Modelling of Activity-Travel Pattern with Support Vector Machine

Activity based travel demand modelling involves lot of uncertainty due to the complex and varying decision making behaviour of each individual. This study contributes to the literature by assessing the suitability of Support Vector Machine (SVM) in modelling the activity pattern and travel behaviour of workers. Activity and travel behaviour of workers consists of decision outcomes, which can be modelled as classification and regression problems. SVM is a good classifier and regressor with good testing and learning capability, hence the present study used SVM for modelling. It was found that support vector machine models are well performing to predict the activity pattern and travel behaviour of workers. The SVM models developed in the study predicts the temporal variation of mode wise work activity generation. Prediction of temporal mode share of commuters is advantageous to policy makers to experiment the implementation of temporary Travel Demand Management (TDM) actions effectively.