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.

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

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

Эксперимент с transfer learning на MLP и двух табличных датасетах

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.

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.

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

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.

used-car prices from specs, R² = 0.957

Classical ML on tabular data: EDA, regression, classification, clustering and a PyTorch MLP

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