Suicide Prediction Project

Website: Suicide Prediction Website

This project focuses on predicting suicidal tendencies using various machine learning models and a custom deep learning model. It leverages advanced classification techniques to analyze text data and identify patterns associated with suicidal intent.

Features

• Implementation of multiple machine learning classifiers:
  • Logistic Regression
  • Multinomial Naive Bayes (MultinomialNB)
  • K-Nearest Neighbors (KNeighborsClassifier)
  • Random Forest Classifier
• Development of a custom deep learning model combining BiLSTM (Bidirectional Long Short-Term Memory) and RNN (Recurrent Neural Network).
• Creation of an ensemble model using a weighted average for probability aggregation.
• Evaluation metrics for performance comparison.

Dataset

The dataset used in this project is sourced from Kaggle Suicide Watch Dataset. It consists of labeled text data indicating suicidal intent. Preprocessing steps include text cleaning, tokenization, and vectorization.

Models and Metrics

1. Logistic Regression

• Accuracy: 0.9325
• Precision: 0.9452
• Recall: 0.9188
• F1-Score: 0.9318

2. Multinomial Naive Bayes (MultinomialNB)

• Accuracy: 0.9052
• Precision: 0.8724
• Recall: 0.9502
• F1-Score: 0.9096

3. K-Nearest Neighbors (KNeighborsClassifier)

• Accuracy: 0.8754
• Precision: 0.9165
• Recall: 0.8272
• F1-Score: 0.8696

4. Random Forest Classifier

• Accuracy: 0.9030
• Precision: 0.8965
• Recall: 0.9123
• F1-Score: 0.9043

5. Custom Deep Learning Model (BiLSTM + RNN)

• Accuracy: 0.9300
• Precision: 0.9031
• Recall: 0.9640
• F1-Score: 0.9326
• Loss: 0.2075

6. Ensemble Model (Weighted Average)

• This model combines the probabilities from multiple classifiers using a weighted average approach, improving overall robustness and predictive performance. Evaluation metrics will be provided upon execution.

Results

The project achieves high accuracy and F1-scores across all models, with the custom BiLSTM + RNN model demonstrating superior performance in recall and F1-score. The ensemble model provides an additional layer of robustness by combining predictions from multiple classifiers.

Conclusion

This project demonstrates the potential of machine learning and deep learning models in sensitive tasks such as suicide prediction. It provides a comprehensive framework for analyzing text data and highlights the importance of balancing accuracy, precision, recall, and F1-score for reliable predictions.