Comparing performance of k-Nearest Neighbors, Parzen Windows and SVM Machine Learning Classifiers on QSAR Biodegradation Data across Multiple Dimensions

Machine learning and pattern recognition are the most popular artificial intelligence techniques to model systems, those can learn from data. These techniques efficiently help in Classification, Regression, Clustering and Anomaly detection etc. k-Nearest Neighbors, Parzen Windows and Support Vector Machine (SVM) are some of the widely used Machine Learning classification techniques. This project aims to experimentally compare several features of these classification techniques using QSAR Biodegradable dataset over different dimensions using Principal Component Analysis (PCA). The results of the experiment demonstrate that SVM performs way better than k-Nearest Neighbor and Parzen window and, k-Nearest Neighbor performs a little better over Parzen window on classification accuracy. Also, I establish that ERBF performs better than other kernel functions (RBF, Polynomial and Linear) when used for SVM.