Performance Analysis of ANN Model for Estimation of Trophic Status Index of Lakes

Abstract

The health of water bodies across the globe is of high concern as the pollution is accelerating rigorously. With the interventions of simple technology, some significant changes could be bought up. Lakes are dying because of high Trophic Index Status which shows the eutrophication level of water bodies. Taking this into account, feed forward back propagation neural network model is used to estimate the Trophic Status Index (TSI) of lakes which could compute the value of TSI with the given parameters; pH, temperature, dissolved oxygen, Secchi disk transparency, chlorophyll and total phosphate. Two learning algorithms; Levenberg Marquardt (LM) and Broyden–Fletcher–Goldfarb–Shanno (BFGS) Quasi Newton were used to train the network, which belongs to different classes. The results were analyzed using mean square error function and further checked for the deviation from actual data. Among both the training algorithm; LM demonstrated better performance with 0.0007 average mean square error for best validation performance and BFGS Quasi Newton shows the average mean square error of 1.07.