Speech Communication & Signal Processing Laboratory Faculty of Electronics and Computer Sciences, USTHB, Algiers, Algeria

---

Abstract—This paper is concerned with the class imbalance problem in activity recognition field which has been known to hinder the learning performance of classification algorithms. To deal this problem, we propose a new version of the multi-class Weighted Support Vector Machines (WSVM) method to perform automatic recognition of activities in a smart home environment. Then, we compare this approach with CRF, k-NN and SVM considered as the reference methods. Our experimental results carried out on various real world imbalanced datasets show that the new WSVM is capable of solving the class imbalance problem by improving the class accuracy of activity classification compared to other methods.

Index Terms—Imbalanced data, support vectors machines, activity recognition, machine learning.

[PDF]

Cite: M’hamed B. Abidine and Belkacem Fergani, "A New Multi-Class WSVM Classification to Imbalanced Human Activity Dataset," Journal of Computers vol. 9, no. 7, pp. 1560-1565, 2014.