A Dynamic Approach for Mining Generalised Sequential Patterns in Time Series Clinical Data Sets

Abstract

Similarity based stream time series is gaining ever-increasing attention due to its importance in many applications such as financial data processing, network monitoring, Web click-stream analysis, sensor data mining, and anomaly detection. These applications require managing data streams, i.e., data composed of continuous, real-time sequence of items. We propose a technique for pattern matching within static patterns and stream time series clinical data sets. The main objective of our project is to ascertain hidden patterns between incoming time series clinical data sets and the set of predetermined clinical patterns. By considering the incoming image data at a particular timestamp, we construct a MultiScale Median model at multiple levels to adapt to the stream time series, characterized by frequent updates. Further, we employ a pruning algorithm, Segment Median Pruning on clinical Image data for pruning all candidate patterns. Experiments have been carried out on retinal disease data set known as Age Related Macula Degeneration (ARMD) and simulation results show that the system is efficient in processing image data sets for making efficient and accurate decision.