KR20180086602A - Apparatus and method for estimating traffic jam area based on machine learning - Google Patents

Abstract

The present invention relates to an apparatus for estimating a traffic jam area based on machine learning and a method thereof. The apparatus includes a data collection part for collecting at least one of weather data, traffic data and road location data, a merged data generating part for merging the traffic data, the road location data, and the weather data to generate merged data, and a machine learning part which learns the merged data based on the machine learning and generates a decision tree model for predicting a traffic jam area for each weather.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus and method for estimating a traffic congestion section using machine learning,

The present invention relates to an apparatus and method for estimating a traffic congestion section using machine learning for estimating a traffic congestion section using machine learning.

Big Data, which emerged as a result of recent advances in information and communication technology, can be defined as data that is difficult to store, manage and analyze with existing software. Through the use of "Big Data" And provision of pre-emptive public services.

Big weather data of these "big data" is "big data" which includes various information related to the weather weather. Big weather data such as analysis of weather data, prediction of future weather information, energy tourism, weather insurance and other industries Can be big data about the weather that can create a new business through the fusion with.

Such weather data is not based on personal information, so it can be free from the problem of personal information protection. Combining such weather data with socio-economic data can bring synergy effects.

Traffic congestion is influenced by many factors such as accidents, construction, weather, etc. Furthermore, accidents are often caused by the carelessness of the driver and sometimes influenced by the weather, but conventionally the influence of weather on traffic There is a problem that can not be understood effectively.

Korean Registered Patent No. 10-0695918 (Mar.

SUMMARY OF THE INVENTION An object of the present invention is to solve the above problem by generating a decision tree model for predicting a traffic congestion period by weather by linking meteorological data and traffic data to machine learning to estimate a traffic congestion period.

The problems to be solved by the present invention are not limited to the above-mentioned problem (s), and another problem (s) not mentioned can be clearly understood by those skilled in the art from the following description.

In order to accomplish the above object, a traffic congestion section estimating apparatus using machine learning according to an embodiment of the present invention includes a data collecting unit for collecting at least one of weather data, traffic data, and road position data, traffic data, A merging data generating unit for merging the position data and the weather data to generate merged data, and a machine learning unit for learning the merged data based on the machine learning and generating a decision tree model for predicting the traffic congestion period by weather .

According to an embodiment of the present invention, since the weather data and the traffic data are used simultaneously for estimating the traffic congestion period, more accurate traffic congestion intervals for each weather can be estimated.

1 is a block diagram illustrating a traffic congestion section estimating apparatus according to an embodiment of the present invention.
2 is a view for explaining weather data utilized by the traffic congestion section estimating apparatus according to the embodiment of the present invention.
3 is a view for explaining traffic data utilized by the traffic congestion section estimating apparatus according to the embodiment of the present invention.
4 is a view for explaining road position data utilized by the traffic congestion section estimating apparatus according to the embodiment of the present invention.
5 is a view for explaining a merged data generating unit included in the traffic congestion interval estimating apparatus according to the embodiment of the present invention.
6 to 8 are views for explaining a machine learning unit included in the traffic congestion section estimating apparatus according to the embodiment of the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings in order to facilitate a person skilled in the art to easily carry out the technical idea of the present invention. . In the drawings, the same reference numerals are used to designate the same or similar components throughout the drawings. In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, an apparatus and method for estimating a traffic congestion section using machine learning according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

1, an apparatus 100 for estimating traffic congestion using machine learning according to an embodiment of the present invention includes a data collecting unit 110, a merge data generating unit 120, and a machine learning unit 130 .

The data collecting unit 110 collects at least one of weather data, traffic data, and road location data.

The merged data generation unit 120 merges the traffic data, the road position data, and the weather data to generate merged data.

The machine learning unit 130 learns the merged data based on the machine learning, and generates a decision tree model for predicting a traffic congestion period by weather.

For example, the apparatus 100 for estimating a traffic congestion zone using machine learning according to an embodiment of the present invention collects weather data and traffic data, and then calculates a traffic congestion interval using machine learning algorithms executed in a distributed system environment It may mean a device for estimating a traffic congestion period.

