Highway disaster risk prediction method system based on multi-factor big data analysis

Abstract

The present application relates to the technical field of data processing, in particular to a highway disaster risk prediction method and system based on multi-factor big data analysis, comprising: obtaining pre-processed multi-dimensional data and real-time monitoring data of each slope region along the target highway; obtaining disaster risk precursor performance values; obtaining disaster risk spatial performance values; determining disaster risk precursor time and spatial performance values based on the disaster risk precursor performance values and the disaster risk spatial performance values; obtaining disaster risk prediction attention based on the disaster risk precursor time and spatial performance values and the three-dimensional coordinates of the monitoring points; determining the prediction weight of each moment according to the disaster risk prediction attention, and obtaining updated real-time monitoring data based on the prediction weight of each moment; inputting the pre-processed multi-dimensional data and the updated real-time monitoring data into a highway disaster risk prediction model to obtain a highway disaster risk prediction result. The present application can improve the accuracy and reliability of highway disaster risk prediction.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and specifically to a method and system for predicting highway disaster risks based on multi-factor big data analysis. Background Technology

[0002] Highway disasters (such as landslides, mudslides, road collapses, snow and ice disasters, and waterlogging) have an increasingly prominent impact on traffic safety and efficiency. Traditional highway disaster risk assessments rely heavily on manual inspections, historical disaster records, or single environmental factor analysis. These methods suffer from limited data sources, long assessment cycles, insufficient real-time performance, and low prediction accuracy, making it difficult to meet the demands of modern highway management departments for more refined, real-time, and intelligent disaster early warning systems.

[0003] Based on "multi-source data fusion," with "quantitative analysis of disaster influencing factors" as the core and "intelligent model prediction" as the means, a comprehensive highway disaster risk prediction system is constructed, encompassing "data acquisition - preprocessing - feature extraction - model training - risk prediction - result output." By integrating multi-dimensional data from meteorology, geology, highway structure, traffic operation, and remote sensing monitoring, the system quantifies the impact weight of each factor on highway disasters. Utilizing big data analytics and machine learning technologies, a predictive model is built to predict highway disaster risk levels, providing targeted support for precise prevention and control.

[0004] However, in the process of highway disaster risk prediction based on multi-factor big data analysis, different types of real-time monitoring data (such as slope displacement, groundwater level, soil moisture content, rainfall, etc.) have different degrees of influence on disaster occurrence, and the data have significant differences in terms of dimensions, noise level, time response characteristics, reliability and spatial representativeness. This can lead to the model being unable to accurately identify key disaster-causing factors, and the prediction results are easily interfered with by noisy data and secondary factors, thereby reducing the accuracy and reliability of highway disaster risk prediction. Summary of the Invention

[0005] This invention provides a method and system for predicting highway disaster risks based on multi-factor big data analysis to solve existing problems.

[0006] The highway disaster risk prediction method system based on multi-factor big data analysis of the present invention adopts the following technical solution:

[0007] One embodiment of the present invention provides a method for predicting highway disaster risk based on multi-factor big data analysis, the method comprising the following steps:

[0008] Acquire preprocessed multidimensional data and real-time monitoring data of each slope area along the target highway; among which, the real-time monitoring data includes monitoring data collected from slope displacement and direction monitoring points, groundwater level monitoring points, and soil moisture content monitoring points;

[0009] Based on the monitoring data collected at each monitoring point, the disaster risk precursor performance value of each monitoring point at each moment is obtained;

[0010] By utilizing the monitoring data of each monitoring point at each time moment, the spatial performance value of disaster risk of each monitoring point at each time moment is obtained;

[0011] Based on the disaster risk precursor performance value and disaster risk spatial performance value of each monitoring point at each time, the spatiotemporal performance value of disaster risk precursors of each monitoring point at each time is determined.

[0012] By utilizing the spatiotemporal performance values ​​of disaster risk precursors at each monitoring point at each moment, as well as the three-dimensional coordinates of each monitoring point, the disaster risk prediction attention level at each moment can be obtained.

[0013] The prediction weight for each moment is determined based on the level of attention to disaster risk prediction at each moment, and updated real-time monitoring data is obtained based on the prediction weight for each moment.

[0014] The preprocessed multidimensional data and the updated real-time monitoring data are input into the highway disaster risk prediction model to obtain the highway disaster risk prediction results.

[0015] Furthermore, the specific steps for acquiring preprocessed multidimensional data and real-time monitoring data of each slope area along the target highway are as follows:

[0016] The target highway is divided into several road segments at preset intervals, and the preset range on both sides of each road segment is defined as the slope area along the target highway.

[0017] Collect multidimensional data and real-time monitoring data for each slope area; the multidimensional data includes meteorological data, geological and topographical data, remote sensing images, highway engineering data, and historical disaster records;

[0018] The format of multidimensional data and real-time monitoring data are converted, coordinates are unified, and time is synchronized to obtain preprocessed multidimensional data and real-time monitoring data.

[0019] Furthermore, the specific steps for obtaining the disaster risk precursor performance value of each monitoring point at each time moment based on the monitoring data collected at each monitoring point are as follows:

[0020] The preprocessed monitoring data collected from each monitoring point is normalized to obtain normalized monitoring data for each monitoring point; the normalized monitoring data for each monitoring point is a time series.

[0021] For each monitoring point, the normalized monitoring data is obtained by subtracting the monitoring data of the previous time from the monitoring data of the next time moment, and the difference is used as the growth value of the next time moment.

[0022] Determine whether the growth value at each subsequent time step is greater than 0. If it is greater than 0, mark the subsequent time step as 1; if it is less than or equal to 0, mark the subsequent time step as 0, thus obtaining the risk growth mark sequence.

[0023] Subsequences continuously marked with 1 in the risk growth marker sequence are considered as continuously growing sequences;

[0024] The length of each continuously growing sequence is calculated, and the sum of the growth values ​​at the next time step corresponding to each continuously growing sequence is used to determine the total amount of each continuously growing sequence.

[0025] The mean of the normalized values ​​of the length of each continuously growing sequence and the normalized values ​​of its total length are used to determine the disaster risk precursor performance value of each continuously growing sequence.

[0026] The disaster risk precursor value of each continuously growing sequence is used as the disaster risk precursor value at the time corresponding to the time marked as 1 in the risk growth marker sequence, and 0 is used as the disaster risk precursor value at the time corresponding to the time marked as 0 in the risk growth marker sequence, so as to obtain the disaster risk precursor value of each monitoring point at each time.

[0027] Furthermore, the specific steps for obtaining the spatial performance value of disaster risk for each monitoring point at each moment using the monitoring data at each monitoring point at each moment are as follows:

[0028] Any slope displacement and direction monitoring point is taken as the first target monitoring point, and the set consisting of the first preset number of slope displacement and direction monitoring points that are closest to the first target monitoring point and the first target monitoring point is determined as the first local monitoring point set.

[0029] Calculate the mean angle between the slope displacement directions of any two monitoring points in the first local monitoring point set at each moment, divide the mean by 180 to obtain the normalized value, and subtract the normalized value from 1 to obtain the disaster risk spatial performance value of the first target monitoring point at each moment.

