Time Series Forecasting using various Machine Learning Models

Abstract

Analysis of time series data is a challenging task in recent times. Statistical analysis of time series data and forecasting with the help of past data is a requirement in current times. The industry is looking forward to accomplishing complete effectiveness in forecasting. There are several established techniques such as auto regressing (AR), moving average (MA), autoregressive moving average (ARMA) and autoregressive integrated moving average (ARIMA) for univariate time series forecasting. For multivariate time series forecasting, the vector autoregression (VAR) model was used. With recent advances in deep learning techniques, prediction tasks can be effectively performed by a neural network and deep learning models can give better results than these established models. This study analyses and compares various established models with deep learning techniques on different datasets and explores whether transformers can be used for time series forecasting to get highly accurate results.