• Performed pre-processing such as modifying column names, visualizing missing values using missmap diagram. • Split the whole dataset into train and test to perform validation of the models generated. • Selected significant features using correlation, rank by importance and recursive feature elimination techniques to generate model with few variables that can explain maximum variability and avoid overfitting. • Implemented 10-fold cross validation techniques and calculated the accuracy, AUC to tune the parameters of the model. • Build following 7 classification models to predict the masking label using the median response time of selected features(Letters). 1. Logistic Regression 2. Naive Bayes 3. K-nearest Neighbor 4. Decision/partition Tree 5. Random Forest 6. SVM 7. Neural Network • Generated an ensemble model by combining the output of all the above seven models to achieve 100% accuracy on the test dataset.

Coursera | John Hopkins University | Reproducible Research | Course project 2