Combination of Feature Selection and Learning Methods for IoT Data Fusion

In this paper, we propose five data fusion schemes for the Internet of Things (IoT) scenario, which are Relief and Perceptron (Re-P), Relief and Genetic Algorithm Particle Swarm Optimization (Re- GAPSO), Genetic Algorithm and Artificial Neural Network (GA-ANN), Rough and Perceptron (Ro-P) and Rough and GAPSO (Ro-GAPSO). All the schemes consist of four stages, including preprocessing the data set based on curve fitting, reducing the data dimension and identifying the most effective feature sets according to data correlation, training classification algorithms, and finally predicting new data based on classification algorithms. The results derived from five compound schemes are investigated and compared with each other with three metrics, namely, Quality of Train (QoT) Accuracy (Ac) and Storage Capacity (SC). While the Re-P scheme is only capable of separating classes that are linearly separable, Re-GAPSO one is a dynamic method, appropriate for constantly changing problems of the real life. On the other hand, GA-ANN is a Wrapper method and despite Relief can adapt itself to the machine learning algorithm. Meanwhile, Ro-P scheme is useful for analyzing vague and imprecise information and, unlike GA-ANN, has less calculative costs. Among these five schemes, Ro-GAPSO is a more precise one, which has less calculative cost and does not become stuck in local minima. Experimental results show that Re-P outperforms other proposed and existing methods in terms of computational time complexity.