How It Works
The science and engineering behind Seismikon's earthquake predictions - explained clearly, without oversimplification.
Seismikon produces earthquake predictions using a combination of natural time analysis, LSTM neural networks, and the Earthquake Potential Score - all grounded in peer-reviewed research by Dr. Anastasios N. Bikos (2024). The pages below explain each concept in plain language while linking to the primary sources for readers who want to go deeper.
Prediction vs. Early Warning
What is the difference between forecasting an earthquake before it happens and alerting people seconds after one starts? Why does the distinction matter?
Read more →How Seismikon Evaluates Predictions
Every prediction is scored against a published set of tolerances for location, magnitude, depth, and time. No result is ever re-scored after the fact.
Read more →What Is Natural Time?
Instead of using clock time, Natural Time counts seismic events - revealing stress accumulation patterns that calendar-based models miss.
Read more →The Earthquake Potential Score (EPS)
A 0–100 index derived from natural time entropy that quantifies how close a region is to its next significant seismic event.
Read more →How AI Is Used
Seismikon uses LSTM neural networks - a type of deep learning - to learn spatio-temporal dependencies in earthquake sequences.
Read more →What Confidence Means
Confidence scores in Seismikon predictions reflect model output probability - not certainty. Here is what the number represents and what it does not.
Read more →False Positives & False Negatives
Why both types of error matter for earthquake prediction, what the cost of each is, and how Seismikon's evaluation protocol handles them.
Read more →One Success Does Not Prove the System
A single correct prediction could be luck. What statistical standards apply before a claim of skill is justified - and how Seismikon addresses that bar.
Read more →Prospective vs. Retrospective Testing
Fitting a model to past data is fundamentally different from making blind forward predictions. Why prospective performance is the gold standard.
Read more →Research & the Operational Service
The relationship between the peer-reviewed Bikos (2024) methodology and the live prediction service - where they agree, and where they diverge.
Read more →What Seismikon Currently Claims
A precise statement of the operational claims, how each one is measured, and which claims remain aspirational pending more prospective data.
Read more →