Statistical Learning

Definisi

Statistical learning adalah pendekatan machine learning berbasis statistik (probabilitas, likelihood, Bayesian inference) yang dominan 1990-2010.

Algoritma

• Linear/Logistic Regression (1800-an, masih populer)
• Naive Bayes (1950-an)
• Decision Trees (ID3, CART — 1980-an)
• k-Nearest Neighbors
• Hidden Markov Models (HMM)
• Conditional Random Fields (CRF)
• Support Vector Machines (SVM) — Vapnik, 1995
• Bayesian Networks — Judea Pearl
• Gaussian Mixture Models (GMM)
• Random Forests (2001)
• Gradient Boosting (XGBoost, LightGBM)

Era

• Dominan: 1990-2010
• Tergeser: 2012+ (AlexNet, deep learning)
• Hybrid: 2020+ — neural network + statistical
• Sekarang: tetap terbaik untuk tabular data

Aplikasi Modern

• XGBoost, LightGBM, CatBoost — dominasi Kaggle tabular
• SVM — masih dipakai untuk small data
• Bayesian methods — uncertainty quantification
• Time series — ARIMA, Prophet
• A/B testing — statistical inference