A prediction model for effective data aggregation materials and fault node classification in WSN

• Main Authors: Amutha, R., Sivasankari, G.G., Venugopal, K.R.
• Format: Conference Proceeding
• Language: English
• Subjects: Data aggregation; Faulty Node Detection; LSTM; WSN
• ISSN: 2214-7853; 2214-7853
• DOI: 10.1016/j.matpr.2021.11.370

With the integration of wireless technology with an embedded technology called Wireless Sensor Network (WSN), the concept of WSN has become even more advanced. WSN is used for monitoring various environmental factors such as pollution, weather, humidity and temperature. It consists of a variety of sensitive microelectronic components for broadcasting and receiving data. The main limitation of WSN is network lifetime due to its battery-operated resources. The key solution to prolong network life is the data aggregation approach. Data aggregation is a process by which the redundant data collected from all the sensor nodes is removed from further processes like routing. Faults present in Sensor Nodes considerably affect the efficiency of data aggregation. There are two ways to improve the process of data aggregation: Early Prediction and Faulty Data Classification. In this work, a novel LSTM based data prediction and classification algorithm is proposed for efficient data aggregation. The algorithm is combined with AdaBoost classifier to increase Classification accuracy. Experimental results show that the proposed approach outperforms other algorithms in terms of Faulty Node Detection, Data Quality, Minimized Energy Consumption, Overhead Time and Live Node Rate etc. Compared to other methods, the proposed approach achieves 5.72% of increased Faulty Node Detection rate, 31% increased Data Quality, 8.2% of Minimized Energy Consumption and 36.5 % increased Live Node Rate.