←Back
250 – Topics in Statistical Learning
Predicting Disease Incidence with Natural Cubic Splines
Sponsor: Section on Statistical Learning and Data Science
Keywords: Cross validation, Smoothing splines, Time series, Dengue fever
Yew-Meng Koh
Hope College
Noah Kochanski
Hope College
High-degree polynomials provide great flexibility and potentially perfect fit of historical time series data. Such flexibility, however, often leads to overfitting and results in models with poor predictive performance. Splines are a low-degree polynomial smoothing method which reduces these overfitting effects. We use a cross-validation method for time series in order to compare the performance of various models which utilize smoothing splines with regard to their forecast accuracy of Singaporean dengue fever counts.