MULTIVARIATE TIME SERIES PATTERN RECOGNITION USING MACHINE LEARNING AND DEEP LEARNING METHODS

Author

Sai Abhishek Devar

Graduation Semester and Year

2019

Language

English

Document Type

Thesis

Degree Name

Master of Science in Industrial Engineering

Department

Industrial and Manufacturing Systems Engineering

First Advisor

Shouyi Wang

Abstract

In this research work, we have implemented machine learning & deep-learning algorithms on real-time multivariate time series datasets in the manufacturing & health care fields. The research work is organized in two case-studies. The case study-1 is about rare event classification in multivariate time series in a pulp and paper manufacturing industry, data was collected of multiple sensors at each stage of production line, the data contains a rare event of paper break that commonly occurs in the industry. For preprocessing we have implemented sliding window approach for calculating first order difference method to capture the variation in the data over the time. The sliding window approach helps to arrange the data for early prediction for instance we can set sliding window parameters predict two or four minutes early as required. Our results indicate that for case study-1 best accuracy score was produced by TensorFlow deep neural network model it was able to predict 50% of failures and 99% of non-failures with an overall accuracy of 75%. In case study-2 we have brain eeg signal data of patients which was collected with the help of Stereo EEG Implantation strategy to measure their ability to remember words shown to him/her after distracting him /her with math problems and other activities. The data was collected at a health-care lab UT-Southwestern Medical Center. The brain eeg signal data collected by the company was preprocessed by using Pearson’s and Spearman’s correlations, extracting bandwidth frequencies and basic statistics from eeg signal data extracted for each event, event in case study-2 refers to a word shown to a patient. We have used minimum redundancy and maximum relevance feature selection method for dimensionality reduction of the data and to get most effective features out of all. For case-study 2 best results were produced by SVM-RBF i.e. 73% accuracy to predict if a patient will remember or not remember a word.

Keywords

Multivariate time series, Rare event, Machine learning, EEG-signal, Sliding window, Classification, Deep learning

Disciplines

Engineering | Operations Research, Systems Engineering and Industrial Engineering

License

This work is licensed under a Creative Commons Attribution-NonCommercial-Share Alike 4.0 International License.

Comments

Degree granted by The University of Texas at Arlington

Recommended Citation

Devar, Sai Abhishek, "MULTIVARIATE TIME SERIES PATTERN RECOGNITION USING MACHINE LEARNING AND DEEP LEARNING METHODS" (2019). Industrial, Manufacturing, and Systems Theses. 19.
https://mavmatrix.uta.edu/industrialmanusys_theses/19