KR101982753B1 - Apparatus and method for predicting degree of risk of train accident - Google Patents

Abstract

The present invention relates to a method for predicting a train accident risk comprising the steps of: receiving new data to be a target of train accident risk prediction; selecting a risk prediction model for predicting a train accident risk for the new data based on a data set including train accident data and running performance data; and predicting a risk of a train accident on the new data based on the selected risk prediction model.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention [0001] The present invention relates to a railway accident risk prediction apparatus,

The present invention relates to an apparatus and method for predicting the risk of a train accident.

Train accidents (railway accidents) are one of the major accidents that cause great loss of lives and massive loss in a single accident. However, in the past, the development level of the risk analysis technology of train accidents has not been satisfactory.

For example, in the past, techniques for analyzing the risk of an accident based on statistical techniques have been known. However, there is a limit to precisely predict the risk of train accidents through statistical technique based accident risk analysis technology.

The background technology of the present application is disclosed in Korean Patent Registration No. 10-1020191.

It is an object of the present invention to provide an apparatus and method for predicting the risk of a train accident which can effectively prevent a risk of a train accident in advance by more accurately predicting the risk of a train accident.

It is to be understood, however, that the technical scope of the embodiments of the present invention is not limited to the above-described technical problems, and other technical problems may exist.

According to a first aspect of the present invention, there is provided a method for predicting a train accident risk, the method comprising: receiving new data to be predicted for a train accident risk; Selecting a risk prediction model for predicting a train accident risk for the new data based on a data set including train accident data and driving performance data; And predicting a risk of a train accident for the new data based on the selected risk prediction model.

According to a first aspect of the present invention, there is provided a method for estimating a risk of a train accident, comprising the steps of: performing pre-processing for normalizing data stored in the data set; And generating a risk prediction model based on the normalized data.

The pre-processing may include: calculating a severity of each record in the data set based on the normalized data; And generating at least one parameter for generating a risk prediction model based on the normalized data, wherein the step of generating the risk prediction model comprises: calculating a risk level value by the calculation of the severity, The risk prediction model can be generated using the parameters.

Also, the step of calculating the severity may calculate the severity in consideration of the scale of the accident damage and the frequency of the accident.

Also, the step of calculating the severity may calculate the severity by considering the delay time.

Further, the train accident risk prediction method according to the first aspect of the present invention further includes performing learning on the risk prediction model by applying a plurality of machine learning algorithms to the generated risk prediction model, Selecting a model may include selecting a risk prediction model to which a machine learning algorithm representing the highest accuracy among the plurality of machine learning algorithms is applied as a risk prediction model for predicting a train accident risk for the new data based on the learning results .

The plurality of machine learning algorithms may also include at least one of a random forest algorithm, a support vector machine (SVM) algorithm, and a k-nearest neighbors (KNN) algorithm. have.

In addition, the predicting step may predict the risk of a train accident with respect to the new data as one of normal, caution, warning, and danger.

Also, the data set includes a plurality of records, and the record may be generated by connecting the train accident data and the driving performance data on the basis of a date of operation, a combination number, and a train number.

According to a second aspect of the present invention, there is provided an apparatus for predicting a train accident risk, comprising: a receiver for receiving new data to be predicted for a train accident risk; A risk prediction model for selecting a risk prediction model for predicting a train accident risk for the new data based on a data set including train accident data and running performance data; And a predictor for predicting a risk of a train accident with respect to the new data based on the selected risk prediction model.

The above-described task solution is merely exemplary and should not be construed as limiting the present disclosure. In addition to the exemplary embodiments described above, there may be additional embodiments in the drawings and the detailed description of the invention.

According to the above-mentioned problem solving means, the risk prediction model generated by using the train accident data and the operation performance can predict the risk of the train accident, and thereby it is possible to prevent the risk of the train accident effectively in advance.

However, the effects obtainable here are not limited to the effects as described above, and other effects may exist.

