A Comparative Case Study on Time Series Prediction
Anıl Özdemir Furkan COŞKUN Selim BALCISOY
AbstractA time series is a sequence collected at consecutive equally spaced points in time. The basic idea behind the time series forecasting is the use of a model to estimate future values based on previously observed ones. Traditionally, statistical methods are used to forecasting time series however, Machine Learning (ML) algorithms have been also proposed as alternatives to statistical methods in past decades. In this paper, we evaluate forecasting performance of different ML algorithms and statistical methods on Turkey automobile sales. Recently, various of work has claimed that traditional statistical methods dominate the ML solutions in terms of time series forecasting. This study discusses different aspects of ML and statistical methods and compare their performance on different time series.