Combining group method of data handling models using artificial bee colony algorithm for time series forecasting
Time series forecasting which uses models to predict future values based on some historical data is an important area of forecasting, and has gained the attention of researchers from various related fields of study. In line with its popularity, various models have been introduced for producing accur...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2019
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/id/eprint/96778/1/NurHaziyatulAdawiyahMSC2019.pdf.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Time series forecasting which uses models to predict future values based on some historical data is an important area of forecasting, and has gained the attention of researchers from various related fields of study. In line with its popularity, various models have been introduced for producing accurate time series forecasts. However, to produce an accurate forecast is not an easy feat especially when dealing with nonlinear data due to the abstract nature of the data. In this study, a model for accurate time series forecasting based on Artificial Bee Colony (ABC) algorithm and Group Method of Data Handling (GMDH) models with variant transfer functions, namely polynomial, sigmoid, radial basis function and tangent was developed. Initially, in this research, the GMDH models were used to forecast the time series data followed by each forecast that was combined using ABC. Then, the ABC produced the weight for each forecast before aggregating the forecasts. To evaluate the performance of the developed GMDH-ABC model, input data on tourism arrivals (Singapore and Indonesia) and airline passengers’ data were processed using the model to produce reliable forecast on the time series data. To validate the evaluation, the performance of the model was compared against benchmark models such as the individual GMDH models, Artificial Neural Network (ANN) model and combined GMDH using simple averaging (GMDH-SA) model. Experimental results showed that the GMDH-ABC model had the highest accuracy compared to the other models, where it managed to reduce the Root Mean Square Error (RMSE) of the conventional GMDH model by 15.78% for Singapore data, 28.2% for Indonesia data and 30.89% for airline data. As a conclusion, these results demonstrated the reliability of the GMDH-ABC model in time series forecasting, and its superiority when compared to the other existing models. |
|---|