Creating Forecasting model
From a workbook, click Actions > ML Dashboard. The ML Dashboard page is displayed. From the Create Model section, click Forecasting and specify the following:
| Feature | ARIMA | Prophet |
|---|---|---|
| Model Type | Classical statistical time series model | Additive model developed by Meta |
| Best Suited For | Data with autocorrelation and consistent structure over time | Time series with clear trend, seasonality, and holiday effects |
| Trend Handling | Captures linear trends via differencing and autoregression | Automatically models piecewise linear or logistic trends |
| Seasonality | Can model seasonality with SARIMA extension | Automatically models multiple seasonalities (daily, weekly, yearly) |
| Interpretability | Transparent model with clearly defined coefficients | Decomposes time series into trend, seasonality, and holiday components |
| Parameter Tuning | Controlled via intuitive parameters (p, d, q) | Minimal manual tuning needed; most settings are automated |
| Data Requirements | Effective on stationary, structured time series | Handles missing values and outliers well |
| Domain Adaptability | Widely used in economics, finance, and control systems | Popular for business, social media, retail, and web traffic forecasting |
| Library Support | Supported by statsmodels and pdarima in Python | Available via the prophet library in Python and R |
| Forecasting Output | Point forecasts with confidence intervals | Forecasts with uncertainty intervals and changepoint detection |
| Deployment Readiness | Lightweight and fast to train | Easy to integrate into production pipelines |
| External Regressors | Supports in extended versions (e.g., ARIMAX) | Direct support for custom regressors |
| Community & Resources | Extensive academic and practical literature | Strong documentation and open-source community support |
Click Create and commence the model training process.