Materials on AI, Python, Git, SQL, SAP/ERP, statistics and time series: explanations, examples, practical tips, sources and applicability limits.
Every word must appear in the article. Use quotes for an exact phrase.
Choose a regression metric for non-negative data by first deciding the prediction goal, then picking a consistent loss such as MAE, RMSE, or pinball loss, and finally comparing against a constant baseline with a D2-style score.
A practical guide to selecting a decision threshold by weighting false‑positive and false‑negative costs, using a held‑out validation set and scikit‑learn confusion matrices.
Prediction intervals aim to contain future individual observations, while confidence intervals target a mean or other parameter, and the two are not interchangeable. Empirical coverage is the held-out fraction inside the interval and width is the distance between quantile limits; a constructed five-value example gives 80 percent coverage and mean width 1.8.
This article explains how to choose effective chunk boundaries and overlap in retrieval-augmented generation (RAG) systems, comparing fixed-size and document-aware strategies. Using a hypothetical support note example, it demonstrates the risks of splitting across semantic units and shows how metadata and overlap can preserve context. The guidance is based on Microsoft Azure documentation for chunking in vector search and RAG workflows.
Build a previous-season reference, evaluate it on later observations and distinguish calendar alignment from leakage and missing data.
Use chronological splits and fit preprocessing only on past data.