KR102042645B1 - Apparatus and method for predicting line risk using compound model - Google Patents

Abstract

The present invention relates to a line risk prediction apparatus and method using a complex model. According to the present invention, the line risk prediction apparatus may comprise the steps of: collecting new data which is subject to line risk prediction; performing preprocessing of the new data; generating a binomial classification model for predicting a line risk for the new data based on a dataset including at least one of accident failure data, track defects, maintenance data, weather data, and facility data; generating a polynomial classification model for predicting the line risk for the new data based on the dataset; and predicting the line risk by combining the binomial classification model and the polynomial classification model.

Description

Apparatus and method for predicting track risk using complex model {APPARATUS AND METHOD FOR PREDICTING LINE RISK USING COMPOUND MODEL}

The present invention relates to an apparatus and method for predicting track risk using a complex model, and more particularly, to a model development and method for predicting track risk using a big data-based machine learning algorithm.

Railroad accidents are one of the most serious accidents that cause great casualties and huge losses in one accident. However, in the related art, the level of development of the risk analysis technology of the railway accident is not appropriate.

For example, there is a conventional technique for analyzing a risk of an accident based on statistical techniques. However, statistical techniques-based accident risk analysis techniques have limitations in accurately predicting line hazards.

The present invention is to solve the problems of the prior art, it is an object of the present invention to provide a method for developing a track risk prediction model that can effectively prevent the risk of the track in advance by more accurately predicting the risk of the track.

However, the technical problem to be achieved by the embodiments of the present application is not limited to the technical problems as described above, and other technical problems may exist.

As a technical means for achieving the above technical problem, the line risk prediction method according to an embodiment of the present application, collecting the new data that is the target of the line risk prediction; Performing preprocessing of the new data; Generating a binomial classification model for predicting track risk for the new data based on a dataset including at least one of accident failure data, track defects, maintenance data, weather data, and facility data; Generating a polynomial classification model for predicting track risk for new data based on the data set; Combining the binomial classification model and the polynomial classification model may include estimating a line risk.

According to one embodiment of the present application, the new data may be track-related data associated with at least one of accident failure data, track defects, maintenance data, weather data, and facility data.

According to the exemplary embodiment of the present application, the data set includes a plurality of records, and the records include accident failure data, line defects, maintenance data, weather data, and facility data based on dates, keystrokes, and up / down codes. It may be to connect.

According to one embodiment of the present application, the step of performing the pre-processing of the new data, the data collector to search the collected data to obtain the data to be analyzed, select the analysis variable to perform the analysis from the obtained data, It may be to perform preprocessing of data corresponding to the selected variable.

According to one embodiment of the present application, the step of performing the pre-processing of the new data may be to normalize the variable values of the data while minimizing the influence of the outliers using the median value and the interquartile range (IQR) using the robust method. .

According to an embodiment of the present disclosure, generating the binomial classification model may include: generating a risk prediction model by setting a binomial risk level value as a dependent variable and setting the selected variable as an independent variable; Learning a generated risk prediction model by applying a plurality of machine learning algorithms; Computing the line risk prediction value by the plurality of machine learning algorithms.

According to one embodiment of the present application, the plurality of machine learning algorithms may include at least one of Random Forest, Logistic Regression, XGBoost Classifier, Balanced Bagging Classifier, Ensemble.

According to an embodiment of the present disclosure, the generating of the polynomial classification model may include generating a risk prediction model by setting the polynomial risk level value as a dependent variable and setting the selected variable as an independent variable; Learning a generated risk prediction model by applying a plurality of machine learning algorithms; Computing the line risk prediction value by the plurality of machine learning algorithms.

According to one embodiment of the present application, the plurality of machine learning algorithms may include at least one of K-Nearest Neighbor (knn), Support Vector Machine (svm), XGBoost Classifier, Balanced Bagging Classifier, Adaboost.

According to one embodiment of the present application, the step of predicting the line risk by combining the binomial classification model and polynomial classification model, in the optimal algorithm selected in consideration of the accuracy and penalty in the complex model plus the binomial classification model and polynomial classification model The derived result may be provided as a track risk prediction result.

As a technical means for achieving the above technical problem, a line risk predicting apparatus according to an embodiment of the present application, the data collection unit for collecting new data that is the target of the line risk prediction, the pre-processing of the data collected by the data collection unit Binomial to generate a binomial classification model for predicting track risk for new data based on a data set including at least one of a data preprocessor, accident failure data, line fault, maintenance data, weather data, and facility data. A classification model unit, a polynomial classification model unit for generating a polynomial classification model for predicting a track risk for new data based on the data set, and a risk prediction unit for predicting a line risk by combining the binomial classification model and the polynomial classification model. can do.