[0030] Any groundwater level monitoring point is taken as the second target monitoring point, and the set consisting of the first preset number of groundwater level monitoring points that are closest to the second target monitoring point and the second target monitoring point is determined as the second local monitoring point set;

[0031] Calculate the normalized value of the variance of groundwater level at each time step for all monitoring points in the second local monitoring point set, and determine the normalized value as the spatial performance value of disaster risk of the second target monitoring point at each time step;

[0032] Any soil moisture content monitoring point is taken as the third target monitoring point, and the set consisting of the first preset number of soil moisture content monitoring points that are closest to the third target monitoring point and the third target monitoring point is determined as the third local monitoring point set.

[0033] Calculate the normalized value of the variance of soil moisture content at each time step for all monitoring points in the third local monitoring point set, and determine the normalized value as the spatial performance value of disaster risk of the third target monitoring point at each time step.

[0034] Furthermore, the specific steps for determining the spatiotemporal performance value of disaster risk precursors at each monitoring point at each moment, based on the disaster risk precursor performance value and disaster risk spatial performance value at each monitoring point at each moment, are as follows:

[0035] The average of the disaster risk precursor manifestation value and the disaster risk spatial manifestation value of each monitoring point at each time moment is determined as the disaster risk precursor spatiotemporal manifestation value of each monitoring point at each time moment.

[0036] Furthermore, the specific steps for obtaining the disaster risk prediction attention level at each moment by utilizing the spatiotemporal performance value of disaster risk precursors at each monitoring point at each moment, and the three-dimensional coordinates of each monitoring point, are as follows:

[0037] Based on the spatiotemporal performance value of disaster risk precursors at each monitoring point at each time, the three-dimensional coordinates and type of each monitoring point, a scatter plot of the spatiotemporal performance of disaster risk precursors at each time is constructed.

[0038] The three-dimensional coordinates of all data points in the scatter plot are normalized to obtain the feature vector of each data point; the feature vector consists of the normalized three-dimensional coordinate values ​​and the spatiotemporal representation values ​​of disaster risk precursors.

[0039] Using the Euclidean distance between the feature vectors of any two data points in the scatter plot as the clustering distance, all data points in the scatter plot are clustered to obtain several point clusters.

[0040] A convex hull detection algorithm is used to obtain the convex hull region of each point cluster, and a polyhedral decomposition method is used to obtain the volume of the convex hull region of each point cluster.

[0041] The risk coefficient of each cluster is obtained by using the spatiotemporal performance values ​​of disaster risk precursors of all data points within each cluster, as well as the monitoring point types and quantities corresponding to all data points;

[0042] The product of the convex hull volume of each point cluster and the risk coefficient is used as the disaster risk concern of each point cluster;

[0043] The disaster risk attention levels of all point clusters are summed to obtain the disaster risk prediction attention level at each moment.

[0044] Furthermore, the specific steps for obtaining the risk coefficient of each data cluster by utilizing the spatiotemporal performance values ​​of disaster risk precursors of all data points within each cluster, as well as the monitoring point types and quantities corresponding to all data points, are as follows:

[0045] The average of the spatiotemporal performance values ​​of disaster risk precursors of all data points within each data cluster is used as the disaster risk level value of each data cluster.

[0046] Count the number of monitoring point types, the total number of monitoring points, and the number of monitoring points of each type for all data points within each point cluster;

[0047] The ratio is obtained by dividing the number of monitoring points in each category by the total number of monitoring points. The ratio is then multiplied by the normalized value of the risk assessment importance of the monitoring data in the corresponding dimension of each type of monitoring point to obtain the risk coefficient of each type of monitoring point.

[0048] The risk coefficients of all monitoring points within each cluster are summed to obtain the risk coefficients of all monitoring points.

[0049] The ratio obtained by dividing the number of monitoring point types within each cluster by 3, and the mean of the risk coefficients of all monitoring points, is used to determine the multidimensional risk synergy coefficient for each cluster.

[0050] The average of the disaster risk level value and the multidimensional risk synergy coefficient of each point cluster is used to determine the risk coefficient of each point cluster.

[0051] Furthermore, the process for obtaining the normalized value of the risk assessment importance of the monitoring data corresponding to each type of monitoring point is as follows:

[0052] The disaster risk precursor performance values ​​of each monitoring point at each time point are summed to obtain the disaster risk concern level of the monitoring data for each monitoring point;

[0053] The absolute value of the difference in disaster risk concern among monitoring data from any two monitoring points of the same type is used as the clustering distance. All monitoring points of the same type are clustered to obtain two clusters.

[0054] Calculate the mean of disaster risk concern for all monitoring points of the same type in each cluster, and identify all monitoring points of the same type in the cluster with the larger mean as high-risk monitoring points of the same type;

[0055] Calculate the mean Euclidean distance between any two high-risk monitoring points of the same type, and the maximum Euclidean distance between any two monitoring points of the same type, and divide the mean by the maximum to obtain the first ratio.

[0056] Subtracting the first ratio from 1 yields the first difference.

[0057] Calculate the ratio of the number of high-risk monitoring points of the same type to the number of monitoring points of the same type, and then calculate the average of the ratio and the first difference, and add 1 to obtain the risk judgment weight;

[0058] The average disaster risk concern of monitoring data from the same type of monitoring points is multiplied by the risk assessment weight to obtain the risk assessment importance of monitoring data for each dimension.

[0059] Calculate the sum of the risk assessment importance of all dimensions of monitoring data, and determine the normalized value of the risk assessment importance of each dimension of monitoring data as the ratio of the risk assessment importance of each dimension of monitoring data to the sum.

[0060] Furthermore, the specific steps for determining the prediction weight for each moment based on the disaster risk prediction attention at each moment, and obtaining updated real-time monitoring data based on the prediction weight at each moment, are as follows:

[0061] The ratio of the disaster risk prediction attention at each moment to the sum of the disaster risk prediction attention at all moments is determined as the prediction weight at each moment.

[0062] Based on the prediction weight at each moment, the normalized real-time monitoring data is weighted and trained using a time series prediction model to obtain the predicted time series data for the future preset period.

[0063] The normalized real-time monitoring data is fused with the predicted time-series data to form the updated real-time monitoring data.

[0064] One embodiment of the present invention provides a highway disaster risk prediction system based on multi-factor big data analysis, the system comprising the following modules:

[0065] The acquisition module is used to acquire preprocessed multidimensional data and real-time monitoring data of each slope area along the target highway; among which, the real-time monitoring data includes monitoring data collected from slope displacement and direction monitoring points, groundwater level monitoring points, and soil moisture content monitoring points;

[0066] The analysis module is used to obtain the disaster risk precursor performance value of each monitoring point at each time point based on the monitoring data collected at each monitoring point;

[0067] By utilizing the monitoring data of each monitoring point at each time moment, the spatial performance value of disaster risk of each monitoring point at each time moment is obtained;

[0068] Based on the disaster risk precursor performance value and disaster risk spatial performance value of each monitoring point at each time, the spatiotemporal performance value of disaster risk precursors of each monitoring point at each time is determined.

