Time-Series Parameter Guidance
Recommendations returned by the AIML time-series parameter-discovery route.
Developer simulator
eazePredict API Simulator
Test time-series parameter discovery, train, and predict payloads before integration. Use sample data, build a request visually, or paste the documented JSON payload.
Test the same payload your backend can send later. Simulator calls are rate-limited and meant for evaluation.
Sample Data
Run a valid prediction payload immediately
Start from a complete prediction request body for the selected model.
Linear Regression sample
Forecast or predict output from documented input signals.
Guided Payload Builder
Build the documented AIML JSON without hand-writing it
These fields map directly to the selected AIML request body: tenant key, model artifact key, output label, cadence, start date, and feature arrays.
CSV import
Import CSV to fill target and input signal values.
Time-series parameter discovery provides a recommendation from the supplied history.
Compare candidate models and validation metrics before production use.
Classification uses bucketed event and metric signals, not row arrays.
Configure the existing classification schema directly. Labels are derived from a future lookahead window by the AIML engine.
Input feature signals
Each signal maps to one schema-validated
featureSignals object.
Target class source
The API generates a YES/NO target from this future signal window.
Predict needs a model ID produced by an earlier successful classification training run. Predict does not send or verify modelFamily, so use the ID returned by the selected training run and avoid reusing the same taskKey plus targetLabel across classifier families. The shape-only sample does not point to a production artifact.
Raw JSON
Edit the exact payload your integration can send
Paste a documented AIML discovery, train, or predict payload. The simulator posts this JSON body directly to the public simulator endpoint for the selected operation and model.
Make sure the JSON matches the selected Predict API payload from the documentation.
Value builder
Build values
Press Enter after each value. We'll build the comma-separated list for you.
CSV import
Paste CSV values
Paste a header row plus numeric values. The CSV stays in this browser and is never uploaded as a file.
Example format
featureOutput.output,historical_demand,average_price_cents
122,120,2450
130,128,2400
137,135,2350
External input signals
Result
Simulator response
Successful responses can be reviewed as a status summary, guidance, chart, table, trust/warning details, and raw JSON.
Ready
Run the simulator to see API output.
Trust Decision Preview
Human-readable trust summary for the latest simulator response.
Developer trust details
Raw trust fields returned by the simulator response.
JSON
Developer trust details
Raw trust fields returned by the simulator response.
trust_summary
{}
trust_summary.modelDisagreement
null
trust_summary.reasonCodes
[]Classification result
Classification-specific API fields are shown here without converting them into a forecast chart.
Prediction trend
Model output across the returned time points.
Training residuals
Holdout-only errors returned in the training response.
Result table
Inspect the exact values behind the charts.
Full simulator response JSON
{}Developer helper
Integration details for the generated request
Use this public simulator route for evaluation. Do not send AIML internal base URLs, service tokens, or secret headers from browser apps.
Operation
Predict/api/v1/predict/simulator/linearPOST
Generated JSON
{}
cURL command
Time-series discovery
ARIMA parameter guidance
Time-series parameter discovery recommends p, d, q, or a candidate ARIMA order from the supplied history. Treat suggestions as diagnostics to compare against validation metrics, not as guaranteed production settings.
| Term | Meaning | Simulator behavior |
|---|---|---|
p |
Autoregressive order: how many prior observations the AR portion uses. | find-p forwards the documented helper payload and renders returned PACF values only when the backend includes them. |
d |
Differencing order: how many times the series should be differenced to become sufficiently stationary. | find-d uses the confirmed ADF-based helper response and renders returned p-values. |
q |
Moving-average order: how many prior forecast errors the MA portion uses. | find-q forwards the documented helper payload and renders returned ACF values only when the backend includes them. |
(p,d,q) |
ARIMA order combining autoregressive, differencing, and moving-average terms. | find-order renders returned heuristic and AIC grid-search details without inventing missing metrics. |
Discovery payloads require timeSeriesInput as a flat numeric array with at least 8 observations; 20+ observations are recommended. Date and frequency fields are not required by the confirmed ARIMA helper schema. CSV import accepts value rows, or timestamp,value rows with timestamps ignored. SARIMA may also require seasonal order (P,D,Q,m); no seasonal-order discovery route is registered, so seasonal options are not shown.