Risk Prediction Based on Time and GPS Patterns

Traffic produces not only pollution but also many incidents resulting in material lost and human injuries or even persons dead. But not all the incidents involve two cars, it may be pedestrian and any type of cycles (motorcycles, bicycles, tricycles, etc.). Most of the approaches try to model traffic accidents using traditional information and avoiding others, such as environmental elements, driver profile, weather, regulations, eventual circumstances like strikes with roadblocks, street reparations, railroads crossings, etc. This paper presents a model for risk prediction, and the impact of varying geographical information details on the precision of the underlying Inference System (a Soft Computing model with a ruled Expert System and a Harmonic System focused on time patterns of events). Its flexibility and robustness has a price: certainly minimal to apriori knowledge. This work outlines the working model implemented as a prototype named KRONOS, and a statistical evaluation of its sensibility to dynamic GPS information. Traffic risk requires this type of flexible and adaptive model due to the high number of alternatives to consider. The model would also be improved by adding certain specific Fuzzy Logic for pattern management during the matching process. The model would also be improved by adding certain specific Fuzzy Logic for pattern management during the matching process.