1 is a diagram showing a schematic configuration of a train accident risk prediction apparatus according to a first aspect of the present invention.
2 is a diagram showing an example of selection of a variable in the train accident risk prediction apparatus according to the first aspect of the present invention.
FIG. 3 is a diagram illustrating a result of predicting a train accident risk for a risk prediction model to which a plurality of machine learning algorithms are applied in the train accident risk prediction apparatus according to the first aspect of the present invention. FIG.
FIG. 4 is a diagram illustrating a plurality of predictive accuracy predictions for each of a plurality of machine learning algorithms in a train accident risk prediction apparatus according to the first aspect of the present invention.
FIG. 5 is a diagram showing an example of the result of predicting a train accident risk according to the application of a risk prediction model to actual new data in the train accident risk prediction apparatus according to the first aspect of the present invention.
FIG. 6 is a diagram schematically showing the configuration of a device for analyzing the relationship between types of railway accident data according to a second aspect of the present invention.
FIG. 7 is a view showing an example of railway accident data obtained by the apparatus for analyzing the correlation between types of railway accident data according to the second aspect of the present invention; FIG.
FIG. 8 is a diagram showing a railway accident data in a two-dimensional matrix in order to apply the SVD technique in the apparatus for analyzing associations between types of railway accident data according to a second aspect of the present invention.
FIG. 9 is a diagram showing an example in which the association analysis result between types of railway accident data is displayed in an easy-to-understand manner by a user in an apparatus for analyzing associations between railway accident data types according to a second aspect of the present invention.
10 is a flowchart of a method for predicting the risk of a train accident according to the third aspect of the present application.
11 is a flowchart illustrating a method of analyzing a correlation between types of railway accident data according to a fourth aspect of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily carry out the present invention. It should be understood, however, that the present invention may be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. In the drawings, the same reference numbers are used throughout the specification to refer to the same or like parts.

Throughout this specification, when an element is referred to as being "connected" to another element, it is intended to be understood that it is not only "directly connected" but also "electrically connected" or "indirectly connected" "Is included.

It will be appreciated that throughout the specification it will be understood that when a member is located on another member "top", "top", "under", "bottom" But also the case where there is another member between the two members as well as the case where they are in contact with each other.

Throughout this specification, when an element is referred to as " including " an element, it is understood that the element may include other elements as well, without departing from the other elements unless specifically stated otherwise.

1 is a diagram showing a schematic configuration of a train accident risk prediction apparatus according to a first aspect of the present invention.

Referring to FIG. 1, the train accident risk prediction apparatus 10 according to the first aspect of the present invention may include a receiving unit 11, a risk prediction model selecting unit 12, and a predicting unit 13.

The receiving unit 11 can receive new data to be subjected to the train accident risk prediction. Here, the new data may mean train-related data related to at least one of train accident data and driving performance data. Also, the term " train "

The risk prediction model selection unit 12 may select a risk prediction model for predicting the train accident risk for new data based on the data set including the train accident data and the running performance data. Models may be referred to herein as models.

Here, the train accident risk prediction apparatus 10 according to the first aspect of the present invention may include a data set generation unit (not shown) for generating a data set. The data set can contain integrated train accident data and operational performance data. According to this, the data stored in the data set can store the train accident data according to the running performance. Train accident data are data related to train accidents and train disturbances in the past, and may be referred to as accident trouble data, rail accident data, and the like. The operation performance data may include information on the number of train operations and the distance of train operation (for example, key number) as the performance related to the operation of the train, but the present invention is not limited thereto.

The data set may also include a plurality of records. Here, the record can be generated by connecting the train accident data and the operation performance data on the basis of the operation date, the combination number, and the train number. In other words, in order to predict the risk of train accidents (risk level), train accident data and train performance data can be stored in the dataset based on the date of operation, the train number, and the train number. At this time, one data stored in the data set connected with the travel date, train number and train number can be regarded as one record.

A detailed explanation for the selection of the risk prediction model (model) is as follows.

The train accident risk prediction apparatus 10 according to the first aspect of the present invention may include a preprocessing unit 14, a risk prediction model generation unit 15, and a learning unit 16.

The preprocessing unit 14 may perform preprocessing for normalizing the data stored in the data set. That is, the preprocessing unit 14 may perform data normalization to adjust the variable values of the data stored in the data set to a predetermined standard. In particular, the preprocessor 14 may be configured to have a non-uniform portion of the data in the data set, i.e., a non-constant 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 The data can be normalized by the MIN / MAX method. The preprocessing unit 14 may cause the normalization through the MIN / MAX method to convert the variable value of the data stored in the dataset to have a value within a range between the minimum value 0 and the maximum value 1.

