Best Practices for Time Series Forecasting (Tutorial)

In a fast-paced world, software systems require autonomic management. To enable accurate and proactive autonomic systems, reliable time series forecasting methods are needed. In this tutorial paper, we guide the reader step-by-step through different forecasting steps. In each step, we highlight best practices and present available approaches. That is, we explain how to pre-process the data and retrieve features. Then, the model selection and fitting steps are shown. Finally, we discuss the forecasting itself and its evaluation. For the individual steps, we provide some basic code snippets in the language R.