A Tree Based Approach for Data Pre-processing and Pattern Matching for Accident Mapping on Road Networks

Abstract

Numbers of road accidents are increasing day by day and it is necessary to keep record of all these accidents, for the purpose of providing road safety. The records maintained by the police might have some inaccurate data. For this purpose, different authors have proposed different techniques, in which the accidents are mapped to the correct road segment. The existing accident mapping algorithms have many limitations such as (1) algorithms developed are for the specific datasets only, (2) improper mapping in case of complex road networks and (3) modular approach is used for different types of roads, instead of any general rule to be applied, (4) learning algorithm developed involve lot of cost and efforts, without guaranteeing the mapping of accident to the exact location. The challenge is to identify this inaccuracy in data and correct it with least amount of cost and effort involved. Significance of the attributes having inaccuracy plays an important role in mapping accidents to the correct location. In this paper, our main aim is to (1) map the accidents on the correct location, (2) finding missing values, (3) calculation of the erroneous values in police recorded data and (4) estimation of the attribute significance. The proposed algorithm can map the accidents to correct location instead of mapping an accident to a road segment, junction, and candidate link. The basic concern is that the algorithm being developed should be generalized and can be applied on any type of a road network. To achieve this aim, we have proposed the minimum bounded region based tree technique, which used for pattern matching methodology. Keywords Accident-mapping Pattern-matching Minimum bounded region (MBR) Missing values Attribute significance