An evolutionary feature selection algorithm for classification of human activities

Abstract

This paper proposes an evolutionary feature selection algorithm to classify human activities. Feature selection is one of the key issues in machine learning, along with classification when some parts of features are not available or have redundant information. It enhances learning accuracy by selecting essential features and eliminating nonessential features. In the proposed algorithm, a feature selection algorithm integrated with an evolutionary algorithm (EA) is developed. We use the wrapper approach, which repeatedly calls the learning algorithm to evaluate the effectiveness of the selected features. Quantum-inspired evolutionary algorithm (QEA) is utilized as an evolutionary algorithm and multi-layer perceptron (MLP) is used as a classifier. The proposed algorithm is applied to classification of the human activities using smartphone sensors.