Study of Activity Recognition Dataset Using Combined Probabilistic and Instance Based Algorithms

Abstract

Wide range of applications involve classification process where supervised learning
approaches play a significant role. Improvement of classification accuracy is one of
the tasks which is most frequently carried out by researchers worldwide. This paper
describes selected statistical and instance-based approaches and presents hybrid classifier
for Human Activity Recognition. Obtained results outperform solely application
of each algorithm paradigm to the dataset and strengths the hypothesis for improving
classification accuracy by using ensembles of classifiers.