Developing a Knowledge-driven Geospatial Information System for Regional Transportation Corridors Accidents Prediction

With increasing traffic density on Iran’s regional corridors, the probability of accident occurrence has increased accordingly. The main objective of this paper is to develop an ontology-driven geospatial information system to induct major crash rules for vehicles accident severity prediction which has the full domain knowledge and logical reasoning ability based on ontology. In the proposed approach, Geographic Information System (GIS) provides a platform both for spatial data analysis and for visualizing relationships between spatial and non-spatial data. Furthermore, the ontology is employed to represent geospatial and attribute domain knowledge related to road, environment and vehicle. Crash rules are acquired by integrating experts knowledge with the rules which are extracted using the Separate-and-Conquer rule induction approach. These rules are transformed to Semantic Web Rule Language (SWRL) syntax for reasoning in crash severity estimation engine. To evaluate the proposed method, a system prototype in the Qazvin-Rasht (Iran) regional transportation corridor as a case study is implemented. The results show that the proposed approach can efficiently induce major crash rules and predict accident prone vehicles crash severity with respect to real-time road, driver and environmental information in the vicinity of vehicle current location.