Comparison and Performance Analysis of Machine Learning Algorithms for the Prediction of Human Actions in a Smart Home Environment.

In this paper, we present a set of experiments to compare the performance of machine learning algorithms for the prediction of human actions in a smart home environment. These experiments use data from the MavPad dataset which was gathered from a real-world environment and tracks the activities of an individual over a 49-day period. We investigated the use of single and multiple groups of sensors for prediction and employed four different machine learning approaches with both accuracy and execution time performance used for the comparisons. The results showed that a group of seven sensors produces the most accurate results and that the Support Vector Machine approach was more accurate over all combinations of input data.