Developing a reliable data analytical method for predicting the velocity profile in small streams is important in that it substantially decreases the amount of money and effort spent on measurement procedures. In recent studies it has been shown that machine learning models can be used to achieve such an important goal. In the proposed framework, a tree-augmented Naïve Bayes approach, a member of the Bayesian network family, is employed to address the aforementioned two issues. Therefore, the proposed study presents novelty in that it explores the relations among the predictor attributes and derives a probabilistic risk score associated with the predictions. The data set of four key stations, in two different basins, are employed and the eight observational variables and calculated non-dimensional parameters were utilized as inputs to the models for estimating the response values, u (point velocities in measured verticals). The results showed that the proposed data-analytical approach yields comparable results when compared to the widely used, powerful machine learning algorithms. More importantly, novel information is gained through exploring the interrelations among the predictors as well as deriving a case-specific probabilistic risk score for the prediction accuracy. These findings can be utilized to help field engineers to improve their decision-making mechanism in small streams.