Study On Growth Model Of Pearl Gentian Grouper Based On Machine Learning

The growth of aquatic products is an important factor affecting their yield,and weight is one of the important indicators.Many factors,such as physical and chemical indexes of water body,internal and external environmental indexes,all affect the growth of aquatic products.Effective regulation and control of these factors can promote aquatic products to increase weight,increase output,and achieve the effect of increasing revenue and saving expenditure.In recent years,a large number of aquaculture data have been accumulated in the factory aquaculture base.The key of this paper is to explore the mapping relationship between many factors and the growth of aquatic products,and to realize the scientific prediction of growth indexes by building models.In this paper,based on the data collected from the seawater circulating aquaculture system,combined with missing data interpolation,data principal component analysis,artificial neural network and other machine learning technologies,the construction of the growth model of the Pearl gentian grouper with body weight index was studied.The main research work is as follows:In the circulating water culture system of the sea water factory culture base,various kinds of culture and production data of Pearl gentian grouper are collected,which cover a variety of growth factors and corresponding weight indicators.In view of the lack of sampling data,in order to ensure the alignment of data based on time series in the process of machine learning and modeling,Singular Spectrum Analysis（SSA）is used to realize the interpolation of missing weight data.After completing the preparation and preprocessing of the sampling data,aiming at the phenomenon of information overlapping and redundancy in the multivariate data,and having a negative impact on the structure and performance of the neural network model,in this paper,Principal Component Analysis（PCA）is used to reduce the dimension of sampling data,and a method of selecting comprehensive growth factors based on regression test is proposed.Combined with the weight of each growth factor in the principal component,the comprehensive growth factor is explained,which lays the foundation for the subsequent growth model construction and future exploration of the correlation law of growth factors.Based on the preprocessing of modeling data and the selection of growth factors,the growth model of Pearl gentian grouper was established by combining Principal Component Analysis and Radial Basis Function（RBF）neural network.The model takes growth factor data as input and weight index as output,and uses supervised learning method to train the model.Compared with the test of stepwise regression and RBF neural network,the growth model of Pearl gentian grouper based on PCA-RBF neural network is the best in performance and precision than the others.To sum up,this paper established the growth model of Pearl gentian grouper,which was in line with the growth law,through the processing and analysis of breeding data by using machine learning technology.The model reflects the relationship between the growth index of aquatic products and the growth factors,which provides a scientific basis for further exploring the culture law of Pearl gentian grouper and improving the economic and ecological benefits of the culture enterprises.