Process Framework for Modeling Multivariate Time Series Data

In multivariate time series (MTS) data, the formation of high-quality, reliable, and statistically sound information by analyzing and interpreting large data set is becoming a challenging task due to its increased complexity and over-fitting problems. Preprocessing steps play an important role in overcoming the performance issues of MTS data analysis. Feature and data subset selections are important preprocessing steps before applying any data mining functionalities like clustering and classification to identify the efficient and valuable predictors and relevant instances that better represents the underlying process of the data. Here we introduced an optimized preprocessing step using process control charts to extract a subset of key instances that form the representative set of the core group and utilized two classification algorithms to analyze the performance. The results are also compared for different test scenarios by adding standard dimensionality reduction methods and numerosity reduction approaches.