[0069] By utilizing the spatiotemporal performance values ​​of disaster risk precursors at each monitoring point at each moment, as well as the three-dimensional coordinates of each monitoring point, the disaster risk prediction attention level at each moment can be obtained.

[0070] The prediction weight for each moment is determined based on the level of attention to disaster risk prediction at each moment, and updated real-time monitoring data is obtained based on the prediction weight for each moment.

[0071] The prediction module is used to input the preprocessed multidimensional data and the updated real-time monitoring data into the highway disaster risk prediction model to obtain the highway disaster risk prediction results.

[0072] The beneficial effects of the technical solution of this invention are as follows: This invention proposes a highway disaster risk prediction method system based on multi-factor big data analysis. First, it analyzes the disaster risk precursor performance values ​​of each monitoring point at each time point, as well as the importance of risk judgment in a single dimension. Then, it performs a comprehensive analysis of multiple dimensions and multiple monitoring points at the same time point to determine the disaster risk prediction focus at each time point, thereby performing weighted prediction. That is, it assigns a larger prediction weight to times with high disaster risk, which can significantly improve the sensitivity and accuracy of the prediction model to key periods, making the prediction results more consistent with the actual disaster evolution law. This invention can improve the accuracy and reliability of highway disaster risk prediction. Attached Figure Description

[0073] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0074] Figure 1 This is a flowchart illustrating the steps of the highway disaster risk prediction method based on multi-factor big data analysis of the present invention.

[0075] Figure 2 This is a block diagram of the highway disaster risk prediction system based on multi-factor big data analysis of the present invention. Detailed Implementation

[0076] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of the highway disaster risk prediction method system based on multi-factor big data analysis proposed by the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0077] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0078] The specific scheme of the highway disaster risk prediction method system based on multi-factor big data analysis provided by the present invention will be described in detail below with reference to the accompanying drawings.

[0079] Please see Figure 1 The diagram illustrates a flowchart of a highway disaster risk prediction method based on multi-factor big data analysis provided by an embodiment of the present invention. The method includes the following steps:

[0080] Step S001: Obtain preprocessed multidimensional data and real-time monitoring data of each slope area along the target highway; among which, the real-time monitoring data includes monitoring data collected from slope displacement and direction monitoring points, groundwater level monitoring points, and soil moisture content monitoring points.

[0081] Step S001 further includes steps S0011-S0013:

[0082] Step S0011: Divide the target highway into several road segments according to preset intervals, and determine the preset range on both sides of each road segment as the slope areas along the target highway.

[0083] It should be noted that: the target highway refers to the highway for which disaster risk prediction is required. The slope area along the target highway is the potential disaster impact zone located outside the main structure of the target highway, and does not include the target highway itself.

[0084] The preset interval is set according to the specific scenario; here, 300 meters is preferred. The preset range is set according to the specific scenario; here, 50 meters is preferred.

[0085] Specifically, for the target highway, the road is divided into sections at 300-meter intervals, and then the area within 50 meters on both sides of each road section is taken as the slope area of ​​each road section, thus obtaining the slope areas along the target highway.

[0086] Step S0012: Collect multidimensional data and real-time monitoring data for each slope area; the multidimensional data includes meteorological data, geological and topographical data, remote sensing images, highway engineering data and historical disaster records.

[0087] It should be noted that: meteorological data usually includes rainfall, rainfall intensity, temperature, wind speed, etc.; geological and topographic data includes slope, aspect, elevation, lithology, fault distribution, etc.; remote sensing data is used to extract information such as vegetation cover and surface changes; highway engineering data includes roadbed type, slope structure, drainage system, etc.; real-time monitoring data includes slope displacement and direction, groundwater level, soil moisture content, etc.

[0088] Step S0013: Perform format conversion, coordinate unification, and time synchronization on the multidimensional data and real-time monitoring data to obtain preprocessed multidimensional data and real-time monitoring data.

[0089] It should be noted that the above data typically comes from different departments and systems, requiring preprocessing such as format conversion, coordinate standardization, and time synchronization. Data acquisition and preprocessing are well-known operations, and specific methods will not be described here.

[0090] Because geological and topographical data, highway engineering data, and historical disaster records are all collected in a single instance and updated periodically, while remote sensing imagery is collected periodically, typically updated monthly or quarterly, these data generally do not require forecasting. Meteorological data provided by weather stations generally includes not only real-time observation data but also weather forecasts for a future period.

[0091] Therefore, this embodiment focuses on accurate prediction of real-time monitoring data to ensure the accuracy of highway disaster risk prediction. Because the monitoring objectives, sensor installation methods, deployment conditions, and data representativeness requirements for slope displacement and direction, groundwater level, and soil moisture content vary significantly, three types of monitoring points are distributed at different locations within the slope area of ​​each road section.

[0092] Among them, Class A monitoring points collect real-time data on slope displacement and direction at every moment, with the outward slope direction as positive, primarily monitoring landslides; Class B monitoring points collect real-time data on groundwater level at every moment; and Class C monitoring points collect real-time data on soil moisture content at every moment. Each class has multiple monitoring points.

[0093] Step S002: Based on the monitoring data collected at each monitoring point, obtain the disaster risk precursor performance value of each monitoring point at each time.

[0094] It should be noted that, in the absence of heavy rainfall, earthquakes, construction disturbances or other abnormal factors, slope displacement, groundwater level and soil moisture content are usually basically stable, with values ​​fluctuating within a relatively fixed range. When monitoring data shows a sudden increase or a continuous upward trend in slope displacement, groundwater level, or soil moisture content, it indicates a decrease in the shear strength of the slope soil, an increase in pore water pressure, and a reduction in the overall stability of the slope. The more significant and prolonged these changes are, the greater the risk of highway disasters. For example, a sudden or continuous increase in slope displacement indicates that a sliding surface may have gradually formed within the slope, the soil shear strength is decreasing, and deformation is accelerating—the most direct and reliable precursor to slope instability. A sudden or continuous increase in groundwater level leads to increased pore water pressure, reduced effective stress, a significant decrease in soil shear strength, and a greater likelihood of potential sliding surfaces—the most typical triggering factor for landslides. A sudden or continuous increase in soil moisture content leads to soil softening, increased density, loss of matrix suction, reduced shear strength, and rainwater infiltration causing the formation or expansion of sliding surfaces. The closer the soil is to saturation, the more prone the slope is to instability. Therefore, this embodiment first determines the precursory manifestations of disaster risk at each monitoring point based on the changes in the corresponding dimensions over time.

[0095] Step S002 further includes steps S0021-S0027:

[0096] Step S0021: Normalize the monitoring data collected from each monitoring point after preprocessing to obtain normalized monitoring data for each monitoring point; wherein, the normalized monitoring data for each monitoring point is a time series.

[0097] It should be noted that each monitoring point corresponds to a three-dimensional coordinate (longitude, latitude, and altitude), and the data collection frequency for each dimension is once per hour, with a collection duration of one year. The minimum-maximum normalization method is used to normalize the data values ​​of each dimension (slope displacement, groundwater level, and soil moisture content) to between 0 and 1, unifying the dimensions for easier subsequent analysis.

