Machine learning models to perform sentiment analysis for mental health statuses.
Training and testing data can be found in kaggle-sentiment-data.csv.
The original model created by Muhammad Faizan, a researcher on Kaggle, can be found in sentiment-analysis-for-mental-health-nlp.ipynb.
This model utilizes logistic regression classification. However, this model was discovered to have data leakage into the test set due to data augmentation occuring before the data was split into train and test sets. A new baseline model was created to move data augmentation after the data is split into train and test sets in order to prevent data leakage into the test set. This model can be found in baseline_sentiment-analysis-for-mental-health-nlp.ipynb.
An improved model can be found in logistic_regression.ipynb.
This model improves upon the baseline model by removing data augmentation, changing hyperparameters, stratifying the status in train-test split, and making slight modifications to preprocessing.
The model can be found in random_forest.ipynb.
The model can be found in naive_bayes_classifier.ipynb.
The model can be found in gradient_boosting_machine.ipynb which runs GridSearchCV to find the best parameters for the model.
The model found in gradient_boosting_machine_best_params.ipynb uses the parameters obtained from gradient_boosting_machine.ipynb to get the best accuracy for this model.
The model can be found in support_vector_machine.ipynb which runs GridSearchCV to find the best parameters for the model.
The model found in support_vector_machines_best_params.ipynb uses the parameters obtained from support_vector_machine.ipynb to get the best accuracy for this model.