Method and apparatus for generating a predictive map for predicting the possibility of sinkhole occurrence

Abstract

A method for generating a prediction map and an apparatus for generating a prediction map that predicts the probability of sinkhole occurrence are disclosed. The method for generating a prediction map may include the steps of: collecting data on whether sinkholes occur by region and data related to sinkhole occurrence by region; performing data preprocessing on the data on whether sinkholes occur by region and data related to sinkhole occurrence by region; generating a prediction model using the data on whether sinkholes occur by region and data related to sinkhole occurrence by region for which data preprocessing has been performed; and obtaining a prediction map from the prediction model that indicates the probability of sinkhole occurrence based on the data related to sinkhole occurrence by region.

Description

Technology Field

[0001] The following disclosure relates to a method and apparatus for generating a prediction map that predicts the likelihood of a sinkhole occurring. Background Technology

[0002] As the density of urban underground utilities increases, the ground weakens, leading to a rapid increase in sinkhole accidents, which are large-scale ground collapses. Under conventional underground safety laws regarding urban ground collapses, investigations and post-accident recovery are carried out only after an accident occurs. However, post-accident recovery cannot be considered an appropriate measure, as it is merely a step taken after significant loss of life and property damage has already occurred. Research has been conducted to predict, prevent, or minimize damage from sinkhole accidents. To minimize damage caused by sinkhole accidents, research has traditionally focused on technologies to predict sinkhole occurrence based on physical measurement data, such as Ground Penetrating Radar (GPR) and soil surveys. Additionally, there are studies that map the risk of sinkhole occurrence using underground facility databases and index-type models.

[0003] A method for generating a prediction map for predicting the likelihood of a sinkhole occurring according to one embodiment may include: collecting data on whether a sinkhole has occurred in each region and data related to the occurrence of a sinkhole in each region; performing data preprocessing on the data on whether a sinkhole has occurred in each region and the data related to the occurrence of a sinkhole in each region; generating a prediction model using the data on whether a sinkhole has occurred in each region and the data related to the occurrence of a sinkhole in each region for which data preprocessing has been performed; and obtaining a prediction map from the prediction model that indicates the likelihood of a sinkhole occurring based on the data related to the occurrence of a sinkhole in each region.

[0004] The step of performing the above data preprocessing may include a step of performing imbalanced data processing by adjusting class weights inversely proportional to the frequency of each class based on regional sinkhole occurrence data, or by extracting bootstraps from minority classes determined based on the regional sinkhole occurrence data.

[0005] The above data on whether sinkholes occur by region may include data indicating whether sinkholes have actually occurred in each region corresponding to each pixel on the map.

[0006] The above prediction model may be a machine learning-based model trained to predict the probability of a sinkhole occurring in each region corresponding to each pixel on a map.

[0007] The above prediction model can generate the prediction map by predicting the probability of a sinkhole occurring in each region corresponding to each pixel on the map, and displaying the predicted probability on each pixel corresponding to the predicted probability.

[0008] The above prediction model can generate the prediction map by mapping location information corresponding to each pixel and the probability of the sinkhole occurring for each pixel.

[0009] The above prediction model can generate the prediction map by displaying it in a predetermined color based on the magnitude of the value of the probability of the sinkhole occurring.

[0010] The above data regarding sinkhole occurrence by region may include at least one of information on water pipe density, water pipe distance map, sewer pipe density, sewer pipe distance map, land use map, road network distance map, road network density map, water system network density map, fault distance map, fault density map, schematic soil map, linear structure density map, or linear structure distance map.

[0011] A prediction map generating device for predicting the probability of a sinkhole occurring according to one embodiment includes a memory and a processor, wherein the memory stores instructions executable by the processor, and when the instructions are executed by the processor, the processor may enable the prediction map generating device to collect data on whether a sinkhole occurs by region and data related to sinkhole occurrence by region, perform data preprocessing on the data on whether a sinkhole occurs by region and the data related to sinkhole occurrence by region, generate a prediction model using the data on whether a sinkhole occurs by region and the data related to sinkhole occurrence by region for which data preprocessing has been performed, and obtain a prediction map indicating the probability of a sinkhole occurring based on the data related to sinkhole occurrence by region from the prediction model. Brief explanation of the drawing