[0098] Step S0022: For the normalized monitoring data of each monitoring point, the difference between the monitoring data of the previous moment and the monitoring data of the next moment is used as the increment value of the next moment.

[0099] It should be noted that for any one of the three types of monitoring points, A, B, and C, for example... For the i-th monitoring point in class A, obtain the corresponding dimension (slope displacement) time series. The difference between the normalized monitoring data value at time x and the normalized monitoring data value at time x-1 is recorded as the growth value at time x. Let the growth value at time 1 be the growth value at time 2.

[0100] Step S0023: Determine whether the growth value at each subsequent time step is greater than 0. If it is greater than 0, mark the subsequent time step as 1; if it is less than or equal to 0, mark the subsequent time step as 0, thus obtaining the risk growth mark sequence.

[0101] It should be noted that: moments with positive growth are marked as 1, and moments with zero or negative growth are marked as 0, thus obtaining... The corresponding 01 sequence is the risk growth marker sequence.

[0102] For example,

[0103] Corresponding dimension (slope displacement) time series: 0.1, 0.3, 0.7, 0.5, 0.3, 0.6, 0.1, 0.3, 0.5;

[0104] Corresponding time points: t1, t2, t3, t4, t5, t6, t7, t8, t9;

[0105] The corresponding growth values ​​at each time point are: 0.2, 0.2, 0.4, -0.2, -0.2, 0.3, -0.5, 0.2, 0.2;

[0106] 01 sequence: 1, 1, 1, 0, 0, 1, 0, 1, 1.

[0107] Step S0024: The subsequences that are continuously marked as 1 in the risk growth marker sequence are taken as the continuous growth sequence.

[0108] It should be noted that: in In a 01 sequence, consecutive adjacent 1s form a 1 sequence, which is a continuously increasing sequence. An isolated 1 also constitutes a 1 sequence.

[0109] Continuing with the example above, there are three continuously increasing sequences: 1, 1, 1; 1; 1, 1.

[0110] Step S0025: Calculate the length of each continuously growing sequence and sum the growth values ​​of each continuously growing sequence at the next time step to determine the total amount of each continuously growing sequence.

[0111] It should be noted that: the length of each 1 sequence is counted and recorded as the duration of continuous growth, and the sum of the growth values ​​corresponding to all 1s in each 1 sequence is counted and recorded as the total continuous growth, that is, the total amount of each continuous growth sequence.

[0112] Continuing with the example above, the length of the continuously growing sequence 1, 1, 1 is 3, and the total length is 0.8;

[0113] The length of the continuously growing sequence 1 is 1, and the total amount is 0.3;

[0114] The length of the continuously growing sequence 1 is 2, and the total length is 0.4.

[0115] Step S0026: The mean of the normalized value of the length of each continuously growing sequence and the normalized value of its total length is used to determine the disaster risk precursor performance value of each continuously growing sequence.

[0116] It should be noted that: using the min-max normalization method, the duration and total duration of continuous growth of all 1-sequences corresponding to all Class A (B or C) monitoring points are normalized to between 0 and 1. The mean of the normalized values ​​of the duration and total duration of continuous growth is recorded as the disaster risk precursor performance value for each 1-sequence in its time period.

[0117] Step S0027: Take the disaster risk precursor performance value of each continuously growing sequence as the disaster risk precursor performance value at the time corresponding to the time marked as 1 in the risk growth marker sequence, and take 0 as the disaster risk precursor performance value at the time corresponding to the time marked as 0 in the risk growth marker sequence, so as to obtain the disaster risk precursor performance value of each monitoring point at each time.

[0118] It should be noted that: the disaster risk precursor performance value of each time period in the 1 sequence is assigned to each moment of that time period. The disaster risk precursor performance value corresponding to each moment of the 0-1 sequence is set to 0. Thus, we obtain... The value of the precursory manifestation of disaster risk at each moment, that is, the value of the precursory manifestation of disaster risk at each monitoring point at each moment.

[0119] Continuing with the example above, the values ​​of t1, t2, t3, t6, t8, and t9 in the corresponding time intervals t1, t2, t3, t6, t8, and t9 correspond to the disaster risk precursor values ​​of the continuously growing sequence, while the disaster risk precursor values ​​of t4, t5, and t7 are 0.

[0120] Step S003: Using the monitoring data of each monitoring point at each time moment, obtain the spatial performance value of disaster risk of each monitoring point at each time moment.

[0121] It should be noted that if slope displacement, groundwater level, and soil moisture content all show a sudden increase simultaneously, or exhibit a continuous upward trend, it indicates a further increase in the risk of highway slope disasters, requiring timely early warning or reinforcement measures. Therefore, it is further necessary to analyze the disaster risk performance of the three types of monitoring data at the same time to determine the risk precursor factors at each moment.

[0122] First, since the above analysis of the temporal changes in monitoring data of each monitoring point determines the disaster risk precursor performance value of each monitoring point at each moment, it is further necessary to combine the spatial change analysis of similar monitoring points at the same moment to determine the temporal spatial performance value of disaster risk precursors of each monitoring point at each moment.

[0123] Step S003 further includes steps S0031-S0036:

[0124] Step S0031: Take any slope displacement and direction monitoring point as the first target monitoring point, and determine the set consisting of the first preset number of slope displacement and direction monitoring points closest to the first target monitoring point and the first target monitoring point as the first local monitoring point set.

[0125] It should be noted that the preset quantity is set according to specific circumstances, and 10 is preferred here. The distance from the first target monitoring point is the Euclidean distance.

[0126] For the first target monitoring point, the 10 monitoring points closest to the first target monitoring point (of the same type as the first target monitoring point) are obtained, and combined with the first target monitoring point to form the first local monitoring point set (11 monitoring points).

[0127] Step S0032: Calculate the mean angle between the slope displacement directions of any two monitoring points in the first local monitoring point set at each moment, divide the mean by 180 to obtain the normalized value, and subtract the normalized value from 1 to obtain the disaster risk spatial performance value of the first target monitoring point at each moment.

[0128] It should be noted that, taking time t as an example, at time t:

[0129] If this is class A, then in the local monitoring point set, obtain the mean of the minimum included angle between the slope displacement directions of any two monitoring points at time t. ,Will Let be the spatial performance value of the disaster risk at the first target monitoring point at time t. This is the normalized value.

[0130] Step S0033: Select any groundwater level monitoring point as the second target monitoring point, and determine the set consisting of the pre-set number of groundwater level monitoring points closest to the second target monitoring point and the second target monitoring point as the second local monitoring point set.

[0131] It should be noted that the preset quantity should be set according to specific circumstances; here, 10 is preferred. The distance to the second target monitoring point is the Euclidean distance.

[0132] For the second target monitoring point, the top 10 monitoring points (of the same type as the second target monitoring point) that are closest to the second target monitoring point are obtained, and combined with the second target monitoring point to form the second local monitoring point set (11 monitoring points).

[0133] Step S0034: Calculate the normalized value of the variance of the groundwater level of all monitoring points in the second local monitoring point set at each time moment, and determine the normalized value as the spatial performance value of the disaster risk of the second target monitoring point at each time moment.

