Time Series Forecasting

Course outline and learning expectations

This is a tutor-led course. The outline shows what your tutor will cover; teaching materials and examinations are provided directly to enrolled students.

What you will learn

• Explain how time-dependent data differs from cross-sectional data
• Identify and visualise trend, seasonality, cycles and noise
• Prepare dated datasets and address missing, irregular or anomalous observations
• Create baseline forecasts with moving averages and exponential smoothing
• Build and interpret ARIMA models
• Build and interpret Prophet forecasts
• Evaluate forecasts using time-based validation and accuracy metrics
• Communicate forecast uncertainty and limitations
• Complete a forecast using a local African time series

Modules

1. 1

   Understanding time series data

   Recognise ordered data and why time dependence matters in rainfall, prices and sales

   Time indexFrequency and granularityAutocorrelation intuitionTime-series visualisation
2. 2

   Trend, seasonality and noise

   Break a series into its main patterns and recognise structural changes

   TrendSeasonalityCyclesNoiseOutliers and change points
3. 3

   Moving averages and smoothing

   Create simple baselines and smooth short-term variation

   Naive forecastsRolling averagesWeighted moving averagesSimple exponential smoothingHolt and Holt-Winters methods
4. 4

   ARIMA models explained simply

   Build and interpret classical statistical forecasts

   StationarityDifferencingAutocorrelation and partial autocorrelationARIMA parametersResidual diagnostics
5. 5

   Prophet for quick forecasting

   Use Prophet, formerly Facebook Prophet, for interpretable forecasts with trend and seasonality

   Prophet data formatTrend and change pointsSeasonalityHoliday and event effectsForecast intervals
6. 6

   Evaluating forecast accuracy

   Compare models without leaking future information into training

   Train-validation splitsWalk-forward validationMAERMSEMAPE limitationsBaseline comparison
7. 7

   Forecasting with limited or messy African data

   Prepare sparse, missing, irregular and changing datasets for responsible forecasting

   Missing observationsIrregular datesOutliersShort historiesExternal driversDocumenting assumptions
8. 8

   Capstone local forecast

   Forecast the next three months of coffee farm-gate prices in Gomba when suitable historical data is available, or use another documented Uganda or African time series

   Problem definitionData sourcing and cleaningModel selectionAccuracy testingThree-month forecastVisualisationLimitations and recommendations