Improving Road Safety by Profiling Different Accident Type

Project Details

Term: 

Fall 2015

Students: 

Angela Hung, Aylada Khunvaranont, Celia Chen, Dobby Yang, Mahsa Ashouri

University: 

NTHU

Presentation: 

Report: 

Business Problem
We all are road users and everyday we see the news about car accidents, which leads us to the concern of what can Transportation Department of Taipei City Government do to decrease the number of the accidents. Therefore, our project focuses on how to identify
what factors or conditions have influence on the certain accident types. With our findings, the government can pay more attentions to those area where accident happened. Moreover, the government can manage their budgets wisely and effectively at where they will fix what; for instance, fixing the road conditions, install more equipment, install more traffic lights and etc.

Data
The raw data we get is from Data Taipei (http://data.taipei/) which has data that is collected by government sectors and is a reliable source. We have combined three years of data set into the data that we will use to analyze.

Analytics Solution
There are several combinations of different software, algorithm and data preparation methods taken to reach the best solution of this business problem. After we tried, decision tree grown by RapidMiner is chosen as the final solution. Although it is not the best performance,
it has the most powerful explanatory ability, which is the most important element in explaining data field.

Recommendations
We found seven exact conditions of BackHit , SideHit , Ped_crossing and S cratch by profiling the data we had. According to the result, we are able to provide some advice for the Transportation
Department of Taipei City Government to improve the road safety efficiently.

Application Area: