Bus Travel Time Extreme Learning Machine Prediction Model Based on Artificial Bee Colony Optimization

The real-time prediction of bus travel time is the key to the construction of intelligent bus system. Considering the factors affecting bus travel time, such as the length of the road section between the stops, the number of intersections, the operation period, the time of getting on and off passengers at the stops, and combining the advantages of artificial bee colony algorithm with few parameters and strong optimization ability and the simple and fast operation of extreme learning machine structure, an optimized extreme learning machine prediction model of bus travel time based on artificial bee colony algorithm was established. Based on the actual GPS data of a bus in Shenzhen City, the travel time record data of the bus stations were extracted, and the simulation experiment was carried out with MATLAB programming. The prediction algorithm proposed in this paper is compared with BP neural network, support vector machine and standard extreme learning machine model in the peak period with the greatest travel time fluctuation. The experimental results show that the proposed algorithm has higher prediction accuracy and better robustness.