A research study submitted in partial fulfillment of the requirements forthedegree of Master of Science inInformation Management, School of Engineering of Technology

Research Studies Project Report (M.Sc.) - Asian Institute of Technology, 2012

In oil retail industries, gas stations are builtto serve households and industries. An operation of a gas station has a very high risk of loss. To reduce thatrisk, managements need to consider several factors influencing revenues of the station before building a new gas station.With the rapid development in information technology, many different data mining approaches are applied to support management{u2019}s decisions. This study focuses on using several data mining techniques to classify the potential of gas stations. The potential means the capability of growth or being without loss; therefore, this study uses sales volume as an indicator of the potential of gas stations. Using classification techniques in the data mining candiscover some hidden knowledge on existing gas station data and other related information and the knowledgecan also be used for helping the management making a decision for building the new station,which is very beneficial.This study conducted 3 experiments, which use artificial neural networks, support vector machine and decision trees. The results show that using artificial neural networks hasthe highest accuracy in classifying the potential of gas stations. It is more than 85% accuracy for all models. On the contrary, using support vector machine and decision trees, both of them getloweraccuracy rate on testing data. Based on these results,the artificial neural networks technique can serve as a decision support tool for classify a potential of a new gas station, which can reduce the human error in the decision-making process or even help to the management to make a decision with very high accuracy.