A comparative Analysis of Machine Learning Classification Approaches for Fountain Data Estimation in Wireless Sensor Networks

Wireless Sensor Networks attract nowadays a great deal of not only research but also the industrial. It is deployed in a variety of area such as military, health care, monitoring. Energy is the main challenge of this network. When providing fountain codes with the assistance of training machine learning models, their ability to accurately determine the needed number of encoded packets significantly improves. In this paper, we discussed and compared our proposed distributed estimation scheme with some machine learning based methods for data classification. Simulations show that our proposed scheme which is based on the Bayesian model looks advantageous over other methods. Consequently, we can determine the needed number of encoded packets to recover initial data with appreciable accuracy and error rate notably low.