The above-mentioned means for solving the problems are merely exemplary and should not be construed as limiting the present application. In addition to the above-described exemplary embodiments, additional embodiments may exist in the drawings and detailed description of the invention.

According to the above-described problem solving means of the present application, the contents of the present invention anticipate the risk of accident in advance through the developed line risk prediction model to expect a proactive prevention effect. In addition, it is possible to create value by applying, analyzing and utilizing the core technology development results in other industries.

1 is a diagram illustrating a schematic configuration of an apparatus for predicting a track risk according to an exemplary embodiment of the present application.
2 is a diagram illustrating data for selecting important variables in a track risk predicting apparatus according to various embodiments of the present disclosure.
3 is an exemplary view summarizing the binomial classification prediction accuracy calculated for each cluster in the line risk prediction apparatus according to various embodiments of the present disclosure.
4 is an exemplary view summarizing the accuracy of polynomial classification prediction calculated for each cluster in the track risk prediction apparatus according to various embodiments of the present disclosure.
5 is a flowchart illustrating a method for predicting a track risk according to various embodiments of the present disclosure.

DETAILED DESCRIPTION Hereinafter, exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings so that those skilled in the art may easily implement the present disclosure. As those skilled in the art would realize, the described embodiments may be modified in various different ways, all without departing from the spirit or scope of the present invention. In the drawings, parts irrelevant to the description are omitted for simplicity of explanation, and like reference numerals designate like parts throughout the specification.

Throughout this specification, when a part is "connected" to another part, it is not only "directly connected" but also "electrically connected" or "indirectly connected" with another element in between. "Includes the case.

Throughout this specification, when a member is said to be located on another member "on", "upper", "top", "bottom", "bottom", "bottom", this means that any member This includes not only the contact but also the presence of another member between the two members.

Throughout this specification, when a part is said to "include" a certain component, it means that it can further include other components, without excluding the other components unless specifically stated otherwise.

Hereinafter, an apparatus and method for predicting track risk of the present application will be described with reference to FIGS. 1 to 5.

1 is a diagram illustrating a schematic configuration of an apparatus for predicting a track risk according to an exemplary embodiment of the present application.

Referring to FIG. 1, the line risk predicting apparatus 10 includes a data collecting unit 11, a data preprocessor 12, a binomial classification model unit 13, a polynomial classification model unit 14, and a risk prediction unit 15. May include

The data collection unit 11 may collect new data that is a target of track risk prediction. In this case, the new data may be track-related data related to at least one of accident failure data, track defects, maintenance data, weather data, and facility data.

According to one embodiment of the present application, the data collector 11 may collect train-related data from an external server through a network. Examples of networks for sharing information between the data collector 11 and external servers include 3rd Generation Partnership Project (3GPP) networks, Long Term Evolution (LTE) networks, 5G networks, World Interoperability for Microwave Access (WIMAX) networks, and wired and wireless networks. Internet, Local Area Network (LAN), Wireless Local Area Network (LAN), Wide Area Network (WAN), Personal Area Network (PAN), Bluetooth (Bluetooth) Network, Wifi Network, Near Field Communication (NFC) Network , Satellite broadcasting network, analog broadcasting network, digital multimedia broadcasting (DMB) network, and the like, but are not limited thereto.

At this time, the line risk prediction apparatus according to an embodiment of the present application of the present application may include a data set generating unit for generating a data set. Datasets can integrate and store accident failure data, track defects, maintenance data, weather data, and facility data. According to this, the new data may be included in the data set and stored.

Here, the accident failure data may be referred to as track accident data, railway accident data, train accident data, railway accident data, and the like as data related to past train accidents and train failures associated with the track. The maintenance data may include information such as data for repairing a defect in a line and data repaired to prevent a defect. The weather data may include information such as weather, temperature, day crossings, wind speed, precipitation, etc. which may cause track defects, distortions, and deformations. Facility data may include information such as tunnels, bridges, and the like. However, the components included in the data set are not limited to those described above, and may further include information related to each data.

In addition, a dataset may constitute a plurality of records. Here, the record may be generated by linking accident failure data with line defects, maintenance data, weather data, and facility data based on date, key, and up / down code. In other words, in the present application, in order to predict the risk of tracks, accident failure data, track defects, maintenance data, weather data, and facility data may be stored in a dataset for each cluster by connecting based on date, key path, and up / down code. . At this time, one data stored in the data set connected based on the date, key definition, and up / down division code may be referred to as one record.

