A data-driven approach towards the full anthropometric measurements prediction via Generalized Regression Neural Networks

Rapid in measurements of brown tide algae cell concentrations using fluorescence spectrometry and generalized regression neural network

The frequent occurrence of brown tide pollution in recent years has brought great losses to the economy of coastal areas. Therefore, accurate and efficient detection of the brown tide algae cell concentration is of great significance to the prevention of marine environmental pollution. In this paper, a combination of three-dimensional fluorescence spectroscopy and generalized regression neural network is used to detect the concentration of Aureococcus anophagefferens (A. anophagefferens). Firstly, the fluorescence spectrometer was used to collect spectra of A. anophagefferens with different growth cycles and different concentrations. In order to reduce the complexity of fluorescence spectral data and improve the efficiency of model calculation, the gradient boosting decision tree (GBDT) algorithm is used to rank the importance of spectral features, and select spectral features with great importance as input and concentration of algal cells as output. In view of the nonlinear relationship between input and output, a generalized regression neural network model optimized by the improved sparrow search algorithm (FASSA-GRNN) was established to predict the concentration of algae cells, The model results show that MSE is 0.0046, MAE is 0.0569, and R² is 0.9955. In addition, the FASSA-GRNN model is compared with the prediction results of the SSA-GRNN, GWO-GRNN, and GRNN models. The results show that the prediction accuracy of FASSA-GRNN is better than other models, and the improved sparrow search algorithm (FASSA) can reach the global optimum faster during the training process. This research provides a new method for predicting the concentration of algae cells.