hierarchical-forecasting

Scalable and user friendly neural 🧠 forecasting algorithms.

PyAF is an Open Source Python library for Automatic Time Series Forecasting built on top of popular pydata modules.

Zero-config time series forecasting for Python. 30+ models, auto selection, Forecast DNA, Rust turbo — one line of code.

Conformal forecasting for hierarchical time series: point forecasts, reconciliation, and calibrated prediction bands.

Functional PoC for temporal hierarchical demand forecasting using exogenous variables, reproducible ML pipelines, and lightweight app/API layers.

A commercial AI-driven platform for real-time and end-of-day forecasting of all Tehran Stock Exchange symbols. Built in collaboration with industry partners and academic advisors, it integrates automated data-ingestion pipelines, deep-learning LSTM models, smart feature extraction (technical data & news data), and rolling 40-minute predictions.

Leakage-safe hierarchical forecasting with bottom-up and MinT reconciliation

Production-grade retail demand forecasting platform with hierarchical MinT reconciliation, Split Conformal Prediction, Tweedie LightGBM, PSI drift telemetry, FastAPI & Streamlit.

Implementing 17 Machine Learning Models in a Hierarchical Data Architecture and Evaluating Their Performance

Demand forecasting for the U.S. power grid and the household meter: every balancing authority in the lower 48 forecast 48 hours ahead with calibrated quantiles, graded against the operator's own day ahead forecast, reconciled across the hierarchy, watched by an anomaly detector, served by a live forecast log.

Practical primitives for time-series forecasting pipelines: conformal intervals, model comparison, hierarchical reconciliation, tuning, validation, drift detection.

Prevision hierarchique de 32 series trimestrielles - approche ascendante, propagation de l incertitude, WinRATS

A demand-planning system that produces coherent forecasts across every level of a retail hierarchy — and translates them into the revenue and forecast-error cost a planner actually decides on.

Hyundai Mobis PIO accessory forecasting | Sanitized UCLA MEng capstone case study.

End-to-end demand planning — 6-model routing ensemble (MAPE 10.3%), capacity planning, demand sensing, S&OP simulation | MinTrace hierarchy, walk-forward CV, conformal prediction | Enterprise: K8s + Helm + Terraform + MLflow + Prometheus/Grafana | 192 tests

Forecasting Pix Across Brazil: global models and coherent geographic reconciliation of municipal instant-payment data (R, targets, DuckDB) — reproducible research compendium

Hierarchical demand forecasting · M5 Walmart · 3049 series · 6 levels · MinT reconciliation · LightGBM · DuckDB · conformal prediction · FastAPI · 135 tests

Demand forecasts whose uncertainty ranges hold when conditions shift, turned into a staffing number at a chosen service level: hierarchical probabilistic forecasting, adaptive conformal intervals, MinT reconciliation and a newsvendor decision layer.

ML a serviço da inferência — Double/Debiased ML para efeitos heterogêneos e forecasting hierárquico com reconciliação MinT no dataset M5.