Document Type

Presentation

Publication Date

8-1-2014

Abstract

This paper shows how to use product unit neural networks (punn's) to derive a data-driven model for the prediction of ecological quality in surface water. In a comparison with other approaches the punns provide by far the best prediction. Moreover they reveal the underlying relations between characteristics and ecological quality. Encouraged by the European Water Framework Directive many measures are taken by Dutch government to improve water quality en ecological quality of surface water. Although it is well known which type of measures are most effective in what cases, it is uncertain what the total effect of a measure is. One way to find out this is to make use of collected data. Over the past ten years data has been collected of ecological quality of Dutch waterbodies. The ecological quality is summarized in 4 ecological quality ratios (EQR's): phytoplankton, macrofauna, aquatic flora, fish. The EQR's are measured as well as related characteristics like water quality (e.g. Phosphorus and Nitrogen), hydromorphology (e.g. meandering) and maintenance. The waterbodies are merged into 8 watertype clusters (e.g. shallow lakes, fast flow streams). Each cluster requires a model for the prediction of the EQR's. For the development of the data-driven model a product unit neural network (punn) was used. A punn is a special type of neural network that uses products instead of weighted sums. Punns have a number of advantages. First of all even a small network provides a good prediction. Moreover punns allow for simplification (pruning). Only the essential terms remain, thus revealing the structure in the underlying dataset. Finally a punn is a simple transportable formula that can be easily implemented in codes. In the paper we will describe the dataset in more detail, introduce product unit neural networks and show the good results for EQR's.

Comments

Session S5-02, Special Session: Computational Intelligence in Data Driven and Hybrid Models and Data Analysis II

 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.