classical-ml

Enhanced Credit Card Fraud Detection using Graph Neural Networks

[Archived] Classical NLP pipeline (2019-2020) predicting Yelp review quality using TF-IDF, FastText, LDA, and traditional ML. Pre-transformer era techniques preserved as a learning resource.

Reproducible writer-specific offline signature verification using RAW and HOG features with One-Class SVM on the CEDAR and UTSig datasets.

Classical Machine Learning models from skicit-learn and datasets by kaggle.com

Interactive Streamlit app for visualizing decision tree classification boundaries and regression curves with live hyperparameter tuning.

Classical-ML pipeline on the Home Credit Default Risk dataset: EDA, feature engineering (SMOTE, target encoding, binning), tree models (decision tree, random forest, XGBoost) with imbalance-aware evaluation, and K-Means customer segmentation

A lightweight NLP-based ticket routing system that classifies support tickets into TECHNICAL, HR, BILLING, or GENERAL. Built using TF‑IDF and Logistic Regression for fast, interpretable, and cost‑efficient performance. Includes a full ML pipeline: data generation, preprocessing, feature extraction, model training, evaluation, and inference.

Эксперимент с transfer learning на MLP и двух табличных датасетах

Building end-to-end machine learning and deep learning projects to gain practical experience across the complete ML workflow.

Detecting Indonesian political hoaxes using deep learning and machine learning models.

Interpretable classical-ML smishing/scam detector — quantifies how Western-trained models fail on India-specific scams (UPI/KYC/FASTag) and fixes the gap with a small India sample. Fully explainable, offline W-vs-A demo. No deep learning.

used-car prices from specs, R² = 0.957

Unified AI Trust & Safety Workbench - Classical ML for evidence verification (FEVER), misinformation detection (LIAR), toxicity analysis (Jigsaw), and length-aware algorithmic fairness.

Urban crime prediction platform using 7M+ Chicago crime records to forecast zone-wise crime risk across time and location with SHAP explainability. Automated ETL pipelines. XGBoost models with Optuna hyperparameter tuning. Isolation Forest for anomaly-based hotspot detection.