Research On Forecasting Methods Of Road Accidents

The road accidents forecasting is the important content of the research of traffic safety. The purpose of road accidents forecasting is to analyze the tendency of road accidents under existing road traffic conditions, evaluate the feasibility and practical effectiveness of traffic safety measures reasonably, control the factors affecting road accidents, and reduce the traffic accidents. The characteristics such as nonlinearity, randomness and uncertainty in traffic system make it difficult to forecast road accidents, the behavioral feature of traffic system. Based on the analysis of existing macroscopic forecasting methods of road accidents, the forecasting methods of road accidents in consideration of multiple attribute characteristic of traffic system is researched in this thesis. The main research contents and conclusions are as follows:l.The characteristic and complexity of the road accidents forecasting is analyzed and expounded in this thesis,2. For the accident data with random fluctuation, the forecasting precision of road accidents is not good. Based on grey forecasting method of roads accidents, the markov chains forecasting method is presented and the grey-markov model for forecasting road accidents is built in this thesis. The grey-markov model for forecasting road accidents has the merits of both grey forecasting and markov chains forecasting which reduces random fluctuation of accident data affecting forecasting precision and widens the applicative scope of grey forecasting.3. In view of the characteristic that the traffic system is a dynamic and time-varying parameter system, the multi-level recursive forecasting method is proposed, and the multi-level recursive forecasting model of road accidents is established in this thesis. In this method, the forecasting of road accidents is divided into two parts: the forecasting of time-varying parameters and the futureforecasting of road accidents based on the forecasting of time-varying parameters. By the precise forecasting of time-varying parameter, the forecasting precision of road accidents can be improved.4. In view of the multiple attribute characteristic of traffic system and deficiency of taking the minimum error sum of squares as the choosing standard of forecasting methods of road accidents, the comprehensive forecasting mode of road accidents is proposed in this thesis. At the same time, the comprehensive forecasting model base on neural network is established. This method can take advantage of the useful information from every single forecasting method of road accidents comprehensively and the forecasting value of road accidents is the nonlinear function of the forecasting value of every single forecasting method. All these can improve the forecasting precision of road accidents.5. A few rules, which should be considered when we choose the forecast models of road accidents is provided in this thesis. And a compact comment on the established forecast models in this thesis is put forward according to the rules.6. Based on the research of the forecasting models and methods of road accidents, the MATLAB's GUI (Graphical User Interface) development technology is applied and the auxiliary software of road accidents forecasting is developed.