A Novel Random Forest-SMOTE Framework with Polynomial Feature Engineering for Early Detection of Gastrointestinal Disorders

Gastrointestinal disorders are a major health concern, and early detection is crucial for effective treatment. This project proposes a novel framework for the early detection of gastrointestinal disorders using machine learning (ML) techniques. The system employs various models, including Logistic Regression, Random Forest (RF), Voting Classifier, TabNet, TabTransformer, and Polynomial Random Forest, integrated with SMOTE (Synthetic Minority Over-sampling Technique) for handling class imbalance. The framework utilizes feature engineering techniques such as polynomial transformations to enhance model performance. The system processes user inputs like age, gender, body weight, and other relevant symptoms, predicting conditions like abdominal cramps, blood in stool, unexplained weight loss, diarrhea, nausea, and bloating. The project is implemented as a web application using Flask for the backend, with frontend development in HTML and CSS. After registering and logging in, users can provide their medical features, and the system will classify the likelihood of specific gastrointestinal disorders. This approach aims to provide an accessible and efficient tool for early diagnosis, improving healthcare outcomes.

Keywords:

Gastrointestinal Disorders, Early Detection, Machine Learning, SMOTE, Polynomial Feature Engineering, Logistic Regression, Random Forest, TabNet, TabTransformer, Voting Classifier, Web Application, Flask, Healthcare.