[0012] FIG. 1 is a diagram illustrating an overview of a prediction map generation system according to one embodiment. FIG. 2 is a flowchart illustrating a method for generating a prediction map according to one embodiment. FIG. 3 is a diagram illustrating a method for generating a prediction map according to another embodiment. FIG. 4 is a diagram illustrating an example of data on whether sinkholes occur by region according to one embodiment. FIGS. 5 to 8 are drawings illustrating an example of data related to the occurrence of sinkholes by region according to one embodiment. FIGS. 9 and FIGS. 10 are drawings illustrating an example of a prediction map according to one embodiment. FIG. 11 is a diagram illustrating the configuration of a prediction map generation device according to one embodiment. Specific details for implementing the invention

[0013] Specific structural or functional descriptions of the embodiments are disclosed for illustrative purposes only and may be modified and implemented in various forms. Accordingly, actual implementations are not limited to the specific embodiments disclosed, and the scope of this specification includes modifications, equivalents, or substitutions included in the technical concept described by the embodiments.

[0014] Terms such as "first" or "second" may be used to describe various components, but these terms should be interpreted solely for the purpose of distinguishing one component from another. For example, the first component may be named the second component, and similarly, the second component may be named the first component.

[0015] When it is stated that a component is "connected" to another component, it should be understood that it may be directly connected to or coupled with that other component, or that there may be other components in between.

[0016] Singular expressions include plural expressions unless the context clearly indicates otherwise. In this document, phrases such as “A or B,” “at least one of A and B,” “at least one of A or B,” “A, B or C,” “at least one of A, B and C,” and “at least one of A, B, or C” may each include any one of the items listed together with the corresponding phrase, or all possible combinations thereof. In this specification, terms such as “comprising” or “having” are intended to designate the existence of the described feature, number, step, action, component, part, or combination thereof, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0017] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an ideal or overly formal sense unless explicitly defined in this specification.

[0018] As used herein, the term "module" may include a unit implemented in hardware, software, or firmware, and may be used interchangeably with terms such as logic, logic block, component, or circuit. A module may be a component formed integrally, or a minimum unit of said component or a part thereof that performs one or more functions. For example, according to one embodiment, a module may be implemented in the form of an application-specific integrated circuit (ASIC).

[0019] As used in this document, the term "part" refers to a software or hardware component, such as an FPGA or ASIC, that performs certain roles. However, "part" is not limited to software or hardware. "Part" may be configured to reside in an addressable storage medium or configured to operate one or more processors. For example, "part" may include components such as software components, object-oriented software components, class components, and task components, as well as processes, functions, attributes, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables. The functions provided within the components and "parts" may be combined into a smaller number of components and "parts" or further separated into additional components and "parts." Furthermore, components and "parts" may be implemented to operate one or more CPUs within a device or secure multimedia card. Additionally, '~part' may include one or more processors.

[0020] Hereinafter, embodiments will be described in detail with reference to the attached drawings. In the description with reference to the attached drawings, identical components are given the same reference numeral regardless of the drawing number, and redundant descriptions thereof will be omitted.

[0021] FIG. 1 is a diagram illustrating an overview of a prediction map generation system according to one embodiment.

[0022] The prediction map generation system described in this specification may be a prediction system based on geospatial artificial intelligence (Geo AI) that fuses a Geographic Information System (GIS) and artificial intelligence (AI) to predict in advance the likelihood of sinkholes (ground subsidence, ground collapse) occurring in urban areas. The prediction map generation system described in this specification may generate a machine learning-based ground subsidence susceptibility map based on ground subsidence-related data held by the Korea Institute of Geoscience and Mineral Resources and underground facility and ground databases held by the government / local governments, etc.

[0023] The prediction map generation system for predicting the likelihood of sinkhole occurrence described herein may include a prediction map generation device that performs a prediction map generation method. The prediction map generation device can accurately predict the likelihood of sinkhole occurrence in a specific area by comprehensively analyzing various heterogeneous data, such as information on urban underground structures, geological and soil characteristics, groundwater changes, the condition of aging infrastructure, civil engineering history, and weather conditions. Based on the prediction results, the prediction map generation device can support local governments and urban management agencies in effectively establishing and operating preventive measures in advance.

[0024] The predictive map generation device can perform precise integrated predictions based on Geo AI. By integrating GIS spatial information and machine learning-based analysis, the predictive map generation device can generate a predictive map displaying the probability and risk of sinkhole occurrence. The predictive map generation device can utilize various data. By fusing heterogeneous data such as underground facilities, ground characteristics, and weather information, the predictive map generation device can maximize the accuracy of sinkhole occurrence predictions.