Also, the preprocessing unit 14 may calculate the severity of each record in the data set based on the normalized data (i.e., the normalized data stored in the data set). At this time, it is possible to derive (calculate) the value of train accident risk level per record by calculating the severity by record.

The preprocessing unit 14 can calculate the severity considering the scale of the accident damage and the frequency of the accident. Particularly, the preprocessing unit 14 can calculate the severity through the multiplication of the accident damage scale and the accident frequency.

In other words, the severity can be estimated by taking into account the frequency of accidents and the scale of accident damage. At this time, the preprocessing unit 14 can calculate the severity of each record in the data set by multiplying the frequency of the accident and the damage scale of each accident (accident damage scale). Here, the magnitude of the accident damage can be calculated by the amount of damage caused in case of a train accident. In addition, the scale of accident damage can be calculated by sum- ming the amount of damage caused by trains in case of a train accident, In other words, the scale of accident damage can be calculated by taking into account the amount of damage caused by trains in addition to the amount of damage caused by trains.

In addition, the preprocessing unit 14 can calculate the severity by taking into account the magnitude of the accident, the frequency of the accident, and the delay time. Here, the delay time may mean a train delay time due to a train accident.

Also, the preprocessing unit 14 may select at least one variable for generation of the risk prediction model based on the normalized data (i.e., the normalized data stored in the data set). In other words, the preprocessing unit 14 can select the parameters (important variables) necessary for predicting the railway accident risk based on the normalized data. The selection of these variables can be made to use only the variables that influence the risk prediction model in generating the risk prediction model (model).

2 is a diagram showing an example of selection of a variable in the train accident risk prediction apparatus according to the first aspect of the present invention.

Referring to FIG. 2, a plurality of variables such as D232, D431, D121, D145, D117, and D112 may be included in the data set as normalized data, for example, with respect to the variable rf_1C2015. For example, as shown in FIG. In this case, the significance of the variable may mean that the larger the value of MeanDecreaseGini (uncertainty), the horizontal axis value, the higher the importance.

The preprocessing unit 14 can select D232 as an example (important) for predicting the railway accident risk in relation to the rf_1C2015 variable. However, the present invention is not limited thereto, and the setting of the number of variables to be selected can be variously set based on user input.

The risk prediction model generation unit 15 can generate a risk prediction model (model) based on the normalized data. Specifically, the risk prediction model generation unit 15 can generate a risk prediction model using the risk level value and the selected variable based on the calculation of the severity based on the normalized data. That is, the risk prediction model generation unit 15 can generate a risk prediction model using the values of the train accident risk level calculated by the severity calculation and the parameters selected for predicting the train accident risk. At this time, the risk prediction model generation unit 15 can generate a risk prediction model by setting the train accident risk level value as a dependent variable and the selected variable as an independent variable.

The learning unit 16 can perform learning on the generated risk prediction model by applying a plurality of machine learning algorithms to the risk prediction model generated by the risk prediction model generation unit 15. [ In other words, the risk prediction model generated by the risk prediction model generation unit 15 can be learned by applying a plurality of machine learning algorithms.

At this time, the risk prediction model generated by the risk prediction model generation unit 15 is applied to each of the plurality of machine learning algorithms, so that a train accident risk (risk level) can be calculated corresponding to each of the plurality of machine learning algorithms. In other words, the generated risk prediction model is applied to each of the plurality of machine learning algorithms, so that the train accident risk level can be determined as a result of applying the machine learning algorithm corresponding to each of the plurality of machine learning algorithms.

Here, the plurality of machine learning algorithms may include at least one of a random forest algorithm, a support vector machine (SVM) algorithm, and a k-nearest neighbors (KNN) algorithm . However, the present invention is not limited thereto, and various machine learning algorithms can be applied.

