Learning and building in Machine Learning and NLP - from classical and supervised ML to neural networks (ANN, CNN), with hands-on NLP using NLTK and Word2Vec.
I'm currently learning AI/ML, with hands-on experience across classical and supervised machine learning, NLP (NLTK, Word2Vec), and neural networks (ANN, CNN). I was also part of an AI/ML research.
A machine learning project predicting Titanic passenger survival. Compares seven classification algorithms - Decision Tree, Logistic Regression, SVM, AdaBoost, Gradient Boosting, Random Forest, and XGBoost - using a clean, modular, leak-free pipeline with feature engineering, class imbalance handling (SMOTE), and hyperparameter tuning.
View on GitHub ->Predicting medical insurance charges by benchmarking 10 regression models, from linear regression to XGBoost, with systematic hyperparameter tuning, reusable scikit-learn pipelines, and domain-driven feature engineering.
View on GitHub ->EDA and ML preprocessing pipeline on 20,000 Indian patient records across 32 diseases. Handles structured missing data, ordinal encoding, multi-label symptom binarization, and feature engineering to produce an ML-ready dataset.
View on GitHub ->An exploratory data analysis project on 7,668 movies released between 1980 and 2020, sourced from IMDb. Covers data cleaning, genre and director analysis, correlation studies (budget vs gross, runtime vs score), and country and decade trends.
View on GitHub ->Predicting body fat percentage from simple body measurements using ElasticNet regression, because BMI alone is not enough. Addresses data leakage, engineers BMI as a feature, and handles multicollinearity through regularization instead of dropping data.
View on GitHub ->A machine learning pipeline predicting stroke risk from patient health data. Uses logistic regression with SMOTE oversampling to handle class imbalance, and optimizes the classification threshold for maximum recall. Built with scikit-learn and imbalanced-learn on the Kaggle Stroke Prediction dataset.
View on GitHub ->