[0025] The predictive map generation device can predict the likelihood of sinkhole occurrence by periodically updating dynamic data, such as continuously changing weather conditions, construction status, and groundwater information. By visualizing risk areas in advance and setting priorities, the device supports the establishment of preemptive and strategic preventive response and management plans. Furthermore, the predictive map generation device is not limited to specific regions and possesses high scalability and universal applicability, as it can be applied to diverse urban environments both domestically and internationally through data-driven expansion.

[0026] The predictive map generation device can collect data on the occurrence of sinkholes by region and data related to sinkhole occurrences by region, and perform data preprocessing on the collected data. The predictive map generation device can predict the probability of sinkhole occurrence by region based on spatial data including at least one of environmental data, geological and ground data, public data, or geospatial data, and can generate a predictive map based on the predicted probability. Here, environmental data may include information on detailed weather conditions such as precipitation, groundwater level, temperature, and humidity. Geological and ground data may include geotechnical information such as soil type, ground strength, and ground subsidence history. Public data may include information on the maintenance status of urban infrastructure and records of past accidents. The predictive map generation device can integrally analyze and process the collected data related to sinkhole occurrences by region using specialized GIS software (e.g., QGIS, ArcGIS Pro). The predictive map generation device can generate a precise predictive map by subdividing the entire urban area into uniform grid-type cells. The predictive map generation device can build an optimized prediction model based on a multi-modeling approach that applies machine learning techniques such as logistic regression, random forest, XG Boost, and deep neural networks. The predictive map generation device can further improve the accuracy of the prediction model that predicts the probability of sinkhole occurrence and reduce the false positive rate by clearly selecting and setting safe areas where the target value is 0—that is, where the value of the data on the occurrence of sinkholes by region is 0. The predictive map generation device can apply advanced oversampling techniques such as SMOTE (Synthetic Minority Over-sampling Technique) and ADASYN (Adaptive Synthetic Sampling Approach) to resolve the imbalance in frequency by class of data.The predictive map generation device can perform quantitative risk assessment and verify prediction accuracy for the clear identification of high-risk areas. The device can provide predictive maps that visualize the probability of sinkhole occurrence—that is, the risk level—and can promote technological scalability by linking the predictive maps with various other services. Through a GIS-based visual interface, the device can provide predictive maps that clearly visualize risk areas where the probability of sinkhole occurrence exceeds a threshold. Based on automated risk alerts and priority management area designation functions, the device can support policy formulation and the optimization of resource allocation for sinkhole response. The device can enhance convenience for practical operations through seamless integration with web-based dashboards and existing local government operational platforms. The device can optimize the predictive model and maximize performance by continuously training the model. The device can continuously improve the predictive performance of the model by periodically training it with data on the occurrence of newly identified sinkholes by region and data related to sinkhole occurrences by region. A prediction map generation device can provide the capability for a prediction model to perform self-adaptive learning and updates based on the influence of external factors such as seasonal changes and climate change.

[0027] Referring to FIG. 1, a prediction map generating device (120) can collect data on whether sinkholes occur by region and data related to sinkhole occurrence by region (110). The data on whether sinkholes occur by region may include information on whether sinkholes have actually occurred by region, and the data related to sinkhole occurrence by region may include information on related factors related to sinkhole occurrence. The prediction map generating device (120) can perform data preprocessing on the data on whether sinkholes occur by region and the data related to sinkhole occurrence by region (110). The prediction map generating device (120) can generate a machine learning-based prediction model (130) using the data on whether sinkholes occur by region and the data related to sinkhole occurrence by region (110) for which data preprocessing has been performed. The prediction map generating device (120) can obtain and output a prediction map (140) indicating the possibility of sinkhole occurrence based on the data related to sinkhole occurrence by region from the generated prediction model (130). The prediction map generating device (120) can supplement the prediction map (140) by overlaying data on whether sinkholes occur by region onto the prediction map (140) which indicates the possibility of sinkhole occurrence.

[0028] FIG. 2 is a flowchart illustrating a method for generating a prediction map according to one embodiment. The method for generating a prediction map can be performed by a prediction map generating device described herein.

