Tekmerion
Ready for pilotsDonated to NKUA

Explainable time-series forecasting software

Chresmos

Explainable Predictions

Forecasts that explain themselves.

Explainable forecasting for series whose cycles drift, built on Locally-Adaptive Karhunen–Loève forecasting (LA-KLF), with its reasoning shown at every step.

Evidence

144golden arrays lock code to paper
180backend tests
656dashboard tests
AAWCAG 2.1 accessibility

How it works

The numerical core implements the accompanying theory paper and is locked to it by a golden test contract of 144 arrays, so what the paper says and what the software does cannot drift apart. On top sit theory-driven tuning, calibrated uncertainty, statistically tested backtests, real-time vintage evaluation and regime alerts, with an audit trail that can be replayed bit for bit. Saved models follow an approval workflow, and reports can be scheduled. Every stored run is sealed in the Bitcoin blockchain with OpenTimestamps, so anyone can prove when a forecast was made and that it has not changed since.

Inside

  • LA-KLF core
  • Calibrated uncertainty
  • Backtests with tests
  • Regime alerts
  • Replayable audit trail
  • Bitcoin-sealed runs
  • API

← Back to the suite

See it on your own data.

We run pilots with banks, funds, energy and shipping firms, public bodies and universities.