[0134] It should be noted that: if this is type B, then in the local monitoring point set, the normalized value of the variance of the groundwater level of all arbitrary monitoring points at time t is obtained and recorded as the disaster risk spatial performance value of the second target monitoring point at time t.

[0135] Among them, the minimum-maximum normalization method is used to normalize the variance of the groundwater level corresponding to all monitoring points in category B to between 0 and 1 at all times, thus eliminating the influence of dimensions.

[0136] Step S0035: Select any soil moisture content monitoring point as the third target monitoring point, and determine the set consisting of the first preset number of soil moisture content monitoring points closest to the third target monitoring point and the third target monitoring point as the third local monitoring point set.

[0137] It should be noted that the preset quantity should be set according to specific circumstances; here, 10 is preferred. The distance to the third target monitoring point is the Euclidean distance.

[0138] For the third target monitoring point, the top 10 monitoring points that are closest to the third target monitoring point (of the same type as the third target monitoring point) are obtained, and combined with the third target monitoring point, a third local monitoring point set (11 monitoring points) is formed.

[0139] Step S0036: Calculate the normalized value of the variance of soil moisture content of all monitoring points in the third local monitoring point set at each time step, and determine the normalized value as the spatial performance value of disaster risk of the third target monitoring point at each time step.

[0140] It should be noted that: if this is type C, then in the local monitoring point set, the normalized value of the variance of soil moisture content of all arbitrary monitoring points at time t is obtained and recorded as the disaster risk spatial performance value of the third target monitoring point at time t.

[0141] Among them, the minimum-maximum normalization method was used to normalize the variance of soil moisture content at all monitoring points in category C to between 0 and 1 at all times, thus eliminating the influence of dimensions.

[0142] Step S004: Based on the disaster risk precursor performance value and disaster risk spatial performance value of each monitoring point at each time, determine the disaster risk precursor time-space performance value of each monitoring point at each time.

[0143] Specifically, this includes: determining the average of the disaster risk precursor performance value and the disaster risk spatial performance value of each monitoring point at each time moment as the disaster risk precursor spatiotemporal performance value of each monitoring point at each time moment.

[0144] It should be noted that for slope displacement, when multiple monitoring points move in the same direction, i.e., the smaller the angle between the slope displacement directions is close to 0, it indicates that they may be located on the same large landslide body, resulting in overall slope deformation and a high risk, requiring close monitoring. Regarding groundwater level and soil moisture content, the greater the difference in the distribution of groundwater level or soil moisture content within the same local area, the more uneven the hydrological conditions and mechanical properties within the slope, and the higher the possibility of local shearing and instability, thus increasing the risk of highway slope disasters.

[0145] At time t, the average of the disaster risk precursor performance value and the disaster risk spatial performance value of each monitoring point is obtained as the disaster risk precursor time-space performance value of each monitoring point.

[0146] Step S005: Using the spatiotemporal performance value of disaster risk precursors at each monitoring point at each moment, and the three-dimensional coordinates of each monitoring point, obtain the disaster risk prediction attention level at each moment.

[0147] Step S005 further includes steps S0051-S0057:

[0148] Step S0051: Based on the spatiotemporal performance value of disaster risk precursors at each monitoring point at each time, the three-dimensional coordinates of each monitoring point, and the type, construct a scatter plot of the spatiotemporal performance of disaster risk precursors at each time.

[0149] It should be noted that: based on the three-dimensional coordinates of all monitoring points in categories A, B, and C, and the spatiotemporal performance value of disaster risk precursors for each monitoring point at time t, a scatter plot of the spatiotemporal performance of disaster risk precursors at time t is constructed. In the scatter plot, each data point is labeled with {timestamp (time t), three-dimensional coordinates, type (ABC), spatiotemporal performance value of disaster risk precursors}.

[0150] Step S0052: Normalize the three-dimensional coordinates of all data points in the scatter plot to obtain the feature vector of each data point; wherein, the feature vector consists of the normalized three-dimensional coordinate value and the spatiotemporal representation value of disaster risk precursors.

[0151] It should be noted that the min-max normalization method is used to normalize the x, y, and y coordinates of all data points in the scatter plot. In the scatter plot, the feature vector of each data point is obtained {normalized three-dimensional coordinates and spatial representation of disaster risk precursors}.

[0152] Step S0053: Using the Euclidean distance between the feature vectors of any two data points in the scatter plot as the clustering distance, cluster all data points in the scatter plot to obtain several point clusters.

[0153] It should be noted that: using the Euclidean distance between the feature vectors of any two data points, a hierarchical clustering algorithm is used to perform clustering operations, resulting in several point clusters.

[0154] Step S0054: Use the convex hull detection algorithm to obtain the convex hull region of each point cluster, and use the polyhedral decomposition method to obtain the volume of the convex hull region of each point cluster.

[0155] It should be noted that in the scatter plot, the convex hull detection algorithm is used to obtain the convex hull region of each point cluster, and the polyhedral decomposition method is used to obtain the volume of the convex hull region of each point cluster.

[0156] Step S0055: Using the spatiotemporal performance values ​​of disaster risk precursors of all data points within each data cluster, as well as the monitoring point types and quantities corresponding to all data points, obtain the risk coefficient of each data cluster.

[0157] Step S0055 further includes steps S0551-S0556:

[0158] Step S0551: Take the average of the spatiotemporal performance values ​​of disaster risk precursors of all data points within each data cluster as the disaster risk level value of each data cluster.

[0159] It should be noted that the average value of the spatiotemporal performance of disaster risk precursors of all data points within each point cluster is obtained and recorded as the disaster risk level value of each point cluster.

[0160] Step S0552: Count the number of monitoring point types, the total number of monitoring points, and the number of monitoring points of each type for all data points within each point cluster.

[0161] It should be noted that: for any given point cluster, the number of monitoring point types existing within the cluster is counted. Then count the number of monitoring points of type j within the point cluster. The total number of monitoring points M in the point cluster.

[0162] Step S0553: Divide the number of monitoring points of each type by the total number of monitoring points to obtain the ratio, and multiply the ratio by the normalized value of the risk assessment importance of the monitoring data of the corresponding dimension of each type of monitoring point to obtain the risk coefficient of each type of monitoring point.

[0163] It should be noted that: Let the risk coefficient be denoted as the risk coefficient of the j-th type of monitoring point within the point cluster. Let be the normalized value of the risk assessment importance of the j-th class in the point cluster.

[0164] The process of obtaining the risk assessment importance normalization value of the monitoring data for each type of monitoring point is as follows: steps S5531-S5538:

[0165] Step S5531: Sum the disaster risk precursor performance values ​​of each monitoring point at each time point to obtain the disaster risk concern level of the monitoring data for each monitoring point.

[0166] It should be noted that in highway slope disaster risk prediction, the contribution of each monitoring dimension (slope displacement, groundwater level, soil moisture content, etc.) to the risk assessment needs to be determined based on the actual monitoring performance of that dimension. If the monitoring data of a certain dimension shows a more obvious disaster risk, it indicates that this factor has a more significant impact on the current slope stability and therefore requires more attention.

