Articles by tag: time-series

Diversity of errors beats quality in ML ensembles

Learn why weak models in ensembles give MAE 8.5 vs 9.5 strong ones. Experiment on time series, selection recommendations. For ML specialists.

Time Series Anomaly Detection ML

Study Isolation Forest, CatBoost and autoencoder for detecting threats in storage logs. Code, metrics, comparison. Apply in infosec without labeling.

Sales Forecast: from Excel to ML in half a year

How to build a reliable sales forecast for business: backtest, TSMixer, ensemble. Error reduction to 3.2%, increase in top management's trust. For DS specialists.