Official Code for "Confidence Matters: Enhancing Medical Image Classification Through Uncertainty-Driven Contrastive Self-distillation" accepted at MICCAI2024

Credit card fraud detection system using logistic regression, EDA, class imbalance analysis, SMOTE, and a Gradio demo.

Customer Churn Prediction using Machine Learning (Imbalanced Classification) Customer Churn Prediction using Machine Learning (Imbalanced Classification)

Fundamentals of Machine Learning Assignment Repository

Predicting company bankruptcy using various machine learning models. The dataset is sourced from Kaggle: Company Bankruptcy Prediction.

Predicting which people would be likely to convert from free users to premium subscribers in the next 6 month period, if they are targeted by our promotional campaign.

(WIP): 'Aporia' in Greek means 'inconsistent'. A Python library that detects and fixes dataset issues using both rule-based methods and ML models. It evaluates dataset quality across multiple metrics, including missing values, duplicates, outliers, class imbalance, and label consistency. It also suggests fixes based on the metric scores.

End-to-end credit risk scoring system using XGBoost, SHAP, and threshold tuning to predict loan defaults and automate lending decisions.

Developing a machine learning model to predict customer churn as it is essential for proactively retaining valuable customers.

Developed an ensemble ML classification model to predict U.S. visa case outcomes (Certified vs Denied) using applicant and employer attributes. Performed EDA, sampling, and model tuning (Random Forest, Gradient Boosting, XGBoost) to improve decision efficiency and identify key policy drivers like education, experience, and wage trends.

Machine learning-based login anomaly detection system that identifies suspicious user behavior using time and IP pattern analysis, simulating real-world SOC monitoring scenarios.

Supervised Learning project from TripleTen

Comparing SMOTE, Class weight and Hybrid techniques for stroke prediction on an imbalanced medical dataset using Logistic Regression and Random Forest.

This project focuses on detecting fraudulent credit card transactions using Machine Learning and Data Analytics. It applies advanced techniques such as EDA (Exploratory Data Analysis), feature engineering, and imbalance handling (SMOTE, undersampling) to improve fraud detection accuracy.

Solution write-up for a Kaggle text-classification challenge: severely imbalanced, obfuscated comment data, classical ML only (no deep learning). Methodology, engineered features, and a stacked LightGBM/sklearn ensemble

End-to-end machine learning workflow on the Combined Cycle Power Plant dataset: data cleaning, EDA, outlier removal, feature engineering, class balancing, and model evaluation for regression and classification. Includes code, visualizations and best practices in a single Jupyter notebook.

This is a production-ready, end-to-end system developed to detect and classify racist tweets using advanced Natural Language Processing (NLP) techniques. Built on top of BERTweet (vinai/bertweet-base) and fine-tuned with a robust, k-fold cross-validation training pipeline, powered by streamlit UI!

A dual-part finance and retail analytics project covering credit default prediction for companies using machine learning (Logistic Regression & Random Forest) and market risk analysis of a five-stock Indian equity portfolio using historical price and return data.

Analysis of bank marketing campaigns using machine learning to predict term deposit subscriptions, optimizing campaign strategies through comparative evaluation of classification models.