ANALYSING ROAD TRAFFIC ACCIDENTS IN MALAWI: A DATA MINING APPROACH

Abstract

One of the significant problems the world faces today is the rate at which road traffic accidents and deaths on the roads are happening. The majority of these accidents occur in developing countries (Ihueze & Onwurah, 2018), and Malawi is no exception. However, to supplement the current safety measures, an analysis of road traffic accidents using data mining techniques was considered. Malawi being a low-income country, it is very crucial to have focus areas when dealing with traffic safety since there are limited budgetary resources. Therefore, this study aimed at digging for patterns in the traffic accident data and modeling the severity of road accidents in Malawi. Using python, three classification algorithms were employed to model the severity of an accident. The algorithms included Decision trees, Logistic regression and Support Vector Machines. These models were evaluated using accuracy, precision, recall, and F1-score. The logistic regression performed better than the other two and it was discovered that the top three attributes that contributed to fatal accidents were accidents involving a moving vehicle and a pedestrian, accidents that occurred at Dawn or Dust, and accidents involving a moving vehicle and a bicycle. Through association rule mining, a series of interesting rules were generated. Road Condition, Weather, posted speed limit and Surface type were the frequent item sets that appeared in all the rules generated.