Development of Environmental Quality Predictor Using Feedforward Artificial Neural Network (Fann) In Matlab Graphical User Interface (Gui)
Loading...
Date
2015-05-01
Authors
Nazira Anisa Rahim
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The environmental conservation efforts always deal with the complexity
problem as it involves a large number of variables. However, in the process
development of the model, the correct input selection for the corresponding output
prediction is so important. Furthermore, traditional reports on the environmental
quality tend to be too technical, presenting monitoring data without providing a
complete and easy to understand facts of the environmental quality. Due to the
redundancy of the environmental datas, input data selection methods were
introduced; Canonical Correspondence Analysis (CCA) and Canonical Correlation
Analysis (CCorA). These approaches could be applied as a feature selection tools
and combined with Feedforward Artificial Neural Networks (FANN) to develop the
graphical prediction interface for the end users. The proposed graphical userinterface
for environmental prediction, will give an indication of the water and air
pollution degree and their qualities, with the terms that are familiar within the
community. To achieve those objectives, this research was divided into three main
phases; determination of input feature selection, FANN model development and
finally, GUI development for offline monitoring. Two case studies were used in
this research which was based on river water and air quality data. The application of
CCA and CCorA to determine the input for the prediction was successfully applied
with 7 (SS, NO3, K, NH3-NL, TS, Zn and Tur) and 3 (humidity, temperature and
wind speed) input variables were selected for Case Study 1 and 2, respectively. The
results show that the developed prediction networks for the environmental quality
prediction system has been executed well for less of input data. The developed
prediction system based on FANN with the combination of CCA and CCorA
generally has generally performed well and helped in simplifying the environmental
prediction system. The final multi-input single output (MISO) models that have been
used to predict the water quality index (WQI) and air pollution index (API) were
successfully developed with the regression values of 0.90 and 0.91 for both of the
networks for the unseen validation data input.