Deep Active Genetic Learning-Based Assessment of Lakes’ Water Quality Using Climate Data

Abstract

Harmful algal blooms pose a threat to the lakes’ habitation and economy in many parts of the United States. State administrators have been devising a state-of-the-art monitoring and forecasting system for these harmful events. The efficacy of a monitoring and forecasting system relies on the performance of algal blooms detection. This paper proposes a deep active genetic learning framework, combining deep active learning as a classifier and a genetic algorithm as a feature selector, for the detection of harmful algal bloom events in the state of New York using climate data. A spatio-temporal weather data point training sample is introduced to retrieve relevant information of both harmful and non-harmful bloom classes. The most informative sample is selected through information entropy criterion to feed the model in classifying harmful blooms, and the most related features are selected using genetic algorithms. The proposed framework provides a classification accuracy of 97.14% with a reduced sample size and feature vector.