EP20: Prof Rob Hyndman: Forecasting COVID, time-series, and why causality doesnt matter as much as you think.

This podcast episode covers a wide range of topics related to forecasting, including the challenges of COVID-19 forecasting, the importance of measuring forecasting accuracy, best practices during a pandemic, choosing the right forecasting horizon, uncertainty and causality in forecasting, building trust in forecasts, and the opportunities and threats of AI in society. Takeaways • Practical problem-solving and interesting mathematics are essential for good forecasting. • Insider trading is unethical and can have legal consequences. • Ensembles of models generally provide better forecasts than individual models. • Probability scoring is used to measure the accuracy of probabilistic forecasts. • Reproducibility and version control are important in data science. • Consider the purpose of the forecast and decisions based on it when choosing the forecasting horizon. • Forecasts should be evaluated and updated over time to account for new information. • Causality can help identify better variables for a model, but it is not always necessary for accurate forecasts. • Explainability is important for building trust in forecasting models. • AI has the potential to improve our lives but also raises ethical concerns.