Water quality models are an important tool for forecasting dissolved oxygen (DO) concentration in water bodies and it is an essential part of every water quality management. The accuracy of the Convolutional Neural Network (CNN) in modeling the concentration of DO was investigated in this research. The parameters selected for this simulation include; Ammonia, Nitrate, pH, suspended solids, temperature, and water turbidity. The model was implemented by using the measured data obtained from uMgeni waterworks, during the year 2014. Various input combinations of these data were applied as inputs to the CNN model. With the presence of artificial intelligence and cheap computational power, many authors have leveraged it for dissolved oxygen prediction with promising results. Many models used fully connected deep neural networks to accomplish their prediction. This research takes a different approach by merging a convolution layer with a fully connected layer (FCL). The convolution layer synthesizes spatial relationships amongst the above data, and thereafter fed into the fully connected layer. Several error statistics, such as the coefficient of correlation (R), mean absolute error (MAE), and mean square error (MSE), were used to compare the performance of the proposed CNN model to that of the FCL model. The CNN model outperformed the FCL models in predicting DO concentration in water bodies, according to the findings. Therefore, results show that this approach improves prediction accuracy more than a model built totally on an FCL architecture.