[0029] Referring to FIG. 2, in step (210), the prediction map generating device can collect data on whether sinkholes have occurred by region and data related to sinkhole occurrence by region. Here, the data on whether sinkholes have occurred by region may include data indicating whether sinkholes have actually occurred by region classified based on predetermined criteria. More specifically, the data on whether sinkholes have occurred by region may include data indicating whether sinkholes have actually occurred by region corresponding to each pixel on the map. The data on whether sinkholes have occurred by region may have a value of 0 or 1. Regions where sinkholes have occurred may have a value of 1, and regions where sinkholes have not occurred may have a value of 0.

[0030] Data on whether sinkholes have occurred by region can be displayed on a map as shown in Fig. 4. Data on whether sinkholes have occurred by region can also be displayed on a prediction map, which will be described later. Based on the data on whether sinkholes have occurred by region, whether sinkholes have actually occurred by region can be visually displayed on a map or a prediction map. The data on whether sinkholes have occurred by region may take the form of text data or may take the form of visual data displayed on a map as shown in Fig. 4.

[0031] Data related to the occurrence of sinkholes by region may include, for example, at least one of information regarding water pipe density, water pipe distance map, sewer pipe density, sewer pipe distance map, land use map, road network distance map, road network density map, water system network density map, fault distance map, fault density map, schematic soil map, linear structure density map, or linear structure distance map, but is not limited to those described in this specification. Data related to the occurrence of sinkholes by region may further include information regarding relevant factors determined to be related to the occurrence of sinkholes based on predetermined criteria. Data related to the occurrence of sinkholes by region may include information related to relevant factors selected as having high importance based on predetermined criteria in relation to predicting sinkhole occurrence. Data on the occurrence of sinkholes by region and data related to the occurrence of sinkholes by region may correspond to each other based on region.

[0032] In step (220), the prediction map generating device may perform data preprocessing on data regarding the occurrence of sinkholes by region and data related to the occurrence of sinkholes by region. The data preprocessing performed here may include, for example, at least one of removing obstruction data present in the data regarding the occurrence of sinkholes by region and data related to the occurrence of sinkholes by region, removing and verifying missing values, detecting data outliers, normalizing data, or processing unbalanced data to resolve data imbalance in the data regarding the occurrence of sinkholes by region, but is not limited to the description in this specification.

[0033] The prediction map generator can perform imbalanced data processing by adjusting class weights inversely proportional to the frequency of each class based on regional sinkhole occurrence data, or by extracting bootstraps from minority classes determined based on regional sinkhole occurrence data. Here, the classes based on regional sinkhole occurrence data may include classes corresponding to sinkhole occurrence and classes corresponding to non-sinkhole occurrence. Generally, the frequency of the class corresponding to sinkhole occurrence may be lower than the frequency of the class corresponding to non-sinkhole occurrence. Therefore, imbalanced data processing may be necessary to improve the performance of the prediction model. The prediction map generator can adjust class weights inversely proportional to the frequency of each class based on regional sinkhole occurrence data by applying the 'class_weight='balanced'' option in the random forest model. The prediction map generator can oversample regional sinkhole occurrence data corresponding to minority classes using ADASYN. ADASYN can be more useful for processing imbalanced data because it generates more samples between distant data points. Although the present specification describes an example of performing imbalanced data processing using option settings of a Random Forest model and ADASYN, it is not limited to the description in the present specification, and other models such as SMOTE or a combination of several models may be used for imbalanced data processing.

[0034] In step (230), the prediction map generating device or the prediction model generating device can generate a prediction model using data on whether sinkholes occur by region and data related to sinkhole occurrence by region, for which data preprocessing has been performed. The prediction map generating device or the prediction model generating device can generate a prediction model that learns the characteristics of the data related to sinkhole occurrence by region corresponding to the data on whether sinkholes occur by region.

[0035] The prediction model may be a machine learning-based model trained to predict the probability of sinkholes occurring in each region corresponding to each pixel on a map and to generate a prediction map that visualizes the probability of sinkholes occurring in each predicted region on the map. The prediction model may be, for example, a random forest (RF) model that was found to have the highest performance in experimental results, but is not limited to the description in this specification.

[0036] In one embodiment, the prediction model may be trained according to a machine learning process. The machine learning process is a process that enables the prediction model to independently recognize patterns in input data and make predictions using given data.

