Comparison of Annual Average Daily Traffic Estimates: Traditional Factor, Statistical, Artificial Neural Network, and Fuzzy Basis Neural Network Approach

Reliable estimation of Annual Average Daily Traffic (AADT) is an essential element of statewide traffic monitoring programs. In the traditional factor approach, one- or two-day short-duration traffic volume counts are converted into AADT estimates by applying certain volume adjustment factors. This paper extends past research by investigating several new methods --- Analysis of Variance (ANOVA), Artificial Neural Network (ANN), and Fuzzy Basis Function Network (FBFN) --- that can assist in deriving AADT estimates. The performances of the traditional factor approach, ANOVA, ANN and FBFN are evaluated by applying them to the continuous traffic volume data from Indiana’s automatic traffic recorder (ATR) sites. The simulation results show that AADT estimates from the ANOVA method are the most consistently satisfactory among the four approaches. The ANN and FBFN methods have very close simulation results to the traditional sixty or eighty-four factor approach. The ANOVA, ANN and FBFN methods are generally superior alternatives to the traditional factor approach for AADT estimation.