**5. Conclusions**

In this study, different ensemble classifiers with different base learner algorithms were implemented to classify six human daily activities based on tri-axial inertial smartphone data. Comparative studies of classification techniques were presented using Bagging, Adaboost, Rotation forest, END and Random subspace with base learner as SVM and RF. The performance measures used to evaluate the classification techniques include overall accuracy, precision, recall, F-measure and ROC. Holdout and 10-fold cross-validation evaluation methods were used in the model evaluation of classification. As seen from the obtained results, Random subspace classifier with SVM gives the best results overall accuracy rate over different ensemble classifiers. The comparison in each activity classification showed the overall results performance of precision, recall, F-measure and ROC and accuracy rate using 10-fold cross-validation method was slightly higher compared to the holdout method. It can be summarized that ensemble classifiers have produced improved performance for the HAR with six different activities such as walking, walking upstairs, walking downstairs, sitting, standing and lying. In future, other methods would be explored to improve the performance.

**Author Contributions:** I.E. contributed to the conception of the idea and the layout of the research plan, K.N. was involved in the construction of the search queries, participated in the literature search, carried out the research and wrote major sections of the manuscript. L.I.I. and I.E. assisted in reviewing while G.C. edited this manuscript. All authors equally contributed to the rest of the paper. All authors read the manuscript and approved its content.

**Funding:** Funds for this research was provided by Ministry of Higher Education (MOHE), Malaysia through FRGS Grant, 0153AB-L28, and Universiti Teknologi PETRONAS.

**Acknowledgments:** The authors would like to thank Universiti Teknologi PETRONAS, Malaysia and HOSEI University, Japan for supporting this work.

**Conflicts of Interest:** The authors declare no conflict of interest.