[0037] In step (240), the prediction map generating device may obtain a prediction map indicating the likelihood of a sinkhole occurring based on data related to sinkhole occurrence by region from a prediction model. The prediction model may generate a prediction map by predicting the likelihood of a sinkhole occurring for each region corresponding to each pixel on the map and displaying the predicted likelihood on each pixel corresponding to the predicted likelihood. The prediction model may generate a prediction map by mapping location information corresponding to each pixel and the likelihood of a sinkhole occurring for each pixel. The likelihood may have a value of, for example, 0 to 1, but is not limited to the description in this specification. A likelihood of 0 corresponds to a case where the likelihood of a sinkhole occurring is 0%, and a likelihood of 1 corresponds to a case where the likelihood of a sinkhole occurring is 100%. Based on whether a sinkhole has occurred by region, the prediction model may generate a prediction map in which regions where sinkholes have actually occurred and regions where sinkholes have not occurred are distinguished and displayed.

[0038] A prediction model can generate a prediction map by displaying areas in predetermined colors based on the magnitude of the value of the probability of a sinkhole occurring. For example, pixels corresponding to the predicted probability can be displayed in predetermined colors corresponding to the category based on the category containing the value of the predicted probability. The predetermined colors may include, for example, at least one of blue, yellow, or red, but are not limited to those described in this specification. For example, the prediction model may display the corresponding pixel on the prediction map in blue as the magnitude of the value of the probability of a sinkhole occurring is smaller, and may display the corresponding pixel on the prediction map in red as the magnitude of the value of the probability of a sinkhole occurring is larger. The prediction model may display the corresponding pixel on the prediction map in yellow when the magnitude of the value of the probability of a sinkhole occurring is moderate based on a predetermined criterion. Based on whether sinkholes have occurred by region, the prediction map may distinguish and display areas where sinkholes have actually occurred and areas where sinkholes have not occurred.

[0039] The prediction map generated by the prediction model may be, for example, as shown in Fig. 9 or Fig. 10.

[0040] FIG. 3 is a diagram illustrating a method for generating a prediction map according to another embodiment.

[0041] Referring to FIG. 3, the prediction map generation method may include a step (310) of generating spatial data and a step (320) of processing data.

[0042] Spatial data may include, for example, digital topographic maps within a city, underground facilities (water supply, sewage, power conduits, etc.), urban development history, and information on infrastructure status. The step (310) of generating spatial data may include steps (311) through (315).

[0043] In step (311), the prediction map generating device may investigate literature related to sinkholes and select a target city to serve as a testbed. The prediction map generating device may select a target city based on the amount of data, the feasibility of collection, etc. In step (312), the prediction map generating device may collect data on the occurrence of sinkholes by region and data related to sinkhole occurrences by region. Data on the occurrence of sinkholes by region and data related to sinkhole occurrences by region may be collected from at least one of the respective relevant institutions or open platforms. Data on the occurrence of sinkholes by region may be a target or target data, and data related to sinkhole occurrences by region may be a feature or feature data. Among the data related to sinkhole occurrences by region, data related to underground facilities may be core data.

[0044] In steps (313) and (314), the prediction map generating device can generate a spatial information system (GIS) database. In step (313), the prediction map generating device can construct a first GIS database based on data regarding whether sinkholes occur by region. The prediction map generating device can construct a first GIS database based on sinkhole occurrence area data and sinkhole non-occurrence area data included in the data regarding whether sinkholes occur by region. Here, the sinkhole non-occurrence area data may be arbitrarily selected from among the sinkhole non-occurrence areas for the entire remaining area excluding a buffer area with a radius of 1 km from the sinkhole occurrence area.

[0045] In step (314), the prediction map generating device may construct a second GIS database based on regional sinkhole occurrence related data. The prediction map generating device may construct a second GIS database based on regional sinkhole occurrence related data that includes information on related factors determined to be related to the occurrence of sinkholes based on predetermined criteria. The related factors determined to be related to the occurrence of sinkholes may include, for example, 45 factors, but are not limited to those described in this specification.

[0046] In step (315), the prediction map generating device distinguishes each region into a sinkhole occurrence region and a sinkhole non-occurrence region and can extract region-specific sinkhole occurrence data from each region. In step (316), the prediction map generating device can remove data from invalid regions, such as landfills, from the region-specific sinkhole occurrence data. In this case, the number of sinkhole non-occurrence region data can be changed from 1,000 to 818.

[0047] The prediction map generation device can generate spatial data by converting the first GIS database and the second GIS database into CSV files in step (317). Spatial data may also be referred to as a spatial dataset. The step (310) of generating spatial data can be performed, for example, using ArcGIS Pro.