[0167] For each monitoring point, the sum of the disaster risk precursor values ​​for all time periods of the 1st sequence is obtained, that is, the sum of the disaster risk precursor values ​​of each monitoring point at each time moment is used as the disaster risk concern of each monitoring point's monitoring data.

[0168] Among them, the larger the total value of disaster risk precursors at each monitoring point over the entire time series, the more attention it needs to pay.

[0169] Step S5532: Using the absolute difference in disaster risk concern between any two monitoring points of the same type as the clustering distance, cluster all monitoring points of the same type to obtain two clusters.

[0170] It should be noted that for all Class A (Class B or Class C) monitoring points, the absolute value of the difference between the disaster risk attention of any two monitoring points is used as the clustering distance. The K-means clustering algorithm is used for clustering, and K is set to 2 to obtain two clusters.

[0171] Step S5533: Calculate the mean of disaster risk concern for all monitoring data of the same type of monitoring points in each cluster, and identify all monitoring points of the same type in the cluster with the larger mean as high-risk monitoring points of the same type.

[0172] It should be noted that: the average disaster risk concern of all monitoring points in each cluster is obtained, and all monitoring points in the cluster corresponding to the larger average are recorded as high-risk monitoring points.

[0173] Step S5534: Calculate the mean of the Euclidean distance between any two high-risk monitoring points of the same type, and the maximum value of the Euclidean distance between any two monitoring points of the same type, and divide the mean by the maximum value to obtain the first ratio.

[0174] It should be noted that: This is denoted as the first ratio. This represents the mean Euclidean distance between any two high-risk monitoring points in category A (or category B or C). It is the maximum value of the Euclidean distance between any two Class A (Class B or Class C) monitoring points.

[0175] Step S5535: Subtract the first ratio from 1 to obtain the first difference.

[0176] It should be noted that: This is denoted as the first difference.

[0177] Step S5536: Calculate the ratio of the number of high-risk monitoring points of the same type to the number of monitoring points of the same type, and calculate the average of the ratio and the first difference, then add 1 to obtain the risk judgment weight.

[0178] It should be noted that: Recorded as risk assessment weight, The number of monitoring points in category A (or category B or C). This refers to the number of high-risk monitoring points in Category A (or Category B or C).

[0179] Step S5537: Multiply the average disaster risk concern of the monitoring data of the same type of monitoring points by the risk judgment weight to obtain the risk judgment importance of the monitoring data of each dimension.

[0180] It should be noted that, taking Category A as an example, the importance of risk assessment for Category A (slope displacement) is... for:

[0181] ;

[0182] in, This represents the average level of disaster risk awareness across all Class A monitoring sites. This reflects the overall risk performance of Category A, while The smaller the value, the more concentrated the locations of high-risk monitoring points in Category A are. The larger the value, the more numerous and concentrated the high-risk monitoring points in Category A are, thus indicating... As The adjustment value is obtained. , The larger the value, the more pronounced the disaster risk indicated by Category A monitoring, meaning that Category A monitoring data is more important in risk assessment. If there is only one high-risk monitoring point in Category A, then the above formula will be... Replace with .

[0183] Similarly, the importance of obtaining risk assessments for categories B and C. and .

[0184] Step S5538: Calculate the sum of the risk assessment importance of all dimensions of monitoring data, and determine the ratio of the risk assessment importance of each dimension of monitoring data to the sum of the sums as the normalized value of the risk assessment importance of each dimension of monitoring data.

[0185] It should be noted that the sum of the risk assessment importance of each category and the risk assessment importance of all categories is used. The ratio of these values ​​is used as a normalized value for assessing the importance of risk for each category.

[0186] Step S0554: Sum the risk coefficients of all monitoring points within each point cluster to obtain the risk coefficients of all monitoring points.

[0187] It should be noted that: Recorded as the risk coefficient for all monitoring points.

[0188] Step S0555: Divide the number of monitoring point types in each cluster by 3 to obtain the ratio, and then use the average risk coefficient of all monitoring points to determine the multidimensional risk synergy coefficient for each cluster.

[0189] It should be noted that the multidimensional risk synergy coefficient for each point cluster... for:

[0190] ;

[0191] Step S0556: Determine the risk coefficient of each point cluster by the average of the disaster risk level value and the multidimensional risk synergy coefficient.

[0192] It should be noted that the average value of the disaster risk level and the multidimensional risk synergy coefficient of each point cluster is obtained and recorded as the risk coefficient of each point cluster.

[0193] Among these, if the disaster risk level value of a point cluster is high, and N is close to 3, it indicates that all dimensions within this point cluster region are at risk at time t, meaning the risk is even higher. However, the importance of each type of data for risk assessment varies. Therefore, the proportion of each type of monitoring point in the point cluster is used as the weight to perform a weighted summation of the risk assessment importance of different types within the point cluster. As an important factor in comprehensive risk assessment, it is used to determine... Adjustments are made to obtain a multidimensional risk synergy coefficient, which is then used to determine the risk coefficient of each point cluster.

[0194] Step S0056: Multiply the volume of the convex hull region of each point cluster by the risk coefficient, and use the disaster risk concern of each point cluster as the disaster risk concern.

[0195] It should be noted that: Record the disaster risk concern level for each point cluster. Let be the volume of the convex hull region of the u-th point cluster. Let be the risk coefficient of the u-th point cluster.

[0196] Step S0057: Sum the disaster risk attention of all point clusters to obtain the disaster risk prediction attention at each moment.

[0197] It should be noted that: Therefore, we can infer the level of attention given to disaster risk prediction at time t. for:

[0198] ;

[0199] In the formula, The number of point clusters in the scatter plot represents the area of ​​the cluster with higher risk coefficient. The larger the area of ​​the cluster, the more significant the disaster risk at time t, and the more attention it needs to be paid to it during prediction.

[0200] Step S006: Determine the prediction weight for each moment based on the disaster risk prediction attention at each moment, and obtain updated real-time monitoring data based on the prediction weight for each moment.

[0201] Step S006 further includes steps S0061-S0063:

[0202] Step S0061: Determine the prediction weight for each moment as the ratio of the disaster risk prediction attention at each moment to the sum of the disaster risk prediction attention at all moments.

[0203] It should be noted that: the sum of disaster risk prediction attention at all times is obtained. The disaster risk prediction attention and sum at each moment The ratio of these values ​​is used as the prediction weight at each time step.

[0204] Step S0062: Based on the prediction weight at each moment, use the time series prediction model to perform weighted training and prediction on the normalized real-time monitoring data to obtain the predicted time series data for the future preset period.

[0205] It should be noted that the time series forecasting model is ARIMA.

[0206] Based on the prediction weight at each moment, weighted ARIMA is used to predict the dimensional time series corresponding to each monitoring point, thereby obtaining the predicted time series for the next month.

[0207] ARIMA (Autoregressive Integral Moving Average) is known to be suitable for predicting linear and stationary sequences and is widely used in engineering monitoring systems. In this embodiment, a weighted ARIMA model is used to predict groundwater level, soil moisture content, and slope displacement trends. Specifically, assigning larger prediction weights to periods of high disaster risk significantly improves the model's sensitivity and accuracy during critical periods, making the prediction results more consistent with the actual evolution of disasters.

