Data Reduction Based on Adaptive Stream Window Size for IoT Data

IoT produces data that is streamed to servers for processing. Such a stream is characterized by redundancy. Data streams are often processed in blocks referred to as windows with a given size known as the window size. The estimation of window size is affecting the accuracy and reduction rate. This paper uses a genetic algorithm to optimize window size for the reduction process. The genetic algorithm is combined with fuzzy subtractive clustering as a fitness function that is new method to choose window size. The experimental results show the impact of suitable window size on the accuracy and reduction rate for each type of sensor.