Fuzzy Decision Tree and Particle Swarm Optimization for Mining of Time Series Data

This paper presents a new approach for power signal time series data mining using Stransform based Kmeans clustering t echnique and fuzzy decision tree. Initially the power signal time series disturbance data are preprocessed through an advanced signal p rocessing tool such as Stransform and various statistical feature s are extracted, which are used as inputs to the Kmeans algorithm f or disturbance event detection. Particle Swarm Optimization (PSO) technique is used to optimize cluster centers which can be inputs to a fuzzy decision tree for pattern classification of time varying database like the power signal data bases.