The risk prediction model selection section 12 selects a risk prediction model (model) to which a machine learning algorithm representing the highest accuracy among a plurality of machine learning algorithms is applied based on the learning result through the learning section 16, It can be selected as a risk prediction model (model) for predicting accident risk. In other words, the risk prediction model selection unit 12 selects, based on the learning results obtained by applying each of the plurality of machine learning algorithms to the generated risk prediction model, the risk prediction model using the machine learning algorithm with the highest accuracy among the plurality of machine learning algorithms The model can be selected for predicting train accident risk for new data.

For example, the risk prediction model generation unit 15 applies the risk prediction model generated by the risk prediction model generation unit 15 to each of the plurality of machine learning algorithms, so that the plurality of machine learning algorithms A risk prediction model can be generated. Thereafter, the learning unit 16 can perform learning for each of the plurality of risk prediction models. Thereafter, the risk prediction model selection unit 12 can select a risk prediction model having the highest accuracy among a plurality of risk prediction models as a risk prediction model for predicting a train accident risk based on learning results.

FIG. 3 is a diagram illustrating a result of predicting a train accident risk for a risk prediction model to which a plurality of machine learning algorithms are applied in the train accident risk prediction apparatus according to the first aspect of the present invention. FIG.

Referring to FIG. 3, (a) to (c) show performance evaluation results of a risk prediction model to which a plurality of machine learning algorithms, random forest, SVM, and KNN, respectively, are applied through a confusion matrix. Here, in each of (a) to (c), the vertical axis represents the value of the true label and represents the value of the predicted label. According to this, the risk prediction model selection unit 12 predicts a risk prediction model for a new data by applying a machine learning algorithm that represents the highest accuracy among a plurality of machine learning algorithms through a confusion matrix As a risk prediction model.

FIG. 4 is a diagram illustrating prediction accuracy of a plurality of risk prediction models (models) corresponding to each of a plurality of machine learning algorithms in the train accident risk prediction apparatus according to the first aspect of the present invention.

 Referring to FIG. 4, for example, the accuracy of the risk prediction model (model) using the random forest algorithm is 60%, the accuracy of the risk prediction model using the SVM algorithm is 53%, the accuracy of the risk prediction model using the KNN algorithm is 52 %. ≪ / RTI > In this case, the risk prediction model selection unit 12 can select a risk prediction model to which a random forest algorithm is applied among a plurality of risk prediction models as a risk prediction model for predicting a train accident risk for new data.

The predicting unit 13 can predict the risk of train accident (risk level) for new data based on the risk prediction model selected by the risk prediction model selecting unit 12. [ That is, the prediction of the train accident risk for new data can be performed based on the risk prediction model having the highest accuracy among the plurality of risk prediction models corresponding to the plurality of machine learning algorithms.

Further, the predicting unit 13 can predict the risk (risk level) of the train accident for the new data to one of normal, caution, warning and danger. Specifically, the predicting unit 13 can predict that the train accident risk (risk level) is 'normal' when the risk value of the train accident risk of the new data based on the risk prediction model is calculated as 1. [ In addition, the predicting unit 13 can predict that the risk of train accident (risk level) is 'around' when the risk value of train accident risk of new data based on the risk prediction model is calculated as 2. [ In addition, the predicting unit 13 can predict that the train accident risk (risk level) is 'warning' when the risk value of the train accident risk of the new data based on the risk prediction model is calculated as 3. [ In addition, the predicting unit 13 can predict that the risk of train accident (risk level) is 'dangerous' when the risk value of train accident risk value of new data based on the risk prediction model is calculated as 4. [

FIG. 5 is a diagram showing an example of the result of predicting a train accident risk according to the application of a risk prediction model to actual new data in the train accident risk prediction apparatus according to the first aspect of the present invention.

Referring to FIG. 5, for example, a date and a combination number associated with new data may be associated with each other in a data set and stored as one piece of data. For example, '20160828' is stored in the first record among the plurality of records stored in the data set in association with the date ID, and '1C2004' may be stored in association with the combination number. 5, the first record is applied to the risk prediction model selected by the risk prediction model selecting unit 12, and as a result, the train accident risk value (predicitions_rf_1) for the first record is calculated as 1 ), The predicting unit 13 can predict that the risk of a train accident (risk level) for new data corresponding to the first record is 'normal'.

