On-line learning and wavelet-based feature extraction methodology for process monitoring using high-dimensional functional data

The recent advances in information technology, such as the various automatic data acquisition systems and sensor systems, have created tremendous opportunities for collecting valuable data. The timely processing of such data for meaningful information remains a challenge. In this dissertation, several data mining procedures that will aid information streaming of high-dimensional functional data are developed.; For on-line implementations, two weighting functions for updating support vector regression parameters are developed. In order to apply these functions for on-line predictions, a new on-line support vector regression algorithm that uses adaptive weighting parameters was presented. The new algorithm uses varying rather than fixed regression parameters for training data. The developed functions and algorithm were applied to two different spectral data and two benchmark time series data. The results show that using adaptive regression parameters performs better than using fixed regression parameters.; In order to reduce the dimension of data with several hundreds or thousands of predictors and enhance prediction accuracy, a novel wavelet-based feature extraction procedure called step-down thresholding procedure for identifying and extracting significant features for a single curve was developed. The procedure is based on multiple hypothesis testing approach and it controls false discovery error rate in order to guide against selecting insignificant features. The procedure was compared to six other data-reduction and data-denoising methods in the literature. This procedure is found to consistently perform better than most of the popular methods.; Many real-world data with high-dimensional explanatory variables also have multiple response variables. In order to select the fewest explanatory variables that can predict each of the response variables better, a two-stage wavelet-based predictive modeling procedure was presented. The first stage uses step-down procedure for multiple curves to extract significant features. The second stage uses non-conjugate Bayesian decision theory approach and simulated annealing search method to select some of the extracted wavelet coefficients that can predict each of the response variables accurately. The two stage procedure was implemented using near-infrared spectroscopy and shaft misalignment data. The features were considerably reduced and the prediction results are encouraging.