0169] Then, based on the statistics, the second test value corresponding to the region can be obtained. If the second test value is greater than the third threshold, the region is determined as a target region.

[0170] In this embodiment of the application, the statistic obtained by the above formula (9) can be standardized to obtain the second detection value, so that it conforms to a certain statistical distribution law, which is convenient for significance judgment. The calculation formula corresponding to the second detection value is as follows (10):

[0171]

[0172] In formula (10), It is a statistic The expected value is calculated using the following formula: It is a statistic The variance.

[0173] In this embodiment, a critical value corresponding to the standard normal distribution (usually ±1.96 in a two-tailed test) can be found based on a set significance level (e.g., 0.05) as the third threshold. If the second detection value is greater than the third threshold, region i can be identified as a hotspot region, indicating that the first hazard data (e.g., mine pressure value) in this region is abnormally high compared to other regions, making it a high-incidence area for mine pressure disasters. It is necessary to analyze the reasons in depth and retrieve corresponding preventive measures from the expert knowledge base for targeted prevention and control. Conversely, if the second detection value is less than the negative third threshold, it is identified as a cold spot region. If it is within the range of the third threshold, it is considered that the region belongs to the normal region in terms of the spatial distribution of mine pressure data and does not show obvious abnormal clustering characteristics.

[0174] Step 305: Determine the spatial characteristics corresponding to the first disaster data based on the relationship between at least one region and the target region.

[0175] In this embodiment of the application, the relationship between at least one region of the first disaster data distribution and the determined target region can be used as the spatial features extracted from the first disaster data.

[0176] Step 306: Generate a disaster prediction model based on the first time features, spatial features, and expert knowledge base.

[0177] For a detailed description of step 306 above, please refer to the description in other embodiments of this application, which will not be repeated here.

[0178] Optionally, in any of the above embodiments of this application, when generating a disaster prediction model, a data model can first be constructed based on first temporal and spatial features, and a knowledge model can be constructed based on an expert knowledge base.

[0179] In this embodiment, the data model can be any of the classical statistical model, machine learning model, or deep learning model. After determining the type of data model to be constructed, the temporal and spatial features are fused. This can be achieved through early fusion methods such as directly concatenating spatiotemporal features into the model, or through late fusion methods such as training independent temporal and spatial sub-models and then integrating the results through weighted averaging or stacking. Model parameters are then adjusted through cross-validation, error analysis, etc., to complete the construction and training of the data model. This allows for the construction of a disaster prediction data model based on historical data. This data model comprehensively considers the temporal and spatial characteristics of disasters and can predict the frequency and intensity of disasters.

[0180] In this embodiment, the knowledge model can formalize and structure knowledge for easier computer processing and application. Knowledge can be extracted from an expert knowledge base, cleaned and formatted, and then represented using methods such as production rules, semantic networks, or frames to construct the model. This model can then be validated and optimized through expert review and test data to obtain the final knowledge model.

[0181] In this embodiment, a knowledge model can be constructed by combining knowledge and experience from an expert knowledge base to correct and optimize the prediction results of the data model. For example, when the data model predicts that there is a risk of mining pressure disaster in a certain area, the knowledge model can retrieve relevant information such as the geological structure and mining activities of the area from the expert knowledge base, gaining a deeper understanding of the disaster's mechanism and influencing factors, thus providing strong support for disaster prediction.

[0182] Then, the data model and knowledge model can be merged to generate a disaster prediction model driven by both data and knowledge.

[0183] In this embodiment, any fusion method such as weighted averaging, stacking, or parameter fusion can be used to combine the data model and the knowledge model, generating a disaster prediction model driven by both data and knowledge, thus achieving dual-driven disaster prediction and early warning. The data model provides quantitative predictions of disaster frequency and intensity, while the knowledge model provides qualitative explanations and analyses of disaster occurrence mechanisms and influencing factors. By effectively integrating the quantitative analysis capabilities of the data model with the qualitative reasoning and explanatory capabilities of the knowledge model, a synergistic mechanism between the knowledge model and the data model is achieved. This ensures that the data model and the expert knowledge model play a dual-driven role in the spatiotemporal fusion research of typical coal mine disasters, continuously improving the accuracy and real-time performance of disaster prediction.

