Application of data mining in road safety | |
| Author | Hossain, Moinul |
| Call Number | AIT Thesis no.TE-05-05 |
| Subject(s) | Data mining Roads Safety measures |
| Note | A thesis submitted in pru.tial fulfillment of the requirements for the degree of Master of Engineering, School of Engineering and Technology |
| Publisher | Asian Institute of Technology |
| Series Statement | Thesis ; no. TE-05-05 |
| Abstract | Road accidents have been the major cause of injuries and fatalities in Thailand for the last few decades. In order to prevent injuries and fatalities, the Epidemiology Division of the Ministry of Public Health of Thailand has developed and maintained an Injury Surveillance (IS) database since 1995. This research study employs the latest know how; the Data Mining technologies on these IS data of 28 public hospitals consisting of 316,868 records for the period of 1999 to 2003 to obtain factors and multivariate relationships that can be used in accident analysis to improve road safety in Thailand. This study uses multivariate data mining tools (including logistic regression, classification and regression trees, SQL queries, etc.) to search the IS database for useful information that might not be previously found and discovers the determinants of injuries and fatalities in road accidents in Thailand. The assessment of severity of injury is one of the most significant factors to determine the severity of accident for road accident studies. The Probability of Survival (PS) model is an important variable to determine the severity of injury due to road accident. Most of the developing countries including Thailand use TRISS (Trauma and Injury Severity Score) model to determine the PS for the accident victims. However, the TRISS model was developed based on the hospital data of US where the emergency services and medical facilities are much better than Thailand. Moreover, the data used for developing the TRISS model was solely based on adult patients. On the contrary, in Thailand, substantial numbers of accident victims are young people. Therefore, the study also develops a Probability of Survival (PS) model using these IS data of 28 hospitals. Finally, the new PS model is validated using the IS data of a hospital (Udon-Thani Hospital) other than these 28 hospitals. The outcome of the data mining identifies that 76% of the accident victims experience injuries or fatalities due to motorcycle accidents. It is found that the young Thai motorcyclists have high tendency to drink and ride as around 54% of the drunk drivers are younger than 25 years. Almost 12% of the drivers belong to the age group of 15-17 years, which is a legal age group for the motorcycle drivers in Thailand and 6.2% of the drunk drivers also belong to this age group. Moreover, 91% of the motorcycle accident victims ignored the use of helmet. The study also conducts in depth analyses of different injury patterns and identifies the determinants of injuries and fatalities in motorcycle accidents. It concluded the head injuries to be the most prominent (39.18%) accident pattern. It is also obtained that use of helmet is being able to alleviate only 6.75% of the life threatening injuries. In the later part, the study also applies data mining to understand the situation of seatbelt use in Thailand and suggests that more than 90% of the accident victims ignored using seatbelt during accidents. Lastly, the study concentrates on pedestrian accidents and exhibits that the children under 10 years old comprise of the most vulnerable pedestrian group in Thai land. Regarding the development of PS model, it is found that the original TRISS model provides poor results in predicting the PS values for the non- survivors as only 32.3% are classified correctly. On the contrary, the new calibrated PS model provides better outcomes as 62.1% of the blunt non-survivors and 57.6% of the penetrating non-survivors are classified correctly. Furthermore, it is also found that the variable `age' requires different levels of classification to obtain more accurate results for Thai injury patients. Moreover, pulse rate is found to be an important variable for determining PS for the blunt injury patients |
| Year | 2006 |
| Corresponding Series Added Entry | Asian Institute of Technology. Thesis ; no. TE-05-05 |
| Type | Thesis |
| School | School of Engineering and Technology (SET) |
| Department | Department of Civil and Infrastucture Engineering (DCIE) |
| Academic Program/FoS | Transportation Engineering (TE) |
| Chairperson(s) | Yordphol Tanaboriboon; |
| Examination Committee(s) | Shinya Hanaoka;Kunnawee Kanitpong; |
| Scholarship Donor(s) | Asian Institute of Technology Fellowship; |
| Degree | Thesis (M.Eng.) - Asian Institute of Technology, 2005 |