The train accident risk prediction apparatus 10 according to the first aspect of the present invention generates a data set including train accident data according to the running performance (that is, including the train accident data and the running performance data) In order to reduce the error of the risk (risk level) prediction, the data set is normalized, the severity of the risk (risk level) of the train accident is determined based on the data set, (Risk level) for new data can be predicted based on the generated risk prediction model by generating a risk prediction model by taking into consideration the selected important variables and the calculated severity. In other words, we can predict the risk of train accidents by using the risk prediction model created by using the train accident data and operation results, and thereby prevent the risk of train accidents effectively in advance.

In addition, the present invention can construct and store big data in the data set by linking the train accident data and the running performance data. In addition, we can predict the risk of train accidents (that is, the risk of train accidents) by using the machine learning algorithm based on the combination number and the train number. In other words, we can achieve a more accurate and improved prediction rate by predicting the risk of train accidents using Big Data based machine learning algorithms.

Hereinafter, an apparatus for analyzing a correlation between train accident data types according to a second aspect of the present invention will be described. Here, the train accident data refers to train accident data included in the data set described above, and may be hereinafter referred to as railway accident data.

In other words, the association analysis technique between the types of railway accident data according to the second aspect of the present invention is a method of constructing a two-dimensional matrix of railway accident data and analyzing the matrix of the singular value decomposition method The present invention relates to a technique for displaying an association value calculated by applying a single value decomposition (SVD).

FIG. 6 is a diagram schematically showing the configuration of a correlation analysis apparatus 100 between types of railway accident data (train accident data) according to a second aspect of the present invention.

Referring to FIG. 6, the correlation analysis system 100 of a railway accident data type according to the second aspect of the present invention may include a processor 110, and the processor 110 may include a redefinition unit 120, 130, an analysis unit 140, and a control unit 150.

The processor 110 may control operations of the redefinition unit 120, the matrix generation unit 130, the analysis unit 140, and the control unit 150.

The redefinition unit 120 can redefine the acquired railway accident data. Before describing this, an example of the railway accident data obtained is as follows.

FIG. 7 is a view showing an example of railway accident data obtained by the apparatus for analyzing the correlation between types of railway accident data according to the second aspect of the present invention; FIG.

Referring to FIG. 7, the railway accident data can be classified into an accident / disability category, a date and time of occurrence, a weather, a relationship affiliation, a railroad category, and a place of occurrence.

More specifically, the information on the type of accident / disability may include a vehicle malfunction, and the information on the date and time may include information such as year, month, day, time, day of the week, And weather and temperature information such as snow, clouds, sunshine, and the like. In addition, the related information of the affiliated affiliation can include information belonging to Seoul, Jeonbuk, etc., and the information related to the railroad type can include information such as a high speed railroad, a general railroad, an urban railroad, The information may include information such as route, reverse section, road / metropolitan area, city / county / district, and the like.

The redefining unit 120 may redefine the railway accident data obtained as shown in FIG. 7, for example, to analyze the correlation between types of railway accident data.

In more detail, the redefinition unit 120 can redefine the number of rows for the railway accident data as the number of accidents. In addition, the redefinition unit 120 may determine whether or not an accident / disability type 1, an accident / disability type 2, year, month, day, time, day, weather, temperature, (Ie, the starting point for thinking that defines viewpoints on things and phenomena), and that the contents belonging to each item should be regarded as members belonging to the point of view You can redefine it. For example, the route may be a viewpoint, and the Gyeongbu line, Honam line, and the like may be members.

The matrix generation unit 130 may generate a two-dimensional matrix using some of the data configuration items in the obtained railway accident data. Alternatively, the matrix generation unit 130 may generate a two-dimensional matrix based on the railway accident data redefined through the redefinition unit 120.

The matrix generation unit 130 may generate a two-dimensional matrix through a pivot function. The pivot function may be a pivot table provided by an Excel (MS Excel) or an on-line analytical processing tool Or the like. The pivot function is one of the data analysis techniques. The example described above is only one example, but the present invention is not limited thereto, and other kinds of analysis techniques may be applied.

An example of the two-dimensional matrix generated through the matrix generation unit 130 may be as shown in FIG.

