Time & Forecasting

• Naive ForecastPredicting 'same as last time' — the baseline every model must beat.
• Run RateExtending today's pace to the full period — and why it usually overstates.
• TrendThe long-run direction underneath the noise.

• Forecast IntervalA range around a forecast saying how uncertain it is.
• ForecastingPredicting future values from past patterns, with honest uncertainty.
• Lag FeaturesUsing past values of a series as inputs to predict its future.
• Mean Absolute Percentage ErrorForecast error as a percentage, and where it misleads.
• Moving AverageSmoothing a series by averaging each point with its neighbours.
• Root Mean Squared ErrorForecast error that punishes big misses more than small ones.
• SeasonalityPatterns that repeat on a calendar: weekly, monthly, yearly.
• Time SeriesData ordered in time, and why the order changes everything.
• Trend, Seasonality and NoiseSplitting a series into direction, repeating pattern and leftover.

• ARIMAA classical forecasting model combining autoregression, differencing and moving averages.
• AutocorrelationHow much a series resembles its own past, at each lag.
• Exponential SmoothingForecasting by weighting recent history more than old history.
• Forecast AccuracyJudging forecasts on held-out data, against a baseline.
• Seasonal AdjustmentRemoving the calendar pattern so the underlying movement is visible.
• StationarityWhen a series' statistical behaviour doesn't drift over time.

Nothing here yet.

Nothing here yet.