The data preprocessor 12 may perform preprocessing of new data. In addition, the data preprocessor 12 searches for data collected by the data collector 11 to acquire data to be analyzed, selects an analysis variable to be analyzed from the acquired data, and corresponds to the selected variable. It may be to perform a pretreatment of. That is, the data preprocessor 12 may perform data normalization to match variable values of data stored in the data set on a predetermined basis.

In addition, the data preprocessing unit 12 may calculate the severity in consideration of the accident damage size and the frequency of each factor of the accident. In particular, the preprocessor 12 may calculate the severity for each record in the data set by multiplying the magnitude of the accident damage and the frequency of each accident factor.

In other words, the severity can be estimated by considering the frequency of accidents and the magnitude of the accident damage. In this case, the preprocessing unit 12 may calculate the severity for each record in the data set by multiplying the frequency of accidents by the damage magnitude (accident damage magnitude) of each accident. Here, the magnitude of the accident damage may be calculated by the magnitude of the physical damage caused during the accident associated with the track. In addition, accident damage can be calculated by adding the amount of operating damage to the amount of material damage incurred during the track-related accident. In other words, the accident damage size can be calculated by considering the amount of business damage in addition to the physical damage caused during the track-related accident.

In addition, the preprocessing unit 12 may calculate the severity in consideration of the delay time in addition to the magnitude of the accident damage, the frequency of the accident. Here, the delay time may mean a train delay time due to an accident related to the track.

The data preprocessor 12 may minimize the influence of an outlier by using a median value and an interquartile range (IQR) using a robust method, and normalize variable values of data. In detail, the data preprocessor 12 may generate an irregular portion of the data in the data set, that is, a non-uniform variable value, to improve the stability and accuracy of the risk prediction model and effectively reduce the error of the data stored in the data set. Having data can be normalized using the robust method. The data preprocessor 12 may normalize using a robust method to minimize the influence of an outlier using a median and an interquartile range (IQR) and normalize variable values of data stored in the dataset.

In addition, the preprocessor 12 may select at least one variable for generating a line risk prediction model based on normalized data (ie, normalized data stored in a dataset). In other words, the data preprocessor 12 may select a variable (important variable) necessary for predicting a track risk based on normalized data. The selection of these variables can be made to use only influential variables in the track risk prediction model in the generation of the track risk prediction model.

According to an embodiment of the present disclosure, the preprocessor 12 may select at least one variable for generation of a risk prediction model based on normalized data (ie, normalized data stored in a dataset). In other words, the preprocessor 12 may select a variable (important variable) necessary for predicting the track risk based on the normalized data. Selection of these variables can be made to use only those variables that are influential in the risk prediction model in the generation of the risk prediction model (model).

2 is a diagram illustrating data for selecting important variables in a track risk predicting apparatus according to various embodiments of the present disclosure.

Referring to FIG. 2, the important variables stored in the data set are normalized data, for example, maximum temperature, daily crossover, three-day average wind speed, maximum wind speed, tunnel presence, maximum precipitation per hour, distortion classification, season, and bridge existence. It may include at least one variable, the amount of precipitation, lasting 33 degrees _3 days, the face, and may include the importance for each variable.

The binomial classification model unit 13 generates a binomial classification model for predicting a track risk for the new data based on a data set including at least one of accident failure data, line defect, maintenance data, weather data, and facility data. can do.

According to an embodiment of the present application, the binomial classification model unit 13 sets a binomial risk level value as a dependent variable, sets the selected variable as an independent variable, generates a risk prediction model, and applies a plurality of machine learning algorithms. In addition, the generated risk prediction model may be learned, and a line risk prediction value may be calculated by the plurality of machine learning algorithms. The model herein can be represented otherwise as a model.

The plurality of machine learning algorithms may include at least one of Random Forest, Logistic Regression, XGBoost Classifier, Balanced Bagging Classifier, and Ensemble. The Random Forest algorithm is a method that multiple decision trees form a forest and average each prediction result as a single result variable. The Logistic Regression algorithm applies a Sigmoid function to a linear model used for linear prediction. An algorithm that solves a classification problem. The Xgboost algorithm is a boost-type algorithm that applies the results to the next tree, while the tree of the random forest is independent. The balanced bagging classifier algorithm collects samples several times and trains each model to aggregate the results.