In other words, the traffic congestion section estimating apparatus 100 using machine learning according to the embodiment of the present invention analyzes a large amount of weather data and traffic data, and predicts how traffic flows on the main road depending on the weather.

The traffic congestion section estimating apparatus 100 using the machine learning according to the embodiment of the present invention can utilize a statistical analysis tool and a distributed processing system to analyze and predict a large amount of weather data and traffic data.

According to one embodiment, the traffic congestion section estimating apparatus 100 using the machine learning can use a representative solution of a processing system widely used for statistical software development and data analysis, such as Hadoop, MapReduce, Hadoop Distributed File System (HDFS), Hadoop (HBASE), Open Relational Distributed Database for Hadoop platform, Yet Another Resource Negotiator (YARN) To efficiently analyze a large amount of data.

For example, the traffic congestion section estimating apparatus 100 using the machine learning can use the Hadoop distributed file system to stably store a large amount of data on the distributed processing system, manage the distributed system state through the H base, You can use the deuce and yarn to process the distributed data and the spark machine learning to analyze the data.

For example, Apache Hadoop (High-Availability Distributed Object-Oriented Platform) is a freeware Java software framework that supports distributed applications running on large computer clusters capable of processing large amounts of data. Apache Hadoop is a subproject of Apache Lucene, which is designed to support Hadoop Distributed File System (HDFS) and MapReduce (Hadoop Distributed File System), which can replace the Google file system, Map Reduce.

The Hadoop distributed file system (HDFS) is a distributed, extensible file system written in the Java language for the Hadoop framework. The Hadoop distributed file system divides and stores large files on multiple machines, By storing, data stability is obtained.

The Hadoop distributed file system eliminates the need to use RAID (RAID) storage on the host, the Hadoop distributed file system has a master / slave structure, and the Hadoop distributed file system cluster has one name And a master server that manages the file system and controls access to the client. Each node in the cluster has one data node, which manages the storage added to the node each time it is executed.

Map Reduce is a software framework that was created in 2004 for the purpose of processing large amounts of data in Google distributed parallel computing. MapReduce is a software framework for large-scale petabytes of data, , And MapReduce is mainly composed of functions such as Map and Reduce which are commonly used in functional programming.

Apache Spark is a general purpose high performance distributed platform, an open source cluster computing framework developed by AMPLab at the University of California at Berkeley, later donated to the Apache Software Foundation and currently managed by the Apache Foundation. .

In contrast to Hadoop's slow implementation of MapReduce operations on a disk-based basis, Spark moves to memory-based and provides fast performance. MapReduce, Streaming data handling, SQL Based data queries, machine learning libraries, and graph data processing.

2 to 4, an embodiment of weather data, traffic data and road position data collected by the data collecting unit 110 will be described.

2 is a view for explaining weather data utilized by the traffic congestion section estimating apparatus according to the embodiment of the present invention. 3 is a view for explaining traffic data utilized by the traffic congestion section estimating apparatus according to the embodiment of the present invention. 4 is a view for explaining road position data utilized by the traffic congestion section estimating apparatus according to the embodiment of the present invention.

For example, the weather data collected by the data collection unit 110 may be as shown in FIG. 2, and weather data such as a point, a date and time, an average temperature, a minimum temperature, a minimum temperature, And the like.

For example, the traffic data collected by the data collecting unit 110 may be as shown in FIG. 3, and may mean data collected in the form of a CSV file.

For example, the road position data collected by the data collecting unit 110 may be as shown in FIG. 4, and may refer to data obtained by collecting road locations in a CSV file form on a standard node link.

In this case, although each data type collected by the data collection unit 110 is described as a CSV file, the present invention is not limited thereto.

Referring to FIG. 5, the merged data generation unit 120 will be described.

5 is a view for explaining a merged data generating unit included in the traffic congestion interval estimating apparatus according to the embodiment of the present invention.

The merge data generation unit 120 generates a merge data model 120 by using the machine learning unit 130 to generate various types of data for data construction and data Data and road position data are merged to generate merged data.

For example, the merge data generator 120 divides the weather data into a plurality of large cities (highways) to pre-process data for each region, and integrates the time of the congestion section and the time of the weather data into one day.

For example, since the weather data is formatted such as the amount of precipitation per day, the amount of sunshine, temperature, and the like, traffic data includes traffic information per 5 minutes, so that the merge data generator 120 can perform preprocessing to merge these data.