[0048] The step of processing data (320) may include steps (321) through (326). The step of processing data (320) may be performed using a statistical model.

[0049] In step (321), the prediction map generating device can perform exploratory data analysis of data regarding the occurrence of sinkholes by region and data related to sinkhole occurrences by region according to the purpose definition. The prediction map generating device can perform exploratory data analysis by determining whether the data regarding the occurrence of sinkholes by region and data related to sinkhole occurrences by region are categorical data or continuous data according to the purpose definition. In step (322), the prediction map generating device can perform data preprocessing on the data regarding the occurrence of sinkholes by region and data related to sinkhole occurrences by region.

[0050] Data preprocessing may include removing specific factors, including obstacle data, from regional sinkhole occurrence-related data, removing and verifying missing values ​​from regional sinkhole occurrence data and regional sinkhole occurrence-related data, and detecting data outliers among regional sinkhole occurrence data and regional sinkhole occurrence-related data.

[0051] In step (323), the prediction map generating device divides at least a portion of the data on whether sinkholes occur by region and the data related to sinkhole occurrence by region into training data and validation data, and can examine the statistical significance of the data on whether sinkholes occur by region and the data related to sinkhole occurrence by region. Additionally, in step (324), the prediction map generating device can perform standardization of the continuous data included in the data on whether sinkholes occur by region and the data related to sinkhole occurrence by region using, for example, StandardScalar.

[0052] In step (325), the prediction map generating device may perform imbalance data processing on regional sinkhole occurrence data and regional sinkhole occurrence related data using at least one of SMOTE or ADASYN. The imbalance data processing performed in step (325) may correspond to the imbalance data processing described in FIG. 2.

[0053] In step (326), the prediction map generating device performs Principal Component Analysis (PCA) of the data through data visualization of the data on whether sinkholes occur in each region and the data related to sinkhole occurrence in each region, and accordingly, can verify the usability of the data on whether sinkholes occur in each region and the data related to sinkhole occurrence in each region.

[0054] FIG. 4 is a diagram illustrating an example of data on whether sinkholes occur by region according to one embodiment.

[0055] Referring to FIG. 4, the data on whether sinkholes occur by region may take the form of visual data in which the occurrence of sinkholes by region is displayed on a map in one example. The data on whether sinkholes occur by region in the form of visual data may be distinguished and displayed as sinkhole occurrence areas and sinkhole non-occurrence areas for each region corresponding to each pixel on the map. For example, sinkhole occurrence areas may be indicated as red dots on the map, and sinkhole non-occurrence areas may be indicated as gray dots on the map, but are not limited to the descriptions in this specification.

[0056] FIGS. 5 to 8 are drawings illustrating an example of data related to the occurrence of sinkholes by region according to one embodiment.

[0057] Regional sinkhole occurrence data, comprising dozens of key factors, includes information on weather conditions, construction status, water pipe density, water pipe distance, sewer pipe density, sewer pipe distance, land use, road network distance, road network density, drainage network density, fault distance, fault density, schematic soil map, prestructure density, or prestructure distance, as well as geological type groups 4, geological grade 5 for ground subsidence vulnerability, soil type, soil drainage grade, soil parent material, topographic classification, effective soil depth, schematic soil map, morphological features, curvature classification, lithology type, forest diameter class, forest age class, forest density, forest classification, flow path length, drainage network distance, prestructure analysis density, prestructure analysis distance, road network analysis density, road network analysis distance, fault analysis distance, fault analysis density, slope, groundwater level_LDensity, and groundwater level_DistanceM. It may also include additional information. Here, each piece of information may be referred to as a related factor, feature, or factor.

[0058] Figure 5 may be a map in which information regarding geological type group 4 is visualized via GIS. Geological type group 4 may include granite series, volcanic rock series, sedimentary rock series, and alluvial deposits. Each group may be distinguished and displayed based on color or shading.

[0059] FIG. 6 may be a map in which information regarding soil drainage grades is visualized via GIS. Soil drainage grades may include, for example, seven grades based on predetermined criteria, but are not limited to those described in this specification. Each grade of soil drainage grade may be distinguished and displayed based on color or shading.

[0060] Figure 7 may be a map in which information regarding slope is visualized via GIS. The slope may be displayed on the map in a predetermined color based on the range to which the value belongs. On the map of Figure 7, areas marked in red may have a steep slope, and areas marked in blue may have a gentle slope.

