Outlier treatments using interolation on Malaysia tourist arrival forecasting: SARIMA and ANN approaches
Outliers are unusual observations that appear in a piece of data that are very different from the rest of the data. The presence of an outlier may directly affect the variance, the model parameters, and the overall estimation, especially during forecasting. To obtain an accurate forecast, any...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English English English |
| Published: |
2020
|
| Subjects: | |
| Online Access: | http://eprints.uthm.edu.my/1095/1/24p%20NORSORAYA%20AZURIN%20WAHIR.pdf http://eprints.uthm.edu.my/1095/2/NORSORAYA%20AZURIN%20WAHIR%20COPYRIGHT%20DECLARATION.pdf http://eprints.uthm.edu.my/1095/3/NORSORAYA%20AZURIN%20WAHIR%20WATERMARK.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Outliers are unusual observations that appear in a piece of data that are very different
from the rest of the data. The presence of an outlier may directly affect the variance,
the model parameters, and the overall estimation, especially during forecasting. To
obtain an accurate forecast, any outliers that are present in the data must be addressed.
This research used monthly Malaysia tourist arrivals from 1998 until 2015 and an
ARIMA outlier detection method to detect outliers on original data. The detected
outliers were regarded as missing values then treated using interpolation method which
are Linear Interpolation and Cubic Spline Interpolation methods. In this study,
SARIMA model and Artificial Neural Network model were used as forecasting tools
using the data before and after outlier treatment. The comparison of forecast
performance between all models were calculated using MSE, MAD, MAPE and R2
including the data before and after outlier treatment. This study found that once the
outlier in the data was treated, ANN model of Cubic Spline Interpolation performs the
best models compare to other models which is 95.65% using R2
validation test. On the
other hand, ANN approach outperforms SARIMA approach on both data for before
and after outlier treatment which are 6.05% and 2.52%. |
|---|