UNIVARIATE AND MULTIVARIATE TIME SERIES REGRESSION USING TRANSFER FUNCTION MODEL AND VECTOR AUTOREGRESSIVE MOVING-AVERAGE WITH EXOGENOUS REGRESSORS (VARMAX)

• Main Author: Widyanti, Devina
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/47689
• Institution: Institut Teknologi Bandung

The Time Series Modeling is used to analyze data by taking into account of the effects of the data in the previous periods. Time Series Modeling can be used for both the univariate and multivariate time series models. The Autoregressive Moving-Average (ARMA) Model is a predictive analysis model of univariate time series data. The ARMA Model only takes into account the endogenous variables in the previous periods without considering the dependency of other variables. The Univariate ARMA Model that is generalized to handle multivariate time series is called the Vector Autoregressive Moving-Average (VARMA) Model. In an event that there are several factors influencing the event, then it is necessary to predict by adding exogenous variables to the model. Time series modeling that involves exogenous variables is also called as time series regression modeling. In this Thesis, the Transfer Function Model is used to model univariate time series regression, while the Vector Autoregressive Moving-Average with Exogenous Regressors (VARMAX) is used to model multivariate time series regression. The VARMAX Model is the ARMAX Model or the Transfer Function Model for multivariate. In this Thesis, a case study of five different data is conducted to determine which data type is suitable for the two models. The case study result shows that both the Function Transfer Model and VARMAX Model can be used for data that have outliers or interventions and data that have patterns or trends. The VARMAX Model is not affected by the size of the data observation, while The Transfer Function Model seems to be more suitable for data with a large observation size.