[0061] Figure 8 may be a map in which information regarding sewer pipe density is visualized via GIS. Sewer pipe density may be displayed on the map in predetermined colors based on the range to which the value belongs. On the map of Figure 8, areas marked in red may have high sewer pipe density, and areas marked in blue may have low sewer pipe density.

[0062] FIGS. 9 and FIGS. 10 are drawings illustrating an example of a prediction map according to one embodiment.

[0063] Referring to FIG. 9, the prediction map may have a two-dimensional form. In the prediction map, areas where sinkholes do not occur and areas where sinkholes occur may be marked with dots of different colors. The prediction map may display a sinkhole risk level corresponding to the probability of sinkhole occurrence described in this specification based on color differences or shading differences. The sinkhole risk level may indicate that a larger value indicates a higher probability of sinkhole occurrence, and a smaller value indicates a lower probability of sinkhole occurrence. The sinkhole risk level may have a value of, for example, 0 to 0.95, and may be displayed in a lighter color on the prediction map as the value is closer to 0, and in a darker color as the value is closer to 0.95.

[0064] Referring to Fig. 10, the prediction map can have a three-dimensional shape.

[0065] The prediction map may display a sinkhole risk corresponding to the probability of a sinkhole occurring as described in this specification based on a difference in color or shade. The sinkhole risk may indicate that a larger value indicates a greater probability of a sinkhole occurring, and a smaller value indicates a lower probability of a sinkhole occurring. The sinkhole risk may have a value of, for example, 0 to 0.95, and may be displayed in blue on the prediction map as the value is closer to 0, and may be displayed in red on the prediction map as the value is closer to 0.95.

[0066] Although the present specification describes a method for generating a prediction map based on a map of the Busan region, the selection of the region is merely an example and is not limited to the embodiments of the present specification. Furthermore, the prediction maps of FIGS. 9 and FIGS. 10 are merely examples and are not limited to the embodiments of the present specification and may have other forms.

[0067] FIG. 11 is a diagram illustrating the configuration of a prediction map generation device according to one embodiment.

[0068] Referring to FIG. 11, the prediction map generating device (1100) may include a processor (1110), memory (1120), and a database (1130). The prediction map generating device (1100) may correspond to the prediction map generating device described herein.

[0069] The memory (1120) is connected to the processor (1110) and can store instructions executable by the processor (1110), data to be computed by the processor (1110), or data processed by the processor (1110). The memory (1120) may include a non-transient computer-readable medium, such as high-speed random access memory and / or a non-volatile computer-readable storage medium (e.g., one or more disk storage devices, flash memory devices, or other non-volatile solid-state memory devices).

[0070] The database (1130) may store the first GIS database and the second GIS database described in this specification. Additionally, the database (1130) may store data necessary for a prediction map generating device to perform a prediction map generating method.

[0071] The processor (1110) may perform one or more operations related to the operation of the prediction map generating device described herein. For example, the processor (1110) may cause the prediction map generating device (1100) to collect data on the occurrence of sinkholes by region and data related to the occurrence of sinkholes by region, and to perform data preprocessing on the data on the occurrence of sinkholes by region and data related to the occurrence of sinkholes by region. The processor (1110) may cause the prediction map generating device (1100) to perform imbalanced data processing by adjusting class weights inversely proportional to the frequency of each class based on the data on the occurrence of sinkholes by region, or by extracting bootstraps from minority classes determined based on the data on the occurrence of sinkholes by region.

[0072] The processor (1110) can enable the prediction map generating device (1100) to generate a prediction model using data on whether sinkholes occur in each region and data related to sinkhole occurrence in each region, for which data preprocessing has been performed, and to obtain a prediction map from the prediction model that indicates the possibility of sinkhole occurrence based on the data related to sinkhole occurrence in each region.

[0073] The hardware device described above may be configured to operate as one or more software modules to perform the operation of the embodiment, and vice versa.

[0074] Although the embodiments have been described above with reference to the limited drawings, those skilled in the art can apply various technical modifications and variations based thereon. For example, suitable results may be achieved even if the described techniques are performed in a different order than described, and / or if the components of the described system, structure, device, circuit, etc. are combined or assembled in a form different from described, or replaced or substituted by other components or equivalents.

