N.A. Banushkina, E.B. Pechatnova
Improving the Efficiency of Traffic Accidents Prediction on Roads Outside of Settlements on the Basis of an Expert System Development
In the paper, the most common methods of traffic accidents (TA) prediction are considered. These methods include extrapolation, prediction with consideration of seasonality, and modeling of accidents frequency. Verification of these methods on real data of TA on roads of the Altai Region for the years 2011–2014 reveals low prediction quality due to the fact that the impact of road conditions on traffic safety has not been considered. Therefore, to improve the prediction quality, a special method is required that considers the multitude of various factors. The accident rate method that determines the impact of various independent factors on TA probability can be taken as a basis. The impact of each factor is expressed by its own coefficient. The total impact is defined as a product of the impact coefficients for the specific test segment of the road. A development technique for the expert system and the knowledge base for TA prediction on roads outside of settlements of the Altai Region is proposed. Its implementation includes the initial sampling of factors that supposedly impact on TA with further elaboration of the algorithmic function.
Key words: traffic accident, predicting methods, factors of influence, knowledge base, expert systems
Full text at PDF
, 924Kb. Language: Russian.