Date of Publication

2023

Document Type

Bachelor's Thesis

Degree Name

Bachelor of Science in Statistics Major in Actuarial Science

Subject Categories

Statistics and Probability

College

College of Science

Department/Unit

Mathematics and Statistics Department

Thesis Advisor

Frumencio F. Co

Defense Panel Chair

Marcus Jude P. San Pedro

Defense Panel Member

Regina M. Tresvalles

Abstract/Summary

Temperature is a key variable for understanding climate variability and change. However, climatological summary statistics based on daily minimum and maximum temperatures contain inhomogeneities that can largely affect estimations done at a local scale. Thus, this study aimed to use a spatiotemporal Gaussian process model (STGPM) to impute and predict three-hourly measurements from neighboring stations in the Greater Manila Area that may be used to obtain summaries of temperature extremes. Due to the computationally expensive nature of this method, temperature extrema predictions using Artificial Neural Network (ANN) methods such as feed-forward back propagation (FFBP), radial basis function network (RBFN), and generalized regression neural network (GRNN) have been obtained as an alternative and for prediction performance comparison. Here, comparisons are made using R2, MSE, RMSE, and IA. As the ANN models are limited to the prediction of temperature extrema, the STGPM inherently gives more information through the imputation and prediction of three-hourly temperature. When comparing performance scores, The STGPM also performs best based on the measures of model performance namely MSE, RMSE, R2, and IA. However, among the ANN models, GRNN performs best in terms of the following measures of model performance: R2, MSE, and RMSE.

Abstract Format

html

Language

English

Format

Electronic

Keywords

Temperature forecasting, Minimum

Upload Full Text

wf_yes

Embargo Period

8-14-2023

Download

Share

COinS