Indirect Measurement of Dissolved Oxygen Based on Algae Growth Factors Using Machine Learning Models

Abstract

Excessive algae growth level has become a major concern in sustaining an acceptable quality marine life. Algae bloom levels significantly affect the amount of dissolved oxygen (DO) in a certain body of water. The amount of DO level indicates if the oxygen in the water is enough to provide and support the ecosystem and if the marine environment is suitable for aquatic organisms for healthy survival. A trophic state assessment was done using dissolved oxygen prediction model relating to algae growth factors is proposed to address this issue. The models are trained and tested using five different estimators, namely: multilinear regression (MLR), artificial neural network regression (ANN-R), support vector machine regressor (SVR), gaussian process regressor (GPR), and k-nearest neighbor regression (KNN-R) based on algae growth factors from a marine culture as attributes; which include temperature, power of hydrogen (pH), and specific conductance of water. The target data used is the DO level. Each of the predictors is remarkably contributing to the algae growth level, while DO indicates the level of algae bloom. The relationship between the two sets of data were produced from the models and will be very important in simplifying systems by minimizing DO sensors needed usually for water quality monitoring. Cross-validation R2 values obtained were: 0.88, 0.91, 0.91, 0.92, and 0.93 respectively as mentioned above.