FIG. 8 is a diagram showing a railway accident data in a two-dimensional matrix in order to apply the SVD technique in the apparatus for analyzing associations between types of railway accident data according to a second aspect of the present invention.

Referring to FIG. 8, for example, the matrix generation unit 130 generates a matrix of data items (high-speed rail, general rail, and urban rail) belonging to the railroad type item and data members (Gyeongbu line, Gyeongwon line, Kyungpun line, A two-dimensional matrix can be generated. For example, based on the two-dimensional matrix of FIG. 3, the number of accidents on the high-speed railway on the Gyeongbu line is 5, the number of accidents on the urban railway on the Gyeongbu line is 11, the number of accidents on the general railway on the Gyeongbu line is 9, The number of accidents can be 10, etc.

The analysis unit 140 can analyze the association between data types corresponding to the two-dimensional matrix by applying a single value decomposition (SVD) to each item of the two-dimensional matrix generated by the matrix generation unit 130 have.

Single Value Decomposition (SVD) is a mathematical method for decomposing a matrix into a plurality of small matrices, and is well known in the related art, and therefore will not be described below.

For example, when the SVD is applied to each item of the two-dimensional matrix as shown in FIG. 8, the analysis unit 140 can analyze the association between the vehicle type (railroad type) and the route as the association between types of railway accident data. In addition, the analysis unit 140 can analyze the association between the vehicle type and the accident / disorder type through the two-dimensional matrix based on the railroad classification item and the accident / disorder classification item. As described above, the analysis unit 140 can analyze the association between two different types, rather than the association between individual items of the accident data.

In general, when SVD is applied to a two-dimensional matrix, similarity in terms of rows or similarity in terms of columns can be calculated. However, since the present invention analyzes an association between two types of items rather than an association between individual items, the analysis unit 140 may calculate the similarity value between the two types, or the analysis unit 140 may calculate the similarity level between the two types It can be judged whether or not it is above a certain level (for example, 70%).

The control unit 150 may control the analysis unit 140 to display the analyzed result on the display screen.

The control unit 150 may cause the similarity value between the two types calculated through the analysis unit 140 to be displayed on the display screen. When the analysis unit 140 determines whether the degree of similarity between two types is greater than or equal to a certain level, the control unit 150 determines whether or not there is a correlation between the types of relevancy determination (Yes / No) based on the result It can be displayed on the display screen.

An example of screen display of the results of the association analysis can be more easily understood with reference to Fig.

FIG. 9 is a diagram showing an example in which the association analysis result between types of railway accident data is displayed in an easy-to-understand manner by a user in an apparatus for analyzing associations between railway accident data types according to a second aspect of the present invention.

Referring to FIG. 9, on the display screen, the degree of association and the degree of association between the object type information and the opposite type can be displayed as a result of the association analysis. More specifically, for example, 'vehicle type-route' information may be displayed on the display screen as object type information, and the correlation between the two types of vehicle type and route in the railway accident data is 'yes' , And the degree of association may be expressed as 85%.

In the embodiment according to the second aspect of the present invention, the association analysis between the two types is performed based only on the railway accident data. However, the present invention is not limited thereto and can be applied to various fields of data.

The association analyzing apparatus 100 according to the second aspect of the present invention can easily calculate the association between types of railway accident data through SVD.

In addition, the association analyzing apparatus 100 according to the second aspect of the present invention redefines the railway accident data, generates a two-dimensional matrix using some items of the data configuration items in the railway accident data, The SVD is applied to each item to analyze the association between the data types corresponding to the two-dimensional matrix and the analysis result is displayed on the display screen. Can be analyzed.

In addition, the association analyzing apparatus 100 according to the second aspect of the present invention displays a method of constructing a two-dimensional matrix of railway accident data and a result of applying the SVD solution to the matrix, thereby allowing the user to intuitively Can be provided in an easy-to-understand form.

10 is a flowchart of a method for predicting the risk of a train accident according to the third aspect of the present application.

The train accident risk prediction method shown in FIG. 10 can be performed by the train accident risk prediction apparatus 10 described above. Therefore, even if omitted from the following description, the description of the train accident risk prediction device 10 can be equally applied to the description of the train accident risk prediction method.

