A modern forecasting dashboard for analyzing daily electricity demand patterns, comparing ARIMA and ETS models, examining STL decomposition, and exploring short-term future demand projections.
This dashboard is designed to help users move from raw demand history to interpretable time-series components and model-based forecasting. It combines interactive charts with short explanations so the results feel easier to understand.
The original data is hourly, but daily aggregation reduces noise and makes medium-term patterns easier to interpret. This is especially useful when comparing forecasting models and decomposition results.
Look for recurring cycles, structural changes, spikes, and changes in recent variability. Those patterns often influence how well ARIMA or ETS performs.
STL stands for Seasonal and Trend decomposition using Loess. It separates a time series into trend, seasonal, and remainder components to make the structure easier to interpret.
A forecasting model often performs better when the structure of the series is understood first. Decomposition helps identify whether the data is dominated by trend, stable seasonality, or irregular noise.
A smaller and less structured remainder suggests the main patterns were captured well. Large erratic remainder values may indicate difficult-to-predict movement.
ARIMA models autocorrelation and differencing structure in a time series. It is often useful when the data shows strong temporal dependence and can be made approximately stationary.
ETS stands for Error, Trend, and Seasonality. It is often effective when the time series has stable level, trend, and seasonal patterns that can be smoothed over time.
The best model is not just the one that looks visually smoother. It should also minimize forecast error metrics and follow the actual series closely during the evaluation period.
These values are model-based projections, not guaranteed outcomes. They are most useful for planning, comparison, and understanding short-term expected demand behavior.
Forecast accuracy depends on how stable the recent pattern is, whether structural changes occurred, and whether the selected model captures the trend and seasonality well.
A time series is a sequence of observations collected over time. In this project, electricity demand is observed repeatedly and modeled as a structured temporal process.
Seasonality refers to patterns that repeat at regular intervals. Electricity demand often shows repeated cycles due to behavioral, weather, and operational factors.
RMSE stands for root mean squared error. Lower RMSE means the forecast is, on average, closer to the actual observed values, with larger errors penalized more heavily.
Different models capture different structures. Comparing ARIMA and ETS helps determine which approach better fits the observed demand dynamics in this dataset.
This dashboard is not only a forecasting tool but also a learning interface: it connects raw demand data, decomposition, forecast comparison, and future projections in one interpretable workflow.