rumale

Why we included this project

Rumale brings a Scikit-Learn-style interface to Ruby, which matters if your team builds models inside the Ruby ecosystem and wants to skip a second language just for machine learning. It covers the standard supervised set, from SVM and logistic regression through ridge, lasso, naive Bayes, decision trees, gradient boosting, and random forests, with clustering and dimensionality reduction such as K-Means, DBSCAN, t-SNE, and PCA also available. The fit/predict and transform conventions mirror Python's familiar patterns, so the mental model carries over cleanly. An n_jobs option enables parallel training through the Parallel gem, and pairing it with Numo::Linalg lets the heavy numerical routines run on OpenBLAS. For Ruby shops already doing web or analysis work, that means training models in-process and persisting them with Marshal, without adding an external service.