PBFP: Probabilistic-Based Fault Prediction Method for Effective Fault Management in Distributed Sensor Network

Abstract

In the past few decades, the world has seen rapid growth in many practical applications in
sensor network, and this network domain has grown even more in terms of uniformity, complexity,
and scalability. However, the use of this technology causes new challenges to the advancement to
build reliable and fault-tolerant applications. As networks become more distributed, centralized
system management, which has barriers to access, flexibility, and growth, has affected the
management system. One of the most important is the prediction of failure in the large amount of
sensor data being collected. Furthermore, such failures may harm the quality of service (QoS)
management of failures. This problem is usually solved by using a temporal and spatial correlation
that predicts faults through a fault prediction algorithm. In this paper, we propose a ProbabilisticBased Fault Prediction (PBFP) method using Bayesian classification to classify the detected failure
data set. This method has been tested in a comprehensive set of data to assess its behavior for various
constraint values. The experiment evaluation yields better results showing that our algorithm could
significantly increase the number of hop responses by slightly reducing the classification accuracy.
Keywords: Fault Prediction, Bayesian, Classification, Distributed Sensor network.