Class Imbalance problem is unavoidable problem foractivity recognition using mobile sensors in real lifescenario when we have very less amount of data for someactivity classes. It can affect the accuracy of thealgorithms for classification. In this paper we measure theimpact of class density in the accuracy of classification inimbalance cases. It is important for understanding theproblem better that can help in finding better solution forclassification in this scenario. Our initial experimentshows that- class imbalance affects the performance ofclassifier negatively and the higher the value of density(lower deviation), the better the performance of theclassifier becomes.
展开▼
