This paper deals with the application of locally weighted learning for forecasting time series corresponding to a wide range of ARMA(p,q) models. The objective of this paper is to explore the feasibility of locally weighted learning in time series forecasting. The study adopted a simulation approach to generate random samples corresponding to different time series models. The samples were divided into two sets: training and test sets. The training set was used to estimate the parameters of the locally weighted learning whereas the test set was used to test its performance. The results of the locally weighted learning were compared to those obtained from using Box-Jenkins modeling approach. The results of the study show that locally weighted learning outperforms Box-Jenkins modeling approach based on the criteria used which are mean squared error (MSE), mean absolute error (MAE) and ratio of the estimated data points closer to actual data points (Ratio).
Al-Shawadfi, G., & Al-Hindi, H. (2021). Locally Weighted Learning for ARMA Time Series.. MEJ- Mansoura Engineering Journal, 28(3), 1-10. doi: 10.21608/bfemu.2021.141480
MLA
Gamal Al-Shawadfi; Hindi A. Al-Hindi. "Locally Weighted Learning for ARMA Time Series.". MEJ- Mansoura Engineering Journal, 28, 3, 2021, 1-10. doi: 10.21608/bfemu.2021.141480
HARVARD
Al-Shawadfi, G., Al-Hindi, H. (2021). 'Locally Weighted Learning for ARMA Time Series.', MEJ- Mansoura Engineering Journal, 28(3), pp. 1-10. doi: 10.21608/bfemu.2021.141480
VANCOUVER
Al-Shawadfi, G., Al-Hindi, H. Locally Weighted Learning for ARMA Time Series.. MEJ- Mansoura Engineering Journal, 2021; 28(3): 1-10. doi: 10.21608/bfemu.2021.141480
)
View on SCiNiTO