Exploring the vague environmental factor via data-driven methods: An application of GA-ANN for environmental investigation

Using some numeric factors to represent certain environmental factors is the essential method to understand the relationships between environmental phenomena and environmental factors. Sometimes environmental researchers need to set some ordinal-scale variables to describe several differences that might exist in certain properties. This research offers one process which is mainly combined by Genetic Algorithm (GA) and Artificial Neural Network (ANN). Through using this GA-ANN process can assist for determining one ordinal-scale variable in environmental experiment. The experiment area in this paper is Feng-Chia night market, Taichung, Taiwan. This research uses ANN to establish a spatial interaction model for simulating the informal commercial phenomena in Feng-Chia night market. And the main goal is to test whether applying GA to determine one ordinal-scale variable can increase the ANN simulating accuracy. By simulation result given in this paper, the GA-ANN can improve the experimental performance via modifying one ordinal-scale variable, and this process can reveal more detail information and potential directions to help researcher clarify the properties of the ordinal-scale variable.