TIME SERIES MODELING USING ARIMA WITHINTERVENTION AND OUTLIERS FACTOR

• Main Author: Tiara Monika, Ficilia
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/82334
• Institution: Institut Teknologi Bandung

In general, the results of time series observations are influenced by the values at the previous time. However, in reality time series are often influenced by external events which can be defined as interventions or outliers. To identify the effect of the intervention on the time series, time series modeling with intervention factors is used. If the time and cause of the event are unknown, then time series modeling with outlier factors can be used. The aim of this research is to determine the influence of intervention factors or outliers on time series and what conditions of time series data are appropriate to model using intervention or outlier time series modeling. Therefore, two types of data are used, namely crime in Bandung City and medium rice prices in Indonesia. The selection of the best model is based on model performance which is compared through AIC, RMSE, and MAPE values. The research results show that the time series model without intervention and outliers provides a better level of prediction accuracy for crime data in Bandung City with a MAPE of 26.15%. Meanwhile, for medium rice price data in Indonesia, the combined model with intervention factors and outliers provides the best prediction performance with a MAPE of 0.76%. The results of this study also show that crime is not significantly influenced by interventions and outliers, while rice prices are quite sensitive to existing interventions and outliers. With certain data characteristics, the ARIMA model with intervention factors and outliers has better performance so that predictions can be made with better accuracy.