Referring to FIG. 10, the method for predicting a train accident risk according to the third aspect of the present invention can receive new data to be subjected to a train accident risk prediction (S11).

Next, in step S12, a risk prediction model for predicting a train accident risk for new data can be selected based on a data set including train accident data and driving performance data.

Here, the data set includes a plurality of records, and the records can be generated by connecting the train accident data and the running performance data on the basis of the operation date, the combination number, and the train number.

Next, in step S13, the risk of train accident (risk level) for the new data can be predicted based on the risk prediction model selected in step S12.

At this time, in step S13, the risk of a train accident with respect to new data can be predicted as one of normal, caution, warning, and danger.

Also, although not shown in the drawings, the method for predicting the risk of a railway accident according to the third aspect of the present invention may include a step of performing a pre-processing for normalizing data stored in a data set.

At this time, the pre-processing step may include calculating the severity of each record in the data set based on the normalized data, and selecting at least one parameter for generating a risk prediction model based on the normalized data . ≪ / RTI >

Here, at the stage of calculating the severity, the severity can be calculated considering the scale of the accident and the frequency of the accident. In the step of calculating the severity, the severity can be calculated considering the delay time.

In addition, the train accident risk prediction method according to the third aspect of the present invention may include generating a risk prediction model based on the normalized data. At this time, in the step of generating the risk prediction model, the risk prediction model can be generated by using the risk level value and the selected variable by the calculation of the severity.

Also, the train accident risk prediction method according to the third aspect of the present invention may include performing learning on the risk prediction model by applying a plurality of machine learning algorithms to the generated risk prediction model.

At this time, in the step of selecting a risk prediction model, a risk prediction model to which a machine learning algorithm showing the highest accuracy among a plurality of machine learning algorithms is applied is used as a risk prediction model for predicting a train accident risk Can be selected.

Here, the plurality of machine learning algorithms may include at least one of a random forest algorithm, a support vector machine (SVM) algorithm, and a k-nearest neighbors (KNN) algorithm .

In the above description, steps S11 to S13 may be further divided into further steps or combined into fewer steps, according to embodiments of the present application. Also, some of the steps may be omitted as necessary, and the order between the steps may be changed.

11 is a flowchart illustrating a method of analyzing a correlation between types of railway accident data according to a fourth aspect of the present invention.

The correlation analysis method between railway accident data types shown in FIG. 11 can be performed by the correlation analysis apparatus 100 between railway accident data types described above. Therefore, even if omitted below, the contents described for the correlation analyzing apparatus 100 between the railway accident data types can be similarly applied to the explanation of the correlation analysis method between railway accident data types.

Referring to FIG. 11, the acquired railway accident data can be redefined in step S21.

Next, in step S22, a two-dimensional matrix can be generated using some items of the data configuration items in the railway accident data.

At this time, in step S22, a two-dimensional matrix may be generated using a pivot table of Excel or a pivot grid of multi-dimensional analysis.

Next, in step S23, the association between the data types corresponding to the two-dimensional matrix can be analyzed by applying the single value decomposition method (SVD) to each item of the two-dimensional matrix generated in step S22.

At this time, in step S23, it can be determined whether or not the association between data types is equal to or greater than a predetermined similarity value.

Next, in step S24, the analysis result in step S23 can be displayed on the display screen.

In the above description, steps S21 to S24 may be further divided into additional steps, or combined in fewer steps, according to embodiments of the present disclosure. Also, some of the steps may be omitted as necessary, and the order between the steps may be changed.

The method for predicting a train accident risk according to an embodiment of the present invention and the method for analyzing a correlation between railway accident data types can be implemented in a form of a program command that can be executed through various computer means and recorded in a computer readable medium. The computer-readable medium may include program instructions, data files, data structures, and the like, alone or in combination. The program instructions recorded on the medium may be those specially designed and constructed for the present invention or may be available to those skilled in the art of computer software. Examples of computer-readable media include magnetic media such as hard disks, floppy disks and magnetic tape; optical media such as CD-ROMs and DVDs; magnetic media 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 machine language code such as those produced by a compiler, as well as high-level language code that can be executed by a computer using an interpreter or the like. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

It will be understood by those of ordinary skill in the art that the foregoing description of the embodiments is for illustrative purposes and that those skilled in the art can easily modify the invention without departing from the spirit or essential characteristics thereof. It is therefore to be understood that the above-described embodiments are illustrative in all aspects and not restrictive. For example, each component described as a single entity may be distributed and implemented, and components described as being distributed may also be implemented in a combined form.

