The Impacts on Fuel Consumption: A Data Mining-Based Analysis

摘要

This paper suggests different driving techniques based on the results of an applied research on the ecodriving domain, supplemented by a huge data set produced from Delhi’s transport system. The data set is based on events automatically extracted from the control area network and enriched with information like GPS coordinates, weather and road data. We use online analytical processing (OLAP) and knowledge discovery (KD) techniques which handles the high volume of data and to determine the major factors that determine the average fuel consumption, and assist in classifying the drivers based on their driving efficiency. Our findings leads to the introduction of simple practices, such as optimal clutch, engine rotation, engine running in idle state and traffic precautions, can reduce fuel consumption on average from 3 to 5l/100 km, meaning a saving of thousands of litre of petrol per day. With the availability of traffic through various traffic sensors, a lot of research effort has been involved in developing traffic prediction techniques, which in turn improve route navigation, traffic minimisation etc. One key boon in traffic prediction is the reliability on prediction models that are constructed on the basis of historical data applied in realtime traffic situations, which may differ from that of the historical data and has a tendency to change over a period of time. We aim in obtaining and proving both shortterm and longterm performance bounds for our online algorithm. The proposed algorithm also works effectively in scenarios where the realized traffic are missing or are available with a delay. In this paper we used a novel online framework that could learn from the current traffic situation in realtime and predict the future traffic by matching the current situation which will be useful for the effective prediction of fuel consumption and a suitable driving scheme.