Comparing the Performances of Logistic Regression and Artificial Neural Network Models in Predicting Swimming Conditions Along the Lincoln Beach Area of Lake Pontchartrain, New Orleans, LA

Keywords: Logistic Regression; Artificial Neural Network

Abstract: Natural bathing beaches serve as a major source of recreation throughout the United States. However, their potential for disease transmission via water contact and/or ingestion is a public health concern. The Lincoln Beach area of Lake Pontchartrain, New Orleans, LA, has for many years been polluted to an extent that swimming and other recreational activities have been drastically curtailed. In an effort to assess water quality and to gauge the recreational viability of Lincoln Beach, indicator organisms are used to help estimate the level of swimmability within a selected time period. A sampling grid of 12 sites along the beach was determined for the collection of environmental data. Indicator organisms include E. coli, Enterococci and Fecal Coliform. In order to study the distribution of the organisms, an integrated rainfall/water runoff sampling was initiated. In addition, physicochemical parameters along with environmental data were recorded. Models are developed to predict the concentration of indicator organisms and make recommendations as to the swimmability (yes/no) of the beach area. This study compares and reviews the performances of the logistic regression and Artificial Neural Network (ANN) models in predicting the concentration of indicator organisms. The ANN based models under consideration include: back-propagation; Bayesian back-propagation and radial basis function.