Model guide

AutoReg (Autoregression)

An autoregression model that uses lagged historical values as direct predictors for the next value in the series. Useful for short-memory sales, demand, usage, revenue, capacity, or service-health signals where recent past behavior strongly influences the near future. Simple, fast, and interpretable, making it a good lightweight baseline before moving to more expressive seasonal or feature-based approaches.

Model ID: timeseries/autoregression Time-series forecasting Trust-aware output guidance

Where to start

Use this guide to connect the route, request, and interpretation together

This page is focused on the practical path for AutoReg (Autoregression): which prediction routes are available, what the request body should communicate, what the response shape means, and how to test the route without exposing service credentials to browser clients.

Good fit

Use this model when historical signal behavior is meaningful enough to forecast the next time window.

Integration guidance

Server-side applications can call eazePredict routes with service credentials. Keep credentials out of browser clients and use your application layer to expose application-facing workflows.

Available endpoints

Verified route paths for AutoReg (Autoregression)

These are the verified paths for this model family. Use the base URL /trace/engine/v1 before the route path in API mode.

POST /timeseries/autoregression

Runs the main model or analysis workflow for this guide.

Relative URL: /trace/engine/v1/timeseries/autoregression

POST /timeseries/autoregression/find_p

Runs a diagnostic helper for choosing time-series order parameters before training or forecasting.

Relative URL: /trace/engine/v1/timeseries/autoregression/find_p

POST /timeseries/autoregression/predict

Loads a saved model artifact and returns prediction rows for the provided feature or time-series context.

Relative URL: /trace/engine/v1/timeseries/autoregression/predict

Example request

Request shape for AutoReg (Autoregression)

This example shows the core JSON shape expected by this guide. In deployed service-to-service calls, include customerKey for protected train, predict, and model-forecast routes, and keep service credentials outside browser clients.

  • timeSeriesFeatureModel identifies the saved time-series model family.
  • timeSeriesInput carries the historical signal values to fit or forecast.
  • forecast, timeSeriesType, and timeSeriesFrequency define the forecast horizon and cadence.
JSON request body 
{
  "customerKey": "eazepredict-demo",
  "forecast": 3,
  "startDate": "Wed, 01 Jan 2025 00:00:00 GMT",
  "timeSeriesFeatureModel": "productUsageForecast",
  "timeSeriesFrequency": 1,
  "timeSeriesInput": {
    "label": "active_users",
    "output": [
      980,
      1020,
      1015,
      1088,
      1130,
      1195,
      1210,
      1265,
      1310,
      1360
    ]
  },
  "timeSeriesType": "d",
  "train": 80
}

Example response

Successful response shape

Responses are JSON and vary by route. Training responses can include artifact metadata and trust summaries. Prediction and forecasting responses return rows keyed by timestamp, signal labels, model output, and any trust-aware warning fields produced by eazePredict.

  • result usually contains model fit, forecast, or analysis rows.
  • prediction appears on saved-model prediction routes.
  • error appears when validation, artifacts, or processing fail.
JSON response body 
{
  "message": "Models trained and saved successfully",
  "result": [
    {
      "active_users": 980,
      "timestamp": "Wed, 01 Jan 2025 10:00:00 GMT",
      "type": "actual"
    },
    {
      "active_users": 1020,
      "timestamp": "Wed, 01 Jan 2025 12:00:00 GMT",
      "type": "actual"
    },
    {
      "active_users": 1015,
      "timestamp": "Wed, 01 Jan 2025 14:00:00 GMT",
      "type": "actual"
    },
    {
      "active_users": 1088,
      "timestamp": "Wed, 01 Jan 2025 16:00:00 GMT",
      "type": "actual"
    },
    {
      "active_users": 1130,
      "timestamp": "Wed, 01 Jan 2025 18:00:00 GMT",
      "type": "actual"
    },
    {
      "active_users": 1195,
      "timestamp": "Wed, 01 Jan 2025 20:00:00 GMT",
      "type": "actual"
    }
  ]
}

Example cURL

Test the route from a server-side app or terminal

This example uses the verified relative path selected for this model guide. Replace the token, deployment host, and request values with deployment-specific values.

cURL example 
curl -X POST "/trace/engine/v1/timeseries/autoregression" \
  -H "Authorization: Basic <aiml-service-token>" \
  -H "Content-Type: application/json" \
  -d '{"customerKey": "eazepredict-demo", "forecast": 3, "startDate": "Wed, 01 Jan 2025 00:00:00 GMT", "timeSeriesFeatureModel": "productUsageForecast", "timeSeriesFrequency": 1, "timeSeriesInput": {"label": "active_users", "output": [980, 1020, 1015, 1088, 1130, 1195, 1210, 1265, 1310, 1360]}, "timeSeriesType": "d", "train": 80}'

Trust behavior

How to interpret trust signals from this model

Numeric output alone should not drive operational action. Use trust-aware fields and context checks to decide whether to escalate, review, compare, or suppress.

Model disagreement

If multiple models produce materially different outputs, treat the result as uncertain and review the operating context.

Unseen regime

If the request falls far outside historical feature ranges or known operating combinations, the prediction may be unreliable.

Suppression behavior

Responses may return null plus a trust_warning instead of a misleading number when unsafe conditions are detected.

Back to model guides API documentation