Performance Evaluation And Comparison Of Classification Techniques For Outcome Estimation In Strategic Board Games

Supervised learning aims to construct a distribution model for class labels with respect to features of prediction. Various machine learning approaches have been developed to analyze classification technique on different kinds of data.The objective of this work is to evaluate and compare the prediction performance of various classification techniques on 3 datasets belonging to strategic board games. This comparison analysis is done by using WEKA, open source software, which is responsible for implementing variety of machine learning algorithms for data-mining application i.e. classification.This work provides basic overview of selected machine learning classification models alongwith a brief description of datasets of three strategic board games. Then, it evaluates and compares the prediction performance of various classifiers using K-fold cross validation test mode. The results are based on several evaluation metrics like accuracy, precision, recall, kappa statistics, mean absolute error, and root mean squared error. Finally, it provides the best classification method for outcome prediction in strategic board games. Tree based LMT, SVM based SMO and K-NN based LBk are observed as the most suitable models for outcome prediction of strategic board games, LMT being the most influential one.