[0075] Therefore, other implementations, other embodiments, and equivalents to the claims also fall within the scope of the claims set forth below. Explanation of the symbols

[0076] 120, 1100: Predictive map generation device 1110: Processor 1120: Memory 1130: Database

Claims

Claim 1 A method for generating a prediction map for predicting the probability of sinkhole occurrence performed by a prediction map generation device comprises: a step of collecting data on the occurrence of sinkholes by region and data related to sinkhole occurrences by region; a step of performing data preprocessing on the data on the occurrence of sinkholes by region and the data related to sinkhole occurrences by region; a step of generating a prediction model using the data on the occurrence of sinkholes by region and the data related to sinkhole occurrences by region for which data preprocessing has been performed; and a step of obtaining a prediction map indicating the probability of sinkhole occurrence based on the data related to sinkhole occurrences by region from the prediction model, wherein the step of performing data preprocessing includes a step of performing imbalanced data processing to resolve imbalances in frequency by class by extracting bootstraps from minority classes determined based on the data on the occurrence of sinkholes by region, and wherein the data related to sinkhole occurrences by region includes a water pipe density map, a water pipe distance map, a sewer pipe density map, a sewer pipe distance map, a land use map, a road network distance map, a road network density map, a water system network density map, a fault distance map, a fault density map, a rough soil map, a linear structure density map, or a linear structure A method for generating a prediction map that includes factors and information selected as having high importance based on predetermined criteria in relation to predicting sinkhole occurrence among information regarding distance maps. Claim 2 A method for generating a prediction map according to claim 1, wherein the step of performing the data preprocessing further includes the step of performing imbalanced data processing by adjusting class weights inversely proportional to the frequency of each class based on the data on whether sinkholes occur by region. Claim 3 A method for generating a prediction map according to claim 1, wherein the data on whether a sinkhole occurs by region includes data indicating whether a sinkhole has actually occurred in each region corresponding to each pixel on a map. Claim 4 A method for generating a prediction map according to claim 1, wherein the prediction model is a machine learning-based model trained to predict the probability of a sinkhole occurring in each region corresponding to each pixel on a map, based on the data regarding the occurrence of a sinkhole by region and the data related to the occurrence of a sinkhole by region for which the data preprocessing has been performed. Claim 5 A method for generating a prediction map according to claim 1, wherein the prediction model predicts the probability of a sinkhole occurring for each region corresponding to each pixel on a map, and generates the prediction map by displaying the predicted probability on each pixel corresponding to the predicted probability. Claim 6 A method for generating a prediction map according to claim 1, wherein the prediction model generates the prediction map by mapping location information corresponding to each pixel and the probability of the sinkhole occurring for each pixel. Claim 7 A method for generating a prediction map according to claim 1, wherein the prediction model generates the prediction map by displaying it in a predetermined color based on the magnitude of the value of the probability of the sinkhole occurring. Claim 8 A method for generating a prediction map according to claim 1, wherein the data related to sinkhole occurrence by region comprises at least one of information regarding water pipe density, water pipe distance map, sewer pipe density map, sewer pipe distance map, land use map, road network distance map, road network density map, water system network density map, fault distance map, fault density map, schematic soil map, linear structure density map, or linear structure distance map. Claim 9 A prediction map generation device for predicting the probability of sinkhole occurrence comprises a memory and a processor, wherein the memory stores instructions executable by the processor, and when the instructions are executed by the processor, the processor enables the prediction map generation device to collect data on the occurrence of sinkholes by region and data related to sinkhole occurrences by region, perform data preprocessing on the data on the occurrence of sinkholes by region and data related to sinkhole occurrences by region, generate a prediction model using the data on the occurrence of sinkholes by region and data related to sinkhole occurrences by region for which data preprocessing has been performed, and obtain a prediction map indicating the probability of sinkhole occurrence based on the data related to sinkhole occurrences by region from the prediction model, wherein the data preprocessing includes imbalanced data processing that resolves the imbalance of frequency by class by extracting bootstraps from a minority class determined based on the data on the occurrence of sinkholes by region, and the data related to sinkhole occurrences by region includes water pipe density, water pipe distance map, sewer pipe density, sewer pipe distance map, land use map, A prediction map generating device comprising factors and information selected as having high importance based on predetermined criteria in relation to predicting sinkhole occurrence among information regarding road network distance maps, road network density maps, water system density maps, fault distance maps, fault density maps, schematic soil maps, linear structure density maps, or linear structure distance maps. Claim 10 delete

Images