For example, the preprocessing method of the merged data generation unit 120 is as shown in FIG. 5, and the merged data generation unit 120 integrates and integrates the road location data including the traffic data and the road ID list of each region Data and weather data can be merged again by date to generate merged data.

The machine learning unit 130 will now be described with reference to FIGS.

6 to 8 are views for explaining a machine learning unit included in the traffic congestion section estimating apparatus according to an embodiment of the present invention.

The machine learning unit 130 can generate a model for predicting the traffic speed and traffic accident according to the weather through the decision tree model using the merged data including the weather data and the traffic data, In order to use this prediction technique, the unit 130 can use a Decision Tree, which is a function included in the C5.0 library provided by the program R. The C5.0 library can use the information acquisition amount in the tree generation algorithm, The present invention is not limited thereto.

Decision tree learning uses a decision tree as a predictive model that connects observation values and target values for an item. This decision tree can be used to show the decision process and the determined decision in a visual and explicit way. have.

For example, the decision tree learning method is a method used in data mining. It aims at creating a model for predicting the value of a target variable based on some input variables. In the decision tree learning method, In the input variable, the branches leading to the child nodes correspond to the possible values of the input variable, and the leaf nodes correspond to the target variable values when the respective input variables have values corresponding to the path from the root node to the leaf node.

When the congestion degree of a given data set is defined as entropy with respect to the amount of information obtained, the state with the highest congestion level can be represented as 1, and the state composed of only one class can be represented as 0, This function can be used to compute the information entropy difference that can occur when sampling an ideal distribution with another distribution approximating the distribution. The method of calculating the entropy is as follows: The method of calculating the information obtaining amount as shown in Equation (1) is expressed by Equation (2) below.

[Equation 1]

&Quot; (2) "

For example, in order to perform the machine learning, the machine learning unit 130 uses the data analyzed in the analysis system. In this case, the result of the training data (training data) and the training decision tree (Decision Tree) Testing Data, which is the data to check, is needed.

For example, the machine learning unit 130 may use a bootstrapping technique to reduce the instability of the existing decision tree model and increase the accuracy.

The bootstrapping technique can be a method of applying a random sampling before hypothesis verification or calculation of a metric, and it is required to allow redundancy.

When the bootstrapping technique is utilized, the machine learning unit 130 may use the data obtained by averaging the operation of extracting data through random sampling by a preset number of times, for example, 6, the gray line in FIG. 6 is randomly sampled data, and the red line may refer to bootstrapping data, but the present invention is not limited thereto .

For example, the machine learning unit 130 can predict a correlation between the temperature, snow, and rain of the weather data and the vehicle speed of the traffic data by a decision tree modeling, The speed (0 to 100 km / h) can be configured as Fast, Normal, or Slow.

For example, the machine learning unit 130 may use a result table in which the testing data is inserted by using the predict () function, and the accuracy and the table of the prediction model are as shown in FIG. 7, The model is as shown in FIG. 8, and the accuracy is 32.2%, which shows that the error rate is low.

That is, the traffic congestion section estimating apparatus 100 using the machine learning according to the embodiment of the present invention can estimate the correlation between the weather data and the traffic data collected through the decision tree modeling prediction result.

That is, the user can utilize the predictive traffic information according to the weather of the road through the traffic congestion section estimating apparatus 100 using the machine learning according to the embodiment of the present invention, and the user predicts and provides the road traffic information As part of the new method, users can guide the optimal route through expected road traffic information.

While the present invention has been described in connection with what is presently considered to be practical exemplary embodiments, it is to be understood that the invention is not limited to the disclosed embodiments, but many variations and modifications may be made without departing from the scope of the present invention. It will be understood that the invention may be practiced.

100: Traffic congestion section estimator using machine learning
110: Data collecting unit
120: Merge data generator
130: Machine learning department

Claims (1)

A data collection unit for collecting at least one of weather data, traffic data, and road location data;
A merged data generation unit for merging the traffic data, the road position data, and the weather data to generate merged data; And
And a machine learning unit that learns the merged data based on the machine learning and generates a decision tree model for predicting a traffic congestion period for each weather.

Images