APPLICATION OF ARTIFICIAL NEURAL NETWORK FOR TIME SERIES DATA FORECASTING
Big data is frequently arranged by its observed time, thus forming a time series. To determine values in the future, it’s important to learn the characteristics of the past values to obtain an appropriate mathematical model. The common model used for time series data forecasting is ARIMA. However...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Project |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/26040 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Big data is frequently arranged by its observed time, thus forming a time series. To determine values in the future, it’s important to learn the characteristics of the past values to obtain an appropriate mathematical model. The common model used for time series data forecasting is ARIMA. However, as a consequence of a more powerful and affordable computational power available, artificial neural network has emerged as an alternative model. Differs from ARIMA which require stationarity assumption of the process, artificial neural networks is highly flexible that it can be applied to any kind of data. This thesis presents neural networks as an approach to model and forecast time series data and how it compares to ARIMA model in terms of prediction for time series data including daily closing price of PT. Unilever Indonesia, Tbk, Tesla, Inc., and PT. Wijaya Karya (Persero), Tbk. The results show that the artificial neural network approach leads to better results than the ARIMA models for the PT. Unilever Indonesia, Tbk and the PT. Wijaya Karya (Persero), Tbk data. However, both neural networks and ARIMA models fail to yield satisfactory results for the Tesla, Inc. data. |
|---|