Investigating the predictive performance of Gaussian process machine learning in time series demographic data handling

Effective handling and modelling of time series data play a crucial role in enhancing the quality of derived information during the research process which includes addressing missing values. Meticulous attention to these data-related tasks is paramount, as the outcomes of the research are directly influenced by the quality and integrity of the processed data. This study employs a non-parametric Gaussian Process Regression (GPR) prediction algorithm in machine learning to investigate the predictive performance of Malaysian demographic data for 1960 to 2021. For robust results, the traditional parametric models were introduced for comparison. The reliability and efficiency of the algorithm are presented. The results show that the GPR with squared exponential covariance function can give the most accurate prediction on the data based on the low mean absolute deviation (MAD) and root mean squared error values (RMSE).