Water is a main factor in aquaculture, its quality plays an important role in fish farming management. The non-linearity, dynamics and non-stability of its parameters make it a very complex system to manage. The classical methods used to judge if the water quality is valid for fish farming or not are not very effective. To have good results, the involvement of technology is necessary. The use of artificial intelligence and a machine learning techniques is a good solution in this context. A DTKNN+ model is proposed in this paper. It is a new hybrid approach that combines decision tree with k-nearest neighbors(KNN). Many machine learning techniques were used in this new approach. The results showed the DTKNN+ effectiveness compared to a simple KNN. Its accuracy score is worth 99.28% and its mean absolute error value did not exceed 0.0071. The error rate is also decreased from 1674 misclassifications on 103544 with KNN to 743 on 103544 with DTKNN+.