An efficient adaptive preprocessing mechanism for streaming sensor data

Wireless Sensor Network (WSN) is fixed in many sensing environments to capture and monitor events. The sensing values come from sensor device that may contain noise, missing values, and redundant features. Noise, missing values and redundant features should be removed from the streamed data using an efficient preprocessing mechanism and then preprocessed data can be provided for further processing such as classification or clustering. If any errors occur in the streaming data then the preprocessing mechanisms should be able to handle the errors adaptively. Many preprocessing techniques are implemented for preprocessing streaming data, to adapt dynamic changes, and to handle different situations. The problem is the preprocessing systems does not efficiently handles different situations and adapt to changes for streamed sensor data. In this research, a new adaptive preprocessing mechanism is proposed that will efficiently handle changes in the incoming streaming data and scenarios are implemented to decouple the preprocessor and predictor in different situations for increasing the prediction accuracy. The proposed system uses PCA (Principal Component Analysis) as preprocessor and Hyperbolic Hopfield Neural Network (HHNN) as predictor. This method provides an efficient and adaptive preprocessing of streaming data.