On the Effects of Input Unreliability on Classifion Algorithms

The abundance of data available on Wireless Sensor Networks makes online processing necessary. In industrial applications, for example, the correct operation of equipment can be the point of interest. The raw sampled data is of minor importance. Classification algorithms can be used to make state classifications based on the available data for devices such as industrial refrigerators. The reliability through redundancy approach used in Wireless Sensor Networks complicates practical realizations of classification algorithms. Individual inputs are susceptible to multiple disturbances like hardware failure, communication failure and battery depletion. In order to demonstrate the effects of input failure on classification algorithms, we have compared three widely used algorithms in multiple error scenarios. The compared algorithms are Feed Forward Neural Networks, naive Bayes classifiers and decision trees. Using a new experimental data-set, we show that the performance under error scenarios degrades less for the naive Bayes classifier than for the two other algorithms.