The scope of the present invention is defined by the appended claims rather than the detailed description, and all changes or modifications derived from the meaning and scope of the claims and their equivalents should be construed as being included within the scope of the present invention.

10: Train accident risk prediction device
11: Receiver
12: Risk prediction model selection
13:

Claims (11)

A method for predicting a train accident risk by a train accident risk prediction device,
Receiving, in the train accident risk prediction device, new data to be a target of train accident risk prediction;
Selecting a risk prediction model for predicting a train accident risk for the new data based on a data set including train accident data and driving performance data in the train accident risk prediction device;
Wherein the train accident risk prediction device includes a plurality of machine learning algorithms for applying a plurality of machine learning algorithms to a risk prediction model generated based on data stored in a normalized data set for selecting the risk prediction model, Generating a plurality of risk prediction models and performing learning for each of the plurality of generated risk prediction models; And
And predicting the risk of a train accident for the new data to one of normal, caution, warning, and risk based on the selected risk prediction model in the train accident risk prediction device,
The data set is generated to include a plurality of records, and the record includes train accident data, information related to past train accidents and train disturbances, and train performance data, which is information related to the number of train runs and train travel distance, Number, and train number and is generated as one record,
Wherein the step of selecting the risk prediction model comprises:
Calculating, based on the learning results, a risk prediction model to which a machine learning algorithm indicating the highest accuracy among the plurality of risk prediction models is applied, in consideration of a plurality of prediction accuracy of each risk prediction model corresponding to each of the plurality of machine learning algorithms, A method for predicting the risk of a train accident, which is selected as a risk prediction model for predicting the risk of train accidents to data. The method according to claim 1,
Performing a pre-processing for normalizing data stored in the data set in the train accident risk prediction device,
The method comprising the steps of: 3. The method of claim 2,
The step of performing the pre-
Estimating a severity of each record in the data set based on the normalized data; And
Selecting at least one variable for generation of a risk prediction model based on the normalized data,
Wherein the risk prediction model generated based on the data stored in the normalized data set is generated using the risk level value by the calculation of the severity and the selected variable. The method of claim 3,
Wherein the step of estimating the severity is to calculate the severity in consideration of the magnitude of the accident damage and the frequency of accidents. 5. The method of claim 4,
Wherein calculating the severity further comprises calculating the severity by further considering the delay time. delete The method according to claim 1,
The plurality of machine learning algorithms
The method comprising at least one of a random forest algorithm, a support vector machine (SVM) algorithm, and a k-nearest neighbors (KNN) algorithm. delete delete A train accident risk prediction apparatus comprising:
A receiver for receiving new data to be a target of a train accident risk prediction;
A risk prediction model for selecting a risk prediction model for predicting a train accident risk for the new data based on a data set including train accident data and running performance data;
Generating a plurality of risk prediction models corresponding to each of the plurality of machine learning algorithms by applying a plurality of machine learning algorithms to the risk prediction model generated based on the data stored in the normalized data set in order to select the risk prediction model A learning unit for performing learning for each of the generated plurality of risk prediction models; And
And a predictor for predicting the risk of a train accident for the new data to one of normal, caution, warning, and risk based on the selected risk prediction model,
The data set is generated to include a plurality of records, and the record includes train accident data, information related to past train accidents and train disturbances, and train performance data, which is information related to the number of train runs and train travel distance, Number, and train number and is generated as one record,
Wherein the risk prediction model selection unit comprises:
Calculating, based on the learning results, a risk prediction model to which a machine learning algorithm indicating the highest accuracy among the plurality of risk prediction models is applied, in consideration of a plurality of prediction accuracy of each risk prediction model corresponding to each of the plurality of machine learning algorithms, A system for predicting a train accident risk, the system being selected as a risk prediction model for predicting a train accident risk for data. A computer-readable recording medium storing a program for executing the method of any one of claims 1 to 5 and a computer.

Images