A Comparative Study of Random Forest & K – Nearest Neighbors on HAR dataset Using Caret
Kella BhanuJyothi, K HimaBindu, D. Suryanarayana
Classifier, rf, ParRF, knn, Accuracy, HAR Dataset, caret package
In this paper, we have compared the accuracy results of the Classification models Random Forest Algorithm and K – Nearest Neighbors on UCI HAR (Human Activity Recognition) dataset and selecting the best classifier among the models to test the dataset. The best classifier is choosing by accuracy of the model. The classifiers for the dataset are built from the methods ‘rf’, ‘parRF’ and ‘knn’ by using Caret package in R. The classifier is used to build on the data which have high quality. If, the data contains incorrect and noise information then the accuracy we will receive is reduced. So, before preparing the classifier the dataset should be preprocessed to get exact accuracy. From the three methods, we found that the method ‘rf’ have high accuracy 93.13% on the trained modelHence, which is considered as a best classifier and applied the method ‘rf’ on the test data of HAR dataset.