Multimodal detection of stress levels that increase academic performance

Date of Publication

4-3-2013

Document Type

Master's Thesis

Degree Name

Master in Computer Science

Subject Categories

Computer Sciences

College

College of Computer Studies

Department/Unit

Computer Science

Thesis Adviser

Rhia Trogo

Defense Panel Chair

Merlin Teodisia Suarez

Defense Panel Member

Jocelyn Cu
Rhia Trogo

Abstract/Summary

This research aims to build a model that identifies the stress levels correlated to the performance of the user in terms of academic work based on the user’s physiological state and throughput. Two experimental set ups were conducted. Non-invasive physiological signals that were used for this research were Blood Volume Pulse (BVP), Galvanic Skin Response (GSR), and Respiration Variability (RS). These signals underwent three signal pre-processing stages: (1) filtration to remove the noise in the data; (2) activity segmentation; (3) fixed windowing and overlapping segmentation; (4) feature extraction; and, (5) normalization of the data. Using kNN = 5, the stress model obtained 79.82% accuracy (Kappa of 0.74) for controlled set up and 84.77% accuracy (Kappa of 0.78) for the naturalistic set up. As for the performance model, kNN = 5 still got the highest result among the other machine learning algorithms. The controlled set up was 75.31% accurate while 84.77% accuracy for the naturalistic set up. The generated tree of the model did show that the inverted U relationship of stress and performance is precise.

Abstract Format

html

Language

English

Format

Electronic

Accession Number

CDTG005346

Shelf Location

Archives, The Learning Commons, 12F Henry Sy, Sr. Hall

Keywords

Stress (Psychology); Academic achievement; Students—Psychology; Stress (Physiology)

Upload Full Text

wf_no

Embargo Period

7-3-2023

This document is currently not available here.

Share

COinS