Human Activity Recognition Using Ensemble Modelling
Amandeep Kaur, Sahil Sharma · Communications in computer and information science · 2016
In pervasive computing, human basic activity recognition has become one of the major challenges as recognizing every day basic activities and then classification of diverse activities using various devices has become arduous task. With the help of various machine learning models and data mining tools prediction has been applied. The dataset has total 10299 labelled activity instances with 561 features, to get the results more optimize we have successfully reduced features to 35 and the results were brilliant. Various machine learning models have been evaluated on the dataset for prediction of human basic activities. Results show that the large features and the reduced features were almost maintained in the terms of accuracy. The best models have been investigated for ensemble learning to get sustainable results on the basis of accuracy to classify the set of common activities carried on whole day. Encouraging results have been obtained with ensemble model. Cross validation has been performed to check the consistency of the ensemble model and accuracy more than 85% has been obtained. Finally, various human activities have been classified using ensemble model with good results.