Enhancing Time Series Analysis with Machine Learning, High-Frequency Data, and Applications in Medicine and Biology

This concluding chapter bridges traditional time series methodology with modern machine learning approaches, focusing on applications involving high-frequency data in fields such as finance, medicine, and biology. The chapter introduces the three main subdivisions of machine learning—supervised learning, unsupervised learning, and reinforcement learning—and discusses their relevance to time series forecasting problems. The tsfGRNN (time series forecasting using Generalized Regression Neural Networks) package in R is presented as a practical tool for rapid, automated forecasting. GRNNs are highlighted for their advantageous properties: single-pass learning, a single tunable parameter, and deterministic results that eliminate the need for multiple training runs. An illustrative example using UK gas consumption data demonstrates the implementation of GRNN-based forecasting with R code. The chapter discusses emerging applications of time series methods in epidemiology for tracking disease incidence and prevalence, including a case study of heatwave mortality analysis using Poisson generalized additive models. Climate and environmental applications are addressed through rainfall time series analysis using seasonal ARFIMA models to capture long-range dependencies in hydrological data. The chapter emphasizes the complementary roles of traditional statistical time series methods and machine learning approaches in addressing complex real-world forecasting challenges.