Prediction of Rural Travel Demand by Spatial Regression and Artificial Neural Network Methods (Tabriz County)

Understanding of the current travel pattern is necessary for identifying and analyzing traffic problems, the movement of people and for developing travel forecasting and prediction models. The prediction of travel demand is the main goal of the long-term transportation planning process for determining strategies about accommodating future travel needs, which may include land use policies or the development of transportation, road construction services and so on. This paper aims to estimate travel demand based on spatial statistics and artificial neural network methods as capable approaches, then compare the results of these methods to determine the accuracy of each to forecaste travel demand. The present research is of applied type and the method of doing it is descriptive-analytical. The statistical population of the study is all rural households in Tabriz County, which is 40,714 households, and using the Cochran's formula, 380 households were selected as the sample population. Data collection was done using a researcher-made questionnaire and based on research indicators, and as a result, the actual travel demand of individuals was estimated. After estimating the actual travel demand, travel estimation was performed using conventional least squares method, geographical weight regression and multilayer perceptron neural network. The result was compared with the actual travel estimates. The results indicate that geographic weight regression (GWR), artificial neural network (ANN) and ordinary least squares (OLS) have the highest to the lowest accuracy in estimating travel demand respectively.