Reducing Data Stream Complexity by Applying Count-Min Algorithm and Discretization Procedure

This paper is a preliminary report on the efficiency of two strategies of data reduction in a data preprocessing stage. In the first experiment, we apply the Count-Min sketching algorithm, while in the second experiment we discretize our data prior to applying the Count-Min algorithm. By conducting a discretization before sketching, the need for the increased number of buckets in sketching is reduced. This preliminary attempt of combining two methods with the same purpose has shown potential. In our experiments, we use sensor data collected to study the environmental fluctuation and its impact on the quality of fresh peaches and nectarines in cold chain.