[0184] In this embodiment of the application, when the probability of a disaster is predicted to be high, the disaster prediction model can issue an early warning signal in a timely manner to notify relevant personnel to take preventive measures, thereby improving the accuracy and reliability of disaster prediction.

[0185] It should be noted that after generating the disaster prediction model in this application, continuous optimization and updating of the model are required. Therefore, this embodiment also provides another method for predicting typical coal mine disasters using a dual-drive approach based on their spatiotemporal characteristics. Figure 4 This is a flowchart illustrating another method for predicting the spatiotemporal characteristics of typical coal mine disasters, provided in an embodiment of this application.

[0186] like Figure 4 As shown, the method for generating this disaster prediction model may include the following steps:

[0187] Step 401: Obtain the data to be updated within the preset time interval.

[0188] The preset time interval can be set according to actual needs, such as one month or one year, or it can be determined according to the frequency of disaster occurrence or the update frequency in the expert knowledge base. The higher the frequency, the shorter the preset time interval can be to ensure the real-time performance of the model prediction. This application does not limit this.

[0189] In this embodiment of the application, the fourth disaster data can be data monitored for newly occurring disasters within a preset time interval after the disaster prediction model is generated, or it can be new knowledge dynamically updated in the expert knowledge base within a preset time interval. The new knowledge data is used as the fourth disaster data for model updates.

[0190] Step 403: Update the disaster prediction model based on the data to be updated.

[0191] In this embodiment, newly collected data can be used periodically for retraining to dynamically update and improve the data model or knowledge model, adapting to new data features and improving the model's accuracy.

[0192] Optionally, the first score corresponding to the updated disaster prediction model can be determined first. Then, if the difference between the first score and the second score corresponding to the disaster prediction model before the update is less than the fourth threshold, the parameters of the updated disaster prediction model can be adjusted, and the first score can be recalculated until the first score and the second score are greater than the fourth threshold, thus completing the update of the disaster prediction model.

[0193] The fourth threshold can be a large value to ensure that the updated model has a significant performance improvement.

[0194] In this embodiment, the performance of the model after each generation or update can be evaluated using any method. If the accuracy and generalization ability of the new model are significantly higher than those of the old model (i.e., the difference between the first score and the second score is greater than the fourth threshold), the updated model can directly replace the old model. Otherwise, if the performance of the new model does not show a significant improvement (i.e., the difference between the first score and the second score is less than the fourth threshold), the model can be adjusted and optimized based on the evaluation results until its performance surpasses that of the old model. Only then can the old model be replaced. This continuously improves the accuracy and real-time performance of disaster prediction, ensures the effectiveness of model updates, avoids ineffective updates, and prevents wasted resources.

[0195] The collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in this application all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0196] It should be noted that personal information collected from users should be used for legitimate and reasonable purposes and should not be shared or sold outside of these legitimate uses. Furthermore, such collection / sharing should only be conducted after receiving the user's informed consent, including but not limited to notifying the user to read the user agreement / user notice and sign an agreement / authorization that includes authorization of relevant user information before the user uses the function. In addition, any necessary steps must be taken to protect and safeguard access to such personal information data and ensure that others with access to personal information data comply with their privacy policies and procedures.

[0197] This application is intended to provide an implementation scheme for users to selectively prevent the use or access to their personal information data. Specifically, this disclosure is intended to provide hardware and / or software to prevent or block access to such personal information data. Once personal information data is no longer needed, risks can be minimized by restricting data collection and deleting data. Furthermore, where applicable, such personal information is de-identified to protect user privacy.

