ARIMA_SAX_Hybrid_Forecaster

CURRENT GOVERNANCE ENTRYPOINT

CURRENT AUTHORITY: RESEARCH_CHARTER.md · RESEARCH_RULES.md · RESEARCH_STATE.md · results/EVIDENCE_LEDGER.md

CURRENT FAMILY: MALE5-ML CURRENT GATE: MALE5-ML-EXECUTION (BLOCKED) SCIENTIFIC STATUS: UNKNOWN — computation blocked; no valid MALE5 real-data result NEXT GATE: MALE5-ML-EXECUTION-RETRY-ON-LARGER-ALLOCATION

HISTORICAL FOUNDATION — SUPERSEDED AS AUTHORITATIVE ENTRYPOINT. The text below documents the original ARIMA + SAX research framework and is preserved intact for the historical record. The current authority is the governance block above plus RESEARCH_CHARTER.md, RESEARCH_RULES.md, RESEARCH_STATE.md, and results/EVIDENCE_LEDGER.md.

ARIMA_SAX_Hybrid_Forecaster

This repository is a research framework for testing ARIMA and SAX hybrid forecasting. It does not claim profitable trading performance and must not be interpreted as a validated trading strategy unless explicitly demonstrated by reproducible out-of-sample evidence.

Owner: chiki2bum2 Platform: https://forge.mql5.io/chiki2bum2

---

Purpose

This is a research project first, not a trading-product project. It builds a reproducible research harness to answer one central scientific question:

Does a hybrid ARIMA + SAX forecasting architecture provide statistically and economically meaningful incremental predictive information compared with appropriate baselines and with each component individually?

The two conceptual references are:

Method	Reference
ARIMA	https://www.mql5.com/en/articles/18247
SAX	https://www.mql5.com/en/code/75543

These are methodology and implementation references only. They are not evidence that either method (or their combination) has trading edge.

Research posture

• No assumption that either model has predictive edge.
• No assumption that combining two models improves performance.
• No live trading / execution in this harness.
• No lookahead is permitted. A failing no-lookahead test is a hard stop.
• No Edge and insufficient analog conditions are legitimate outputs, not implementation failures.

Project structure

docs/       Design and research notes
configs/    Freezed experiment / model configurations
src/        Source code (python statistical/runtime engine)
tests/      Unit, integration and no-lookahead research tests
experiments/Experiment run definitions
results/    Reproducible output artifacts (git-ignored)
scripts/    Runner / report entry points

All user-visible research claims must separate OBSERVED vs INFERRED vs UNKNOWN vs DECISION.

See RESEARCH_PROTOCOL.md for the governance rules.

Status

This repository is in first-phase research bootstrap. See CHANGELOG.md.