Parametric and Non-Parametric Mixture of Regression Models for Agricultural Economic Data

The initiative is expected to result in an increase in the Gross Domestic Product (GDP) as an outcome of a greater understanding of the association between agricultural factors and economic development. Therefore, the relationships between crop yield, arable land, and GDP are examined in ten Southeast Asian countries from 1960 to 2015. Consequently, techniques such as the Baye factor, a mixture regression model, and a Gaussian process mixture model have been used to analyze the correlations between two or three variables in linear and nonlinear systems. The specific goals of this thesis are to first find the structure of relationships between variables and then construct optimal clusters for the best clusters of relationships between variables discovered through the process of research. 

The Bayes factors are used to develop a non-parametric model based on the interaction between economic growth, arable land, and GDP per capita in time series data. The experimental results reveal that the dynamic model predicts changes in variables over time. Furthermore, the use of a mixture model for the modelling of heterogeneous and multi-population data has been investigated. This thesis represents a significant strategy for improving our insight into the relationship between agricultural factors and GDP. The prediction equations can be fitted to a revised method of the mixture regression model with two or three variables based on some of the aspects of inference about the unknowns in the models, and we examine the mixture of more complex models, including the representation of maximum likelihood estimation through an expectation-maximisation algorithm (EM). Additionally, the results show the prediction models that are suited to the three variables used in the Gaussain process for estimation purposes. To that end, this work studies how to determine the best explicit function and it discovers suitable models for a group of clusters.