PREDICTING WILLINGNESS TO SHIFT TOWARDS BICYCLE USING MACHINE LEARNING MODELS AND QUANTIFYING ITS ECONOMIC BENEFITS: EVIDENCES FROM KOLLAM COASTAL AREA

Abstract

The increase in vehicle ownership among the cities is mainly due to the predicted economic growth. These growth leads to the global warming, emission of greenhouse gas, hike in price of different fuels and traffic congestion. Due to this problems in the environmental and transportation, the health and the life style of the society is badly affected. So to reduce these effects, one of the solutions is switching to non – motorized traffic such as bicycling and walking. In this thesis, a study was conducted to know the people’s willingness to shift towards non-motorized transports. The study also determined the different factors such as socio economic, environmental and transportation system characteristics that affect the commuters’ choice on using bicycle as mode of travel. For the purpose of the model development, about 868 household surveys were collected in Thangassery-Thanni coastal road stretch through household interview survey. The user perception survey was conducted for identifying the perception of society about the use of bicycle as main or feeder mode. The Adaboost Classifier, a machine learning technique was used to analyze the important attributes that influences the shifting towards bicycle the study incorporated three supervised machine learning algorithms (K-Nearest Neighbor, Support vector machine, Random Forest) to predict willingness to shift towards bicycle as a mode of transport. The results indicated that Random Forest model outperformed with an accuracy of 0.94. A health benefits analysis in terms of mortality when bicycle is used as mode of travel was carried out using Health Economic Assessment Tool (HEAT) for 1% mode shift. The total economic value of carbon emission on all pathways is monetized as 3870 USD (302556 INR) and considered as the health benefit that can be saved per day while cycling in the study stretch