Abstract

A novel construction final account prediction method based on relevance vector regression algorithm (RVR) optimized by ant colony optimization (ACO) with two-stage updating pheromone is presented in this paper, the ACO algorithm has been used to find global minimum, modifications based on the pheromone trail are then applied to each vector. RVR optimized by ACO with two-stage updating pheromone has more robust than classical relevance vector regression algorithm. The experimental results show that the prediction performance for construction final account of the prediction models trained by the training sample sets with 4 dimensional input vector than that of the prediction models trained by the training sample sets with other dimensional input vector, and the prediction performance for construction final account of TSUPACO-RVR than that of classical RVR regardless of dimensions of input vector in the training sample sets.

Paper Details