Regression analysis and machine learning models for finding optimal growth parameters of Сherax quadricarinatus in closed water supply installations

Abstract

The article examines the dependence of the growth value for a promising aquaculture object - Australian Redclaw crayfish (Cherax quadricarinatus) – depending on a number of growing parameters in the installation of a closed water supply system recirculating aquaculture system. To analyze this dependence, the authors use methods of linear regression analysis, as well as approaches related to the creation of machine learning models based on bagging algorithms and random forest. Based on the results of the analysis, a conclusion was made regarding the best models. Thus, simple linear models were not accurate enough to build an effective predictive model, while ensemble models based on decision trees demonstrated a significantly higher degree of accuracy in constructing a regression relationship. At the same time, of the two studied algorithms of ensemble models, taking into account the sample size, the algorithm based on bagging turned out to be relatively more accurate than the algorithm based on a random forest. At the same time, both approaches are considered promising for solving aquaculture problems associated with predicting the efficiency of crayfish farming depending on the cultivation conditions.