3 is an exemplary view summarizing the binomial classification prediction accuracy calculated for each cluster in the line risk prediction apparatus according to various embodiments of the present disclosure.

Referring to FIG. 3, the binomial classification model unit 13 may classify the binomial risk level into two types. For example, binomial risk levels can be classified as attention or risk. In addition, each cluster stored in the dataset may represent an interest and a risk value.

The binomial classification model unit 13 applies a plurality of machine learning algorithms to each of the generated line risk prediction models, so that the binomial classification model unit 13 may calculate a line risk prediction value corresponding to each of the plurality of machine learning algorithms. have. In other words, the generated line risk prediction model is applied to each of the plurality of machine learning algorithms, thereby predicting the line risk as a result of the application of the machine learning algorithm corresponding to each of the plurality of machine learning algorithms.

The binomial classification model unit 13 may select the risk prediction model to which the optimal machine learning algorithm is applied as the line risk prediction model in consideration of the accuracy and the penalty among the plurality of machine learning algorithms based on the learning result.

The polynomial classification model unit 14 may generate a polynomial classification model for predicting a line risk for new data based on the data set.

According to one embodiment of the present application, the polynomial classification model unit 14 sets the polynomial risk level value as a dependent variable, sets the selected variable as an independent variable, generates a risk prediction model, and applies a plurality of machine learning algorithms. In addition, the generated risk prediction model may be learned, and a line risk prediction value may be calculated by the plurality of machine learning algorithms. That is, the risk prediction model generated by the polynomial classification model 14 can be learned by applying a plurality of machine learning algorithms.

The plurality of machine learning algorithms may include at least one of K-Nearest Neighbor (knn), Support Vector Machine (svm), XGBoost Classifier, Balanced Bagging Classifier, and Adaboost. The K-Nearest Neighbor algorithm classifies by referring to the labels of K different data that are close to the data, and the Support Vector Machine measures the distance between each data to find the center between the two data and obtain the optimal Hyper Plane. By dividing the two categories. Adaboost is an algorithm that improves performance by reflecting the error of previous weak model result to the weight of other weak model.

4 is an exemplary view summarizing the accuracy of polynomial classification prediction calculated for each cluster in the track risk prediction apparatus according to various embodiments of the present disclosure.

Referring to FIG. 4, the polynomial classification model unit 14 may divide the polynomial risk level into a plurality of terms. For example, polynomial risk level values can be divided into attention, boundary, and severity. In addition, the line risk prediction model generated by the polynomial classification model 14 is applied to each of the plurality of machine learning algorithms, so that the line risk prediction value may be calculated corresponding to each of the plurality of machine learning algorithms. In other words, the generated line risk prediction model is applied to each of the plurality of machine learning algorithms, thereby predicting the line risk as a result of the application of the machine learning algorithm corresponding to each of the plurality of machine learning algorithms.

Based on the learning results, the polynomial classification model unit 14 may select a risk prediction model to which an optimal machine learning algorithm is applied as a line risk prediction model in consideration of accuracy and penalty among a plurality of machine learning algorithms. If the K-Nearest Neighbor (knn) algorithm is 70% accurate, the Support Vector Machine (svm) algorithm is 80% accurate, and the penalty of the two algorithms is the same, then the polynomial classification model 14 provides high-precision support. The risk prediction model to which the vector machine (svm) algorithm is applied can be selected as the line risk prediction model.

The risk prediction unit 15 may predict the line risk by combining the binomial classification model and the polynomial classification model. At this time, the risk prediction unit 15 may predict the line risk as one of a plurality of terms of the polynomial risk level value. have. For example, the risk prediction unit 15 may predict the line risk as one of caution, boundary, and severity.

According to an exemplary embodiment of the present application, the risk prediction unit 15 provides a result value obtained from an optimal algorithm selected in consideration of accuracy and penalty in a complex model including a binomial classification model and a polynomial classification model as a line risk prediction result. It may be. The risk prediction unit 15 generates an algorithm that combines the line risk prediction model selected by the binomial classification model unit 13 and the line risk prediction model selected by the polynomial classification model unit 14, and converts the result value derived from the algorithm into a line. It may be provided as a risk prediction result.

The risk prediction unit 15 may predict the line risk based on a complex model including a binomial classification model and a polynomial classification model. For example, a binary risk level value can be predicted based on a track risk prediction model generated based on a balanced bagging classifier algorithm. In addition, it is possible to predict the track risk level value based on the multinomial track risk prediction model generated based on the K-Nearest Neighbor algorithm. The risk prediction unit 15 may provide the result value of the optimal algorithm selected from the optimal algorithm selected by considering the accuracy and penalty in the complex model including the binomial classification model and the polynomial classification model as the line risk prediction result of the corresponding key definition.

