Analyzing of traffic accident data play an important role in identifying the factors that affecting the repeated accidents and trying to reduce them. Accidents frequencies and their causes are different from one location to another and also differ from time to time in the same location. Data mining techniques such as clustering and classification are widely used in the analysis of road accident data. Therefore, this study proposes a framework to analyze times of accident frequencies for highway locations. The proposal framework consists of clustering technique and classification trees. The k-means algorithm is applied to a set of frequencies of highway locations accidents within 24 hours to find out when and where accidents occur frequently. These frequencies were extracted from 358,448 accident records in Britain between 2013 and 2015. As a result of clustering technique, four clusters were ranked in descending order according to the accidents rate for location within the cluster. After that, the decision tree (DT) algorithm is applied to the resulting clusters to extract the decision rules as the cluster name represents the class value for all tuples contained. However, extracting decision rules (DRs) from the DT is restricted by the DT's structure, which does not allow us to extract more knowledge from a specific dataset. To overcome this problem, in our study, we develop an ensemble method to generate several DTs in order to extract more valid rules. The DRs obtained were used for identifying the causes of road accidents within each cluster.
Title = "Analysis of Accident Times for Highway Locations Using K-Means Clustering and Decision Rules Extracted from Decision Trees.",
Journal ="International Journal of Computer Applications Technology and Research(IJCATR)",
Volume = "7",
Pages ="1 - 34",
Year = "2018",
Authors ="Ali Moslah Aljofey , Khalil Alwagih"}
The paper proposes a framework to analyze times of accident frequencies for highway locations
The k-means algorithm is applied to find out when and where accidents occur frequently
Many of decision rules were extracted from several decision trees applied to the resulting clusters
The extracted rules were used for identifying the causes of accidents within each cluster.