Implementation of Adaboost for the detection of the toxic response behaviour of zebrafish (Danio Rerio)

The movement behaviour of zebrafish (Danio rerio) schools was observed in response to treatment with copper at a 24 h half-lethal concentration. The behavioural characteristic parameters, which were continuously recorded into a SQL (Structured Query Language) Server database by a digital image processing system both before and after the treatment, had significant changes. Subsequently, the Adaboost algorithm was implemented to solve the data vector classification problem in normal and abnormal water. Furthermore, to evaluate the accuracy and timeliness of the classifiers, Adaboost was compared with a back-propagation neural network (BPNN) and support vector machine (SVM). The results clearly demonstrated that the prediction accuracy of the Gentle Adaboost and Real Adaboost algorithms were over 93%, which was better than the Modest Adaboost, the BPNN and the SVM. In addition, the time requirement was also acceptable. In conclusion, Adaboost is a useful computational method for the classification of water quality.