Improved SAX time series data representation based on Relative Frequency and K-Nearest Neighbor Algorithm

In this paper we propose a new approach based on Symbolic Aggregate approximation (SAX), called improved iSAX to recognize efficient and accurate discovery of the important patterns, essential for time series data. The original SAX approach allows a very high-quality dimensionality reduction and distance measures to be defined on the symbolic approach and it is based on PAA (Piecewise Aggregate Approximation) representation for dimensionality reduction that minimizes dimensionality. The proposed improved SAX, called iSAX includes the Relative Frequency and K-Nearest Neighbor (RFknn) Algorithm. The main task of the algorithm is to determine the sufficient number of intervals represented as symbolic (alphabet size) that can ensure efficient mining process and a good knowledge model is obtained without major loss of knowledge. We show that iSAX can improve representation preciseness without losing symbolic nature of the original SAX representation. The iSAX is compared with the original SAX and PAA representation, and demonstrate its quality improvement. Ten time series rainfall data sets were used. The experimental results showed that iSAX gives better term of representation and minimum Euclidean Distance.