Auto-Mutual Information Forecastability — M4 Competition Analysis (v1.0.0)
Abstract
Analysis code and result artefacts for “Horizon-resolved Forecastability of Time Series via Auto Mutual Information” (Scientific Reports). Version 3 retains every file of version 2 unchanged (the canonical run run_20260801_203622_73477, which produced all forecasts) and adds sr_revision_2026-10.zip, an extended analysis campaign run on the frozen outputs of that run without regenerating any forecast: AMI on differenced and STL-remainder series; the Spearman autocorrelation as a second comparator with paired bootstrap intervals; partial correlations on five series-level controls; RMSSE, one-step-scaled MASE, HMSE and QLIKE sensitivities; a pre-specified benchmark-skill analysis against seasonal naive, naive-1 and drift references; two further N-BEATS fits; two-origin stability; and synthetic estimator checks under serial dependence (AR(1) grid with Theiler windows, AR(2), ARCH(1)). The archive includes scripts, inputs, every output table, a manifest with per-file SHA-256 hashes, seeds and software versions, the dated analysis amendment, and the figure script. README.md maps every table and figure in the manuscript and its Supplementary Information to the file and script that produces it.
Fields of Study & Subjects
Time Series Analysis and Forecasting
Forecasting Techniques and Applications
Complex Systems and Time Series Analysis
Publication details
- Journal / source
- Zenodo (CERN European Organization for Nuclear Research)
- Publication date
- 1 March 2026
- Language
- EN