Predictive Modeling

Predictive Time Series

Question: How is the selected metric likely to move over the next few months? Data: monthly historical series with trend and seasonality patterns. Method: SARIMA or Holt-Winters forecasting with back-fit error checks, confidence intervals, residual diagnostics, and seasonal decomposition.

Dataset

FMCG 2022-2024

Daily aggregated sales and quantity

1. Analysis Name

Predictive Time Series

This page estimates how the selected business metric is likely to evolve over the forecast horizon, using historical trend and seasonal behavior.

2. Problem Context

What problem this page answers

The problem is to project near-term outcomes with enough transparency to support planning. The page makes the tradeoff explicit by showing actual history first, then the fitted forecast, then the uncertainty and residual diagnostics behind it.

3. Observed Data

Observed history and descriptive statistics

The first view isolates the actual historical series before any forecast is applied. The table summarizes the central tendency and spread of the observed metric.

Descriptive Statistics

4. Workflow

How the forecast answer is built

The workflow moves from projected path to error inspection and seasonal decomposition so the forecast can be evaluated, not just displayed.

01

Fit the model

Estimate the future path using SARIMA or Holt-Winters.

02

Check residuals

Inspect the remaining error to see whether the model is missing structure.

03

Explain seasonality

Separate long-term trend, repeating seasonality, and residual noise.

Forecast Summary Table

Trend Signal

-

Model Performance

MAE: -
MAPE: -
AIC: -
Feature Holdout R2: -
Feature Holdout RMSE: -
Feature Holdout MAPE: -

Top Feature Drivers

5. Conclusion

Recommended forecast answer

Why this is the best answer