Extensions of the Neural Network Models into Applications and Comparisons with General Linear Models

This dissertation is designed to answer the following questions: (1) Which measurement model is better to contribute to the research model in different areas? (2) Within a given model, how does the data size influence the performance of a neural network (NN) and some other methods? (3) Compared to partial least square (PLS), ordinary least square (OLS), XGBoost, how is the performance of NN? Essay 1 systematically compares PLS-SEM to ANN and builds the hybrid vehicle purchasing intention model (HVPIM). It investigates different models that have been previously applied to study the theory of planned behavior (TPB). The methods find those factors that significantly correlated with consumer purchase intention. Essay 2 posits, develops, and tests a PNN model with healthcare data. A research survey is designed and distributed to undergraduate students from a major research school in the U.S. southwest region. Research hypotheses are tested using PLS-SEM and PNN. Essay 3 targets on testing the performance of the NN model with panel data from the soccer transfer market. To achieve this purpose, the essay posits and develops an empirical test built on game theory. The NN model is tested and compared to OLS and XGBoost. As the research compares the different types of NN models with other linear models, NN does have extensive applications and outperforms some other methods. The contribution of this research falls into both academia and industry. While theoretical science and practical applications may be mutually complementary, real demand is often the driving force behind technological growth. Currently, the practicality of neural networks has many accomplishments, such as smart control, pattern recognition, robot control, etc. It can be expected that using NN algorithms to process the necessary data will improve its applicability in real life