Comparative Study of Publicly Available Classification Techniques

Sarma, Banashree (2019) Comparative Study of Publicly Available Classification Techniques. Masters thesis, Indian Institute of Science Education and Research, Kolkata.

PDF (MS dissertation of Banashree Sarma (14MS090)) 14MS090.pdf - Submitted Version Restricted to Repository staff only Download (9MB)

Abstract

Classification is a machine learning technique used to classify instances into their respective groups in accordance with some constrains. Several publicly used classification algorithms including LDA, QDA, Logistic Regression, Naive Bayes, SVM, Kernel discriminant analysis , Decision Tree, Bagging and Random Forest are discussed. Finally,a comparative study of publicly used classification algorithms is made . Various dataset are simulated in favour of some classfication technqiues and the results are verified. The difference between different methods along with their features , limitations and accuracies are discussed. I have shown that linear classifiers works well when the data are linearly separable and performs extremely bad in non-linearly separated data. Non-parametric classifiers works quite well in non-linearly separated data. I have used accuracy as evaluative measures. All the work and simulation is done in R.

Actions (login required)

View Item