Comparative application of artificial neural networks and genetic algorithms for multivariate time-series modelling of algal blooms in freshwater lakes

Abstract

The paper compares potentials and achievements of artificial neural networks and genetic algorithms in terms of forecasting and understanding of algal blooms in Lake Kasumigaura (Japan). Despite the complex and nonlinear nature of ecological data, artificial neural networks allow seven-days-ahead predictions of timing and magnitudes of algal blooms with reasonable accuracy. Genetic algorithms possess the capability to evolve, refine and hybridize numerical and linguistic models. Examples presented in the paper show that models explicitly synthesized by genetic algorithms not only perform better in seven-days-ahead predictions of algal blooms than artificial neural network models, but provide more transparency for explanation as well.

Prediction of algal blooms using genetic programming

“…Similar conclusions on the variables affecting algal bloom modeling has been arrived at by other researchers, for example, Recknagel et al (2002), Bobbin and Recknagel (1999), Lee et al (2003) and Muttil and Lee (2005). Figure 4 shows the plot of prediction obtained with model 10 in Table 5 (presented as Eq (7) above) and the actual Chy t+14 .…”

“…Recknagel et al (1997) did a study on predictive potential of phytoplankton models by ANN and compared with other models such as AD HOC inductive models and found that predictive accuracy improved with increased event and time resolution of data. Recknagel et al (2002) compare the potential of ANN and GA in terms of forecasting and understanding of algal blooms in Lake Kasumigaura, Japan and found that models evolved by GA performs better than ANN models and provide more transparency for physical explanation as well. Lui et al (2007) studied modeling of algal bloom with vector autoregressive model with exogenous variables in Hong Kong.…”

Validity and Errors in Water Quality Data — A Review

“…Lately, machine learning algorithms like the artificial neural networks (ANNs) [67,[76][77][78], and genetic algorithms (GA) [76,79] have gained momentum in water quality monitoring studies. [41] pointed out that these technique generally yields the best parameter estimates in the data set with the least amount of missing data.…”

Data mining and multivariate statistical analysis for ecological system in coastal waters

“…These include artificial neural networks (Recknagel et al 1997(Recknagel et al , 2002Yabunaka et al 1997;Maier et al 1998;Scardi & Harding 1999;Karul et al 2000;Jeong et al 2001;Scardi 2001;Wei et al 2001;Lee et al 2003), evolutionary based techniques (Bobbin & Recknagel 2001;Recknagel et al 2002;Jeong et al 2003;Chau 2005;Muttil & Lee 2005), fuzzy and neuro-fuzzy techniques (Maier et al 2001;Chen & Mynett 2003), and so on. DM techniques with the goal of "description" have also been used, but to a lesser extent.…”