Geneva Forecasting API: An Expert System for Time-Series Forecasts

Source: Dev.to

Overview

Time‑series forecasting has been around for decades, and the Geneva Forecasting Engine has been quietly powering forecasts for over 25 years. Although few developers know the name, the engine has been the core forecasting component inside many Oracle products (e.g., Oracle OLAP, where DML statements such as FCOPEN, FCSET, and FCEXEC rely on Geneva for demand planning and statistical forecasting).

Access

RoadMap Technologies now offers the same proven engine to all developers via a modern, lightweight API.

• Portal:
• Free tier: 1,000 forecasts per month – no credit card required.

Getting Started

# geneva_forecast example
from geneva_forecast import GenevaClient

client = GenevaClient(
    api_url="https://api.roadmap-tech.com",
    api_key="gva_xxxxxx",
)

result = client.forecast(
    data=[
        118, 153, 137, 186, 129, 164,
        152, 207, 156, 178, 168, 225,
        167, 205, 176, 248, 185, 213,
        199, 272, 196, 244, 218, 298
    ],
    horizon=12,
    wave_periods=[4],
    confidence_level=0.95,
)

print(result.forecast)

How It Works

Geneva is a rule‑driven engine that automatically evaluates ten statistical forecasting methods, each with best‑fit parameter tuning and seasonality detection. It selects the strongest performer for a given series and returns the forecast—typically in under 200 ms.

Expert System

When no specific method is supplied, Geneva’s built‑in Expert System:

1. Tests all ten methods against the historical data.
2. Chooses the optimal model based on performance metrics.
3. Returns the forecast and full visibility into the decision process, including:
   • The selected method_name
   • MAPE, MAD, and RMSE metrics
   • Whether smoothing or seasonality handling was applied

Features

• Automatic model selection via expert system rules.
• Performance metrics (MAPE, MAD, RMSE) included in the response.
• Confidence intervals (configurable confidence level).
• Fast response time (≈ 200 ms per request).
• Lightweight API – no need to manage a custom modeling pipeline.

Use Cases

• Dashboard visualizations
• Inventory and demand planning
• Internal tooling that requires quick, reliable forecasts

Further Information

For questions or more details, visit the portal:

Stay tuned for upcoming posts covering advanced capabilities such as probabilistic forecasting and agentic AI extensions.