auc-roc

Loan Default Prediction using PySpark, with jobs scheduled by Apache Airflow and Integration with Spark using Apache Livy

Chest X-ray screening research with explicit distribution shift evaluation, per-disease binary models and validated Grad-CAM localization.

A machine learning solution for churn prediction using CatBoost, achieving a 0.8464 AUC-ROC through feature engineering and hyperparameter optimization.

A machine learning web application and classification pipeline detecting real vs. fake news articles. Built with a Streamlit interface and an optimized Python NLP pipeline using TF-IDF vectorization and Logistic Regression. Achieves a 98.57% stratified test accuracy, 98.63% F1-score, and a 0.999 AUC-ROC.

A classification project using the Heart Disease UCI dataset

The final project finished 05-27-2026 as part of the TripleTen Data Science program using real-world data and mimicking real-world project requirements. Task was to train a machine learning model to predict which Interconnect customers are soon to terminate their contracts to offer special deals and reduce churn rate. Best AUC-ROC was 0.885.

End-to-end credit risk ML pipeline — loan default prediction, FICO-style credit scorecard & threshold optimization on 117K+ applicants. Python · MySQL · Power BI. AUC-ROC 0.8428 · $6.6M projected net benefit.

Predicts telecom customer churn with machine learning and an interactive Streamlit app. Features include single/batch predictions, dashboards, and actionable insights for improved retention.

Parkinson's Disease Prediction using Machine Learning with a web-based interface built with React, Flask, and Node.js for real-time predictions and user history tracking.

Binary classification pipeline predicting next-day stock direction using technical indicators

Semantic coherence verification for distributed AI inference. Omega metric validated AUC 0.9539 on Wikipedia ES.

Prediction Model, Bias and Uncertainty

Credit card fraud detection with anomaly detection and deep learning.

Эта работа - проект для сессии старшего курса.Но в ней сосредоточены основные требования работадателей

A predictive healthcare model to detect early signs of diabetes or heart disease using medical datasets. The project features EDA, feature selection, model training (SVM, Gradient Boosting, Neural Networks), and insights for clinical decision-making.

A project finished 01-22-2026 as part of the TripleTen Data Science program using real-world data and mimicking real-world project requirements. Task was to train an ML model to predict when BetaBank customers would terminate their contract in order to offer special deals and reduce churn rate. F1 score of 0.640 and AUC-ROC of 0.855. Req. F1>0.59.

Benchmarking 10+ ML algorithms to predict diabetes risk with 82% AUC featuring Snowflake and dbt

End-to-end customer churn prediction using Logistic Regression and Random Forest — Telco dataset