The track risk predicting apparatus of the present application may be configured to store big data in a data set by linking accident failure data with line defects. In addition, the present application can be machine learning algorithm to predict the line risk. In other words, the present application can predict the risk of the track using a big data-based machine learning algorithm, thereby achieving a more accurate and improved prediction rate.

5 is a flowchart illustrating a method for predicting a track risk according to various embodiments of the present disclosure.

The track risk predicting method shown in FIG. 5 may be performed by the track risk predicting apparatus 10 described above. Therefore, even if omitted below, the content described with respect to the line risk predicting apparatus 10 may be equally applicable to the description of the line risk predicting method.

Referring to FIG. 5, the line risk prediction method may collect new data that is a target of line risk prediction in the line risk prediction device. (S100)

The new data may be track related data associated with at least one of accident failure data, track defects, maintenance data, weather data, and facility data.

Next, preprocessing of the new data collected in step S100 may be performed. (S200)

In operation S200, the data collected by the data collecting unit 11 searches for data to be analyzed, selects an analysis variable to be analyzed from the acquired data, and performs preprocessing of data corresponding to the selected variable. can do.

In step S200, the preprocessing may be to normalize the variable values of the data while minimizing the influence of the outliers using the median value and the interquartile range (IQR) using the robust method.

Next, a binomial classification model for predicting a line risk for the new data may be generated based on a data set including at least one of accident failure data, track defects, maintenance data, weather data, and facility data. (S300)

At this time, the data set includes a plurality of records, the record may be to connect the accident failure data, track defects, maintenance data, weather data, facility data based on the date, keystroke and vertical division code.

In step S300, the risk prediction model may be generated by setting the binomial risk level value as a dependent variable and setting the selected variable as an independent variable, and applying the plurality of machine learning algorithms to learn the generated risk prediction model. In operation S300, a line risk prediction value may be calculated by the plurality of machine learning algorithms.

In this case, the plurality of machine learning algorithms may include at least one of Random Forest, Logistic Regression, XGBoost Classifier, Balanced Bagging Classifier, and Ensemble.

Next, a polynomial classification model for predicting line risk for new data may be generated based on the data set. (S400)

In step S400, the risk prediction model may be generated by setting the polynomial risk level value as the dependent variable and the selected variable as the independent variable, and applying the plurality of machine learning algorithms to learn the generated risk prediction model. The method may further include calculating a line risk prediction value by the plurality of machine learning algorithms.

In this case, the plurality of machine learning algorithms may include at least one of K-Nearest Neighbor (knn), Support Vector Machine (svm), XGBoost Classifier, Balanced Bagging Classifier, and Adaboost.

Next, the line risk may be predicted by combining the binomial classification model and the polynomial classification model. (S500)

In step S500, the line risk is predicted by combining the binomial classification model and the polynomial classification model, and the resultant value derived from the optimal algorithm selected by considering the accuracy and penalty in the complex model including the binomial classification model and the polynomial classification model. Can be provided as a prediction result.

In the above description, steps S100 to S500 may be further divided into additional steps or combined into fewer steps, according to an embodiment of the present disclosure. In addition, some steps may be omitted as necessary, and the order between the steps may be changed.

The track risk predicting method performed by the track risk predicting apparatus according to an exemplary embodiment of the present disclosure may be implemented in the form of program instructions that may be executed by various computer means and recorded in a computer readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. Program instructions recorded on the media may be those specially designed and constructed for the purposes of the present invention, or they may be of the kind well-known and available to those having skill in the computer software arts. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tape, optical media such as CD-ROMs, DVDs, and magnetic disks, such as floppy disks. Magneto-optical media, and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine code generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like. The hardware device described above may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

In addition, the above-mentioned line risk prediction method performed by the above-described line risk predicting apparatus may be implemented in the form of a computer program or an application executed by a computer stored in a recording medium.

The above description of the present application is intended for illustration, and it will be understood by those skilled in the art that the present invention may be easily modified in other specific forms without changing the technical spirit or essential features of the present application. Therefore, it should be understood that the embodiments described above are exemplary in all respects and not restrictive. For example, each component described as a single type may be implemented in a distributed manner, and similarly, components described as distributed may be implemented in a combined form.