[0208] Step S0063: The normalized real-time monitoring data is fused with the predicted time-series data to form the updated real-time monitoring data.

[0209] It should be noted that the updated real-time monitoring data for the slope area of ​​each road segment is constructed by combining the real-time monitoring data of each road segment with the predicted time series of all monitoring points for the next month.

[0210] Step S007: Input the preprocessed multidimensional data and the updated real-time monitoring data into the highway disaster risk prediction model to obtain the highway disaster risk prediction results.

[0211] It should be noted that the highway disaster risk prediction model was constructed using a machine learning model (random forest), which classifies risk levels by training on historical disaster samples. During model training, methods such as cross-validation were employed to optimize parameters and improve prediction accuracy.

[0212] Meteorological data, geological and topographical data, remote sensing images, highway engineering data, historical disaster records, and updated real-time monitoring data from the slope area of ​​each road section are input into the highway disaster risk prediction model, and the disaster risk prediction results for each road section are output.

[0213] In existing technologies, highway disaster risk prediction models based on the random forest algorithm have been widely used. This method typically includes steps such as data collection, feature selection, model training, probability prediction, and risk level classification. By using multi-source data such as slope, aspect, lithology, rainfall, vegetation cover, and roadbed type as input, the random forest algorithm is used to train historical disaster samples to obtain the probability of disaster occurrence along the highway, which is then further classified into different risk levels, i.e., low, medium, high, and extremely high risk levels according to a preset classification standard.

[0214] Please see Figure 2The diagram illustrates a block diagram of a highway disaster risk prediction system based on multi-factor big data analysis according to an embodiment of the present invention. The system includes the following modules:

[0215] The acquisition module 100 is used to acquire preprocessed multidimensional data and real-time monitoring data of each slope area along the target highway; among which, the real-time monitoring data includes monitoring data collected from slope displacement and direction monitoring points, groundwater level monitoring points, and soil moisture content monitoring points;

[0216] Analysis module 200 is used to obtain the disaster risk precursor performance value of each monitoring point at each time based on the monitoring data collected at each monitoring point;

[0217] By utilizing the monitoring data of each monitoring point at each time moment, the spatial performance value of disaster risk of each monitoring point at each time moment is obtained;

[0218] Based on the disaster risk precursor performance value and disaster risk spatial performance value of each monitoring point at each time, the spatiotemporal performance value of disaster risk precursors of each monitoring point at each time is determined.

[0219] By utilizing the spatiotemporal performance values ​​of disaster risk precursors at each monitoring point at each moment, as well as the three-dimensional coordinates of each monitoring point, the disaster risk prediction attention level at each moment can be obtained.

[0220] The prediction weight for each moment is determined based on the level of attention to disaster risk prediction at each moment, and updated real-time monitoring data is obtained based on the prediction weight for each moment.

[0221] The prediction module 300 is used to input the preprocessed multidimensional data and the updated real-time monitoring data into the highway disaster risk prediction model to obtain the highway disaster risk prediction results.

[0222] In summary, in this embodiment of the invention, the disaster risk precursor performance value of each monitoring point at each time moment is first analyzed, along with the importance of risk assessment in a single dimension. Then, a comprehensive analysis of multiple dimensions and multiple monitoring points at the same time moment is performed to determine the disaster risk prediction attention level at each time moment, thereby performing weighted prediction. That is, a larger prediction weight is assigned to times with high disaster risk, which can significantly improve the sensitivity and prediction accuracy of the prediction model to key periods, making the prediction results more consistent with the actual disaster evolution pattern.

[0223] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting highway disaster risk based on multi-factor big data analysis, characterized in that, The method includes the following steps: Acquire preprocessed multidimensional data and real-time monitoring data of each slope area along the target highway; among which, the real-time monitoring data includes monitoring data collected from slope displacement and direction monitoring points, groundwater level monitoring points, and soil moisture content monitoring points; Based on the monitoring data collected at each monitoring point, the disaster risk precursor performance value of each monitoring point at each time moment is obtained, specifically including: The preprocessed monitoring data collected from each monitoring point is normalized to obtain normalized monitoring data for each monitoring point; the normalized monitoring data for each monitoring point is a time series. For each monitoring point, the normalized monitoring data is obtained by subtracting the monitoring data of the previous time from the monitoring data of the next time moment, and the difference is used as the growth value of the next time moment. Determine whether the growth value at each subsequent time step is greater than 0. If it is greater than 0, mark the subsequent time step as 1; if it is less than or equal to 0, mark the subsequent time step as 0, thus obtaining the risk growth mark sequence. Subsequences continuously marked with 1 in the risk growth marker sequence are considered as continuously growing sequences; The length of each continuously growing sequence is calculated, and the sum of the growth values ​​at the next time step corresponding to each continuously growing sequence is used to determine the total amount of each continuously growing sequence. The mean of the normalized values ​​of the length of each continuously growing sequence and the normalized values ​​of its total length are used to determine the disaster risk precursor performance value of each continuously growing sequence. The disaster risk precursor performance value of each continuously growing sequence is used as the disaster risk precursor performance value at the time corresponding to the time marked as 1 in the risk growth marker sequence, and 0 is used as the disaster risk precursor performance value at the time corresponding to the time marked as 0 in the risk growth marker sequence, so as to obtain the disaster risk precursor performance value of each monitoring point at each time. Using the monitoring data of each monitoring point at each time point, the spatial performance value of disaster risk for each monitoring point at each time point is obtained, specifically including: Any slope displacement and direction monitoring point is taken as the first target monitoring point, and the set consisting of the first preset number of slope displacement and direction monitoring points that are closest to the first target monitoring point and the first target monitoring point is determined as the first local monitoring point set. Calculate the mean angle between the slope displacement directions of any two monitoring points in the first local monitoring point set at each moment, divide the mean by 180 to obtain the normalized value, and subtract the normalized value from 1 to obtain the disaster risk spatial performance value of the first target monitoring point at each moment. Any groundwater level monitoring point is taken as the second target monitoring point, and the set consisting of the first preset number of groundwater level monitoring points that are closest to the second target monitoring point and the second target monitoring point is determined as the second local monitoring point set; Calculate the normalized value of the variance of groundwater level at each time step for all monitoring points in the second local monitoring point set, and determine the normalized value as the spatial performance value of disaster risk of the second target monitoring point at each time step; Any soil moisture content monitoring point is taken as the third target monitoring point, and the set consisting of the first preset number of soil moisture content monitoring points that are closest to the third target monitoring point and the third target monitoring point is determined as the third local monitoring point set. Calculate the normalized value of the variance of soil moisture content at each time step for all monitoring points in the third local monitoring point set, and determine the normalized value as the spatial performance value of disaster risk of the third target monitoring point at each time step; Based on the disaster risk precursor performance value and disaster risk spatial performance value of each monitoring point at each time moment, the spatiotemporal performance value of disaster risk precursors at each monitoring point at each time moment is determined, specifically including: The average of the disaster risk precursor performance value and the disaster risk spatial performance value of each monitoring point at each time moment is determined as the disaster risk precursor spatiotemporal performance value of each monitoring point at each time moment. By utilizing the spatiotemporal performance values ​​of disaster risk precursors at each monitoring point at each time moment, and the three-dimensional coordinates of each monitoring point, the disaster risk prediction attention level at each time moment is obtained, specifically including: Based on the spatiotemporal performance value of disaster risk precursors at each monitoring point at each time, the three-dimensional coordinates and type of each monitoring point, a scatter plot of the spatiotemporal performance of disaster risk precursors at each time is constructed. The three-dimensional coordinates of all data points in the scatter plot are normalized to obtain the feature vector of each data point; the feature vector consists of the normalized three-dimensional coordinate values ​​and the spatiotemporal representation values ​​of disaster risk precursors. Using the Euclidean distance between the feature vectors of any two data points in the scatter plot as the clustering distance, all data points in the scatter plot are clustered to obtain several point clusters. A convex hull detection algorithm is used to obtain the convex hull region of each point cluster, and a polyhedral decomposition method is used to obtain the volume of the convex hull region of each point cluster. The risk coefficient of each cluster is obtained by using the spatiotemporal performance values ​​of disaster risk precursors of all data points within each cluster, as well as the monitoring point types and quantities corresponding to all data points; The product of the convex hull volume of each point cluster and the risk coefficient is used as the disaster risk concern of each point cluster; The disaster risk attention of all point clusters is summed to obtain the disaster risk prediction attention at each moment; The prediction weight for each moment is determined based on the level of attention to disaster risk prediction at each moment, and updated real-time monitoring data is obtained based on the prediction weight for each moment. The preprocessed multidimensional data and the updated real-time monitoring data are input into the highway disaster risk prediction model to obtain the highway disaster risk prediction results.

