Scribd
Upload
23 views4 pages

Stress Prediction Documentation

Random

Uploaded by

Zyam Maqsood
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
23 views4 pages

Stress Prediction Documentation

Random

Uploaded by

Zyam Maqsood
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

Human Stress Prediction

Group members:
• Zyam Maqsood
• Mohammad Hamza
• Syed Muneeb Naqvi

Course:
Comp-404 Artificial Intelligence

Dataset:
https://www.kaggle.com/datasets/kreeshrajani/human-stress-prediction/discussion/428141

Models implemented:
• Support Vector Machines (SVM)
• k-Nearest Neighbors (kNN)
Introduction:
Human stress is a critical psychological factor that affects well-being and productivity.
Automated stress prediction can aid in timely interventions, providing insights to healthcare
professionals and individuals alike. This project explores the application of Support Vector Machine
(SVM) and K-Nearest Neighbors (KNN) algorithms for classifying stress levels from textual data.
The dataset used for this project contains stress-related textual information labeled for binary
classification.

Data Description:
The dataset, Human Stress Prediction Dataset, consists of textual data labeled with stress
indicators:

• Columns:
◦ text: Contains text input describing situations, emotions, or expressions.
◦ label: Binary indicator of stress (1 for stressed and 0 for non-stressed).
• Preprocessing Steps:
◦ Converted text to lowercase.
◦ Removed punctuation marks (e.g., periods, commas, exclamation marks).
• Data Distribution:
◦ Total records: 568
◦ Non-stressed (0): 263 records

◦ Stressed (1): 305 records

Methodology:
• Preprocessing:
◦ The text data was cleaned to remove punctuation and converted to lowercase for
uniformity.
◦ No stop-word removal or stemming/lemmatization was applied.
• Feature Extraction:
◦ TF-IDF Vectorizer: Text data was transformed into numerical vectors using a maximum
feature limit of 5000 terms.
• Model Training:
◦ Support Vector Machine (SVM): A linear kernel was used to classify text as stressed or
non-stressed. This kernel is suitable for linearly separable data and efficient for high-
dimensional spaces.
◦ K-Nearest Neighbors (KNN): A neighborhood-based classification method, using 5
neighbors (n_neighbors=5) for prediction.
• Model Evaluation
◦ Models were evaluated using:
▪ Accuracy Score: Percentage of correct predictions.
▪ Classification Report: Includes precision, recall, F1-score, and support for both
classes.

Results:
• SVM Classifier
◦ Accuracy: 74.82%
◦ Classification Report:
Precision Recall f1-score support
0 0.73 0.73 0.73 263
1 0.77 0.76 0.77 305
Accuracy 0.75 568
Macro avg 0.75 0.75 0.75 568
Weighted avg 0.75 0.75 0.75 568

◦ Insight:
▪ The SVM model demonstrates robust performance, particularly for the stressed class
(1).
▪ A well-balanced precision and recall indicate reliable predictions across both classes.

• KNN Classifier
◦ Accuracy: 65.85%
◦ Classification Report:
Precision Recall f1-score support
0 0.72 0.43 0.54 263
1 0.63 0.86 0.73 305
Accuracy 0.66 568
Macro avg 0.68 0.64 0.63 568
Weighted avg 0.68 0.66 0.64 568
◦ Insights:

▪ While KNN demonstrates high recall for the stressed class (1), its performance on
the non-stressed class (0) is limited, with a lower precision and recall.
▪ Sensitivity to class imbalance and reliance on neighborhood distances likely
impacted its overall accuracy.

• Model Comparison

Metric SVM KNN


Accuracy 74.82% 65.85%
Precision (avg) 75% 68%
Recall (avg) 75% 64%
F1-Score (avg) 75% 63%

Conclusion:
• The SVM classifier outperformed the KNN classifier in terms of accuracy and classification
metrics, making it a more reliable model for stress prediction.
• KNN struggled with lower precision and recall for the non-stressed class, highlighting its
sensitivity to class imbalance and feature scaling.
• These results indicate that TF-IDF combined with SVM is effective for high-dimensional
text classification tasks such as stress detection.

You might also like