A researchsubmitted in partial fulfillment of the requirements for thed egree of Masterof Science in Information Management, School of Engineering and Technology

Research studies project report (M. Sc.) - Asian Institute of Technology, 2016

Mining techniques have proven to be effective in exploring data. In this report, the efficiency of several data-mining methods is explored. In particular, we apply these methods to the predictive modelling competition Titanic: Machine Learning from Disaster currently active at kaggle.com, a website for such competitions. This particular competition is a classification challenge to build a model to predict which passengers on the Titanic survived. The focus of our approach is comparing different data-mining techniques such as K-neighbourhood, Logistic Regression, Support Vector Machine, XGBoost, Linear Regression, Stochastic Gradient Decent, Decision Tree, Naive Bayes and Random Forest algorithms. Results indicate that the predictors' gender, ticket price, embarked port, age, title, and passenger class are the most important variables to predict survival of the passengers. According to the results, Random Forest classifier has gained the highest accuracy of nine classifiers with a score: "0.80861" (322 out of 3667) top 10% on the Titanic: Machine Learning from Disaster Competition.