2. The highway disaster risk prediction method based on multi-factor big data analysis according to claim 1, characterized in that, The specific steps for obtaining preprocessed multidimensional data and real-time monitoring data of each slope area along the target highway are as follows: The target highway is divided into several road segments at preset intervals, and the preset range on both sides of each road segment is defined as the slope area along the target highway. Collect multidimensional data and real-time monitoring data for each slope area; the multidimensional data includes meteorological data, geological and topographical data, remote sensing images, highway engineering data and historical disaster records; The format of multidimensional data and real-time monitoring data are converted, coordinates are unified, and time is synchronized to obtain preprocessed multidimensional data and real-time monitoring data.

3. The highway disaster risk prediction method based on multi-factor big data analysis according to claim 1, characterized in that, The specific steps for obtaining the risk coefficient of each data cluster by utilizing the spatiotemporal performance values ​​of disaster risk precursors of all data points within each cluster, as well as the monitoring point types and quantities corresponding to all data points, are as follows: The average of the spatiotemporal performance values ​​of disaster risk precursors of all data points within each data cluster is used as the disaster risk level value of each data cluster. Count the number of monitoring point types, the total number of monitoring points, and the number of monitoring points of each type for all data points within each point cluster; The ratio is obtained by dividing the number of monitoring points in each category by the total number of monitoring points. The ratio is then multiplied by the normalized value of the risk assessment importance of the monitoring data in the corresponding dimension of each type of monitoring point to obtain the risk coefficient of each type of monitoring point. The risk coefficients of all monitoring points within each cluster are summed to obtain the risk coefficients of all monitoring points. The ratio obtained by dividing the number of monitoring point types within each cluster by 3, and the mean of the risk coefficients of all monitoring points, is used to determine the multidimensional risk synergy coefficient for each cluster. The average of the disaster risk level value and the multidimensional risk synergy coefficient of each point cluster is used to determine the risk coefficient of each point cluster.

4. The highway disaster risk prediction method based on multi-factor big data analysis according to claim 3, characterized in that, The process for obtaining the risk assessment importance normalization value of the monitoring data for each type of monitoring point is as follows: The disaster risk precursor performance values ​​of each monitoring point at each time point are summed to obtain the disaster risk concern level of the monitoring data for each monitoring point; The absolute value of the difference in disaster risk concern among monitoring data from any two monitoring points of the same type is used as the clustering distance. All monitoring points of the same type are clustered to obtain two clusters. Calculate the mean of disaster risk concern for all monitoring points of the same type in each cluster, and identify all monitoring points of the same type in the clusters where the mean is greater than the mean of the other cluster as high-risk monitoring points of the same type; Calculate the mean Euclidean distance between any two high-risk monitoring points of the same type, and the maximum Euclidean distance between any two monitoring points of the same type, and divide the mean by the maximum to obtain the first ratio. Subtracting the first ratio from 1 yields the first difference. Calculate the ratio of the number of high-risk monitoring points of the same type to the number of monitoring points of the same type, and then calculate the average of the ratio and the first difference, and add 1 to obtain the risk judgment weight; The average disaster risk concern of monitoring data from the same type of monitoring points is multiplied by the risk assessment weight to obtain the risk assessment importance of monitoring data for each dimension. Calculate the sum of the risk assessment importance of all dimensions of monitoring data, and determine the normalized value of the risk assessment importance of each dimension of monitoring data as the ratio of the risk assessment importance of each dimension of monitoring data to the sum.

5. The highway disaster risk prediction method based on multi-factor big data analysis according to claim 1, characterized in that, The specific steps involved in determining the prediction weight for each moment based on the disaster risk prediction attention level, and obtaining updated real-time monitoring data based on the prediction weight for each moment, are as follows: The ratio of the disaster risk prediction attention at each moment to the sum of the disaster risk prediction attention at all moments is determined as the prediction weight at each moment. Based on the prediction weight at each moment, the normalized real-time monitoring data is trained and predicted using a time series prediction model to obtain the predicted time series data for the future preset period. The normalized real-time monitoring data is fused with the predicted time-series data to form the updated real-time monitoring data.

6. A highway disaster risk prediction system based on multi-factor big data analysis, used in the highway disaster risk prediction method based on multi-factor big data analysis as described in any one of claims 1 to 5, characterized in that, The system includes the following modules: The acquisition module is used to acquire preprocessed multidimensional data and real-time monitoring data of each slope area along the target highway; among which, the real-time monitoring data includes monitoring data collected from slope displacement and direction monitoring points, groundwater level monitoring points, and soil moisture content monitoring points; The analysis module is used to obtain the disaster risk precursor performance value of each monitoring point at each time point based on the monitoring data collected at each monitoring point; By utilizing the monitoring data of each monitoring point at each time moment, the spatial performance value of disaster risk of each monitoring point at each time moment is obtained; Based on the disaster risk precursor performance value and disaster risk spatial performance value of each monitoring point at each time, the spatiotemporal performance value of disaster risk precursors at each monitoring point at each time is determined. By utilizing the spatiotemporal performance values ​​of disaster risk precursors at each monitoring point at each moment, as well as the three-dimensional coordinates of each monitoring point, the disaster risk prediction attention level at each moment can be obtained. The prediction weight for each moment is determined based on the level of attention to disaster risk prediction at each moment, and updated real-time monitoring data is obtained based on the prediction weight for each moment. The prediction module is used to input the preprocessed multidimensional data and the updated real-time monitoring data into the highway disaster risk prediction model to obtain the highway disaster risk prediction results.

Images