The scope of the present application is indicated by the following claims rather than the above description, and it should be construed that all changes or modifications derived from the meaning and scope of the claims and their equivalents are included in the scope of the present application.

10: Track risk prediction device
11: data collector
12: data preprocessor
13: Binomial classification model
14: polynomial classification model
15: Risk Prediction

Claims (12)

In the track risk prediction method by the track risk prediction device,
Collecting new data subject to track risk prediction;
Performing preprocessing of the new data;
Based on the data set including at least one of accident failure data, line defect, maintenance data, weather data, and facility data, the binomial risk level value is set as a dependent variable, and the selected variable is set as an independent variable. Generating a binomial classification model for predicting track risk for the vehicle;
Generating a polynomial classification model for predicting a line risk for new data by setting a polynomial risk level value as a dependent variable and setting the selected variable as an independent variable based on the data set;
Learning the binomial classification model and the polynomial classification model by applying a plurality of machine learning algorithms to the generated binomial classification model and the polynomial classification model;
Selecting a binomial classification model and a polynomial classification model to which an optimal machine learning algorithm is applied in consideration of accuracy and penalty among the plurality of machine learning algorithms based on the learning result as a line risk prediction model;
Predicting a line risk by combining the selected binomial classification model and the polynomial classification model;
The dataset comprises a plurality of records,
The record is generated by connecting accident failure data, line defects, maintenance data, weather data, and facility data based on date, key, top and bottom codes,
Predicting the line risk,
And generating a complex model algorithm combining the selected binomial classification model and the polynomial classification model, and providing a result value derived from the algorithm of the complex model as a track risk prediction result. The method of claim 1,
And the new data is track-related data associated with at least one of accident failure data, track defects, maintenance data, weather data, and facility data. delete The method of claim 1,
Performing preprocessing of the new data,
Acquiring the data to be analyzed by searching the collected data in the step of collecting the new data,
Selecting an analysis variable to be analyzed from the acquired data and performing preprocessing of data corresponding to the selected variable. The method of claim 4, wherein
Performing preprocessing of the new data,
Robust method to predict the risk of the line using the median and interquartile range (IQR) to minimize the impact of outliers and to normalize the variable values in the data. delete The method of claim 1,
The plurality of machine learning algorithms include at least one of Random Forest, Logistic Regression, XGBoost Classifier, Balanced Bagging Classifier, Ensemble,
Track risk prediction method. The method of claim 7, wherein
Generating the polynomial classification model,
Generating a risk prediction model by setting a polynomial risk level value as a dependent variable and setting the selected variable as an independent variable;
Learning a generated risk prediction model by applying a plurality of machine learning algorithms;
Calculating a line risk prediction value using the plurality of machine learning algorithms;
Track risk prediction method comprising a. The method of claim 8,
The plurality of machine learning algorithms include at least one of K-Nearest Neighbor (knn), Support Vector Machine (svm), XGBoost Classifier, Balanced Bagging Classifier, Adaboost,
Track risk prediction method. delete A data collection unit collecting new data that is a target of track risk prediction;
A data preprocessor configured to preprocess the data collected by the data collector;
Based on the data set including at least one of accident failure data, line defect, maintenance data, weather data, and facility data, the binomial risk level value is set as the dependent variable, and the selected variable is set as the independent variable. A binomial classification model unit for generating a binomial classification model for predicting track risk;
A polynomial classification model unit for generating a polynomial classification model for predicting a line risk for new data by setting a polynomial risk level value as a dependent variable and setting the selected variable as an independent variable based on the data set;
Including a risk prediction unit for predicting the line risk by combining the binomial classification model and polynomial classification model,
The dataset comprises a plurality of records,
The record is generated by connecting accident failure data, line defects, maintenance data, weather data, and facility data based on date, key path, and up / down code.
The binomial classification model unit and the polynomial classification model unit,
A plurality of machine learning algorithms are applied to the generated binomial classification model and the polynomial classification model to learn the binomial classification model and the polynomial classification model, and based on the learning results, accuracy and penalty among the plurality of machine learning algorithms are calculated. Considering the binomial classification model and the polynomial classification model to which the optimal machine learning algorithm is applied as the line risk prediction model,
The risk prediction unit,
And generating a complex model algorithm combining the binomial classification model and the polynomial classification model, and providing a result value derived from the algorithm of the complex model as a track risk prediction result. A computer-readable recording medium having recorded thereon a program for executing the method of any one of claims 1, 2, 4, 5 and 7-9.

Images