[0198] In the foregoing descriptions of the embodiments, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0199] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0200] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0201] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0202] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0203] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0204] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0205] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A dual-driven prediction method for the spatiotemporal characteristics of typical coal mine disasters, characterized in that, include: Acquire historical data on the first disasters detected, as well as a pre-built expert knowledge base; Based on the time information corresponding to the first disaster data, extract the first time feature corresponding to the first disaster data; Statistical analysis is performed on the first disaster data to determine the spatial characteristics corresponding to the first disaster data; A disaster prediction model is generated based on the first temporal feature, the spatial feature, and the expert knowledge base. The step of extracting the first time feature corresponding to the first disaster data based on the time information corresponding to the first disaster data includes: Based on the time information corresponding to the first disaster data, the first disaster data is clustered to determine at least one dataset; Perform time series analysis on the second disaster data in each dataset to determine the second temporal characteristics corresponding to the dataset; Based on the second time features corresponding to all datasets, determine the first time features corresponding to the first disaster data; The step of performing time series analysis on the second disaster data in each dataset to determine the second time feature corresponding to the dataset includes: Based on the dispersion of the second disaster data and the corresponding time information, determine the associated time period of the second disaster data and its corresponding association strength; Calculate the local anomaly factor corresponding to each of the second disaster data to identify anomalies in the second disaster data; The second temporal feature corresponding to the dataset is obtained from the associated time period, the associated strength corresponding to the associated time period, and the outliers; The step of statistically analyzing the first disaster data to determine the spatial characteristics corresponding to the first disaster data includes: Based on the environmental information in the first disaster data, at least one region where the first disaster data is distributed and the relationship between the at least one region are determined; Based on the size of the first disaster data corresponding to each region, the target region among the at least one region is determined; The spatial characteristics corresponding to the first disaster data are determined based on the relationships between the at least one region and the target region; The step of generating a disaster prediction model based on the first temporal feature, the spatial feature, and the expert knowledge base includes: Based on the first temporal feature and the spatial feature, a data model is constructed; Based on the aforementioned expert knowledge base, a knowledge model is constructed; The data model and the knowledge model are fused to generate a disaster prediction model driven by both data and knowledge.

2. The method as described in claim 1, characterized in that, The step of calculating the local anomaly factor corresponding to each of the second disaster data points to identify outliers in the second disaster data includes: Determine the first distance between each of the second disaster data and other second disaster data; Based on a preset value of k and all first distances corresponding to the second disaster data, the neighborhood and second distance corresponding to the second disaster data are determined, wherein the neighborhood contains at least k third disaster data, and k is a positive integer; Based on the second distance and the first distance corresponding to all third disaster data, the first local reachability density corresponding to the second disaster data is determined; Based on the first local reachability density and the second local reachability density corresponding to all the third disaster data, determine the local anomaly factor corresponding to the second disaster data; If the local anomaly factor is greater than the first threshold, the second disaster data is identified as an anomaly point, and the operation of determining the area and second distance corresponding to the next second disaster data is returned until all anomalies in the second disaster data are identified.

3. The method as described in claim 2, characterized in that, The step of determining the neighborhood and second distance corresponding to the second disaster data based on a preset k value and all first distances corresponding to the second disaster data includes: Sort all the first distances, and determine the kth first distance in the sorting results as the second distance; All other second disaster data whose corresponding first distance is less than or equal to the second distance are identified as third disaster data; Based on the third disaster data and the second distance, the neighborhood corresponding to the second disaster data is obtained.

4. The method as described in claim 1, characterized in that, The step of determining at least one region where the first disaster data is distributed and the relationship between the at least one region based on the environmental information in the first disaster data includes: The environmental information in the first disaster data is analyzed to determine at least one area where the first disaster data is distributed; Based on the third distance between every two regions in the at least one region, a spatial weight matrix corresponding to the first disaster data is generated; Based on the spatial weight matrix and the first disaster data corresponding to each region, the Moran index corresponding to the at least one region is obtained; Based on the Moran index, the first test value is obtained; If the first test value is greater than the second threshold, the relationship between the at least one region is determined based on the sign of the Moran index.

5. The method as described in claim 4, characterized in that, The step of determining the target area among the at least one area based on the size of the first disaster data corresponding to each area includes: Based on the spatial weight matrix and the first disaster data corresponding to each region, determine the statistics corresponding to each region; Based on the statistic, the second test value corresponding to the region is obtained; If the second test value is greater than the third threshold, the region is identified as a target region.

6. The method as described in claim 1, characterized in that, Following the generation of the disaster prediction model, the following is also included: Retrieve data to be updated within a preset time interval; The disaster prediction model is updated based on the data to be updated.

7. The method as described in claim 6, characterized in that, Following the update of the disaster prediction model, the following is also included: Determine the first score corresponding to the updated disaster prediction model; If the difference between the first score and the second score corresponding to the disaster prediction model before the update is less than the fourth threshold, the parameters of the updated disaster prediction model are adjusted, and the first score is recalculated until the difference between the first score and the second score is greater than the fourth threshold, thus completing the update of the disaster prediction model.

Images