CS 235 Project Report: UCSD Data Mining Contest

In this report, we summarize the results of our participation in the UCSD Data Mining Competition. We describe the testing phases due to which we achieved a very high accuracy score (91.6%). We show that resampling the training set was a critical part in our study, and that combined with the Multilayer Perceptron algorithm were the reasons for our success. Finally, we discuss all the classifiers we used, and try to conclude our inferences from the testing experience. I. INTRODUCTION Data Mining is the process of pattern learning, by analyzing data. This kind of information extraction became an important tool as huge amount of data is now gathered in most of the applications. Thus, various data mining methods are proposed and employed in a vast number of fields from biology to