Bayesian networks (BN) belong to the category of probabilistic graph models and are used to represent knowledge about uncertain domains (Perkusich et al., 2017). Bayesian networks represent a joint probability distribution over a set of variables (Freire, Perkusich, Saraiva, Almeida, & Perkusich, 2018). Dragicevic et al. (2017) suggested that the BN model is a suitable estimation method in an agile software development methodology as it does not have an impact on agility and can be applied in an early planning phase successfully. The BN model is useful in making predictions and diagnostics with ambiguous data to determine the probability of an event (Dragicevic et al., 2017). Estimators use the method to incorporate causal factors to determine conditional probability is estimations.
The BN is a model that describes probabilistic relationships between causally related variables. The advantages of a BN are suitability for small projects, and it provides results based on incomplete data sets (Zare F., Zare H., & Fallahnezhad, 2016). The BN model's additional advantages are the explicit treatment of uncertainty and support for decision analysis (Perkusich et al., 2017). The use of BN can be advantageous in effort estimation because probability distributions can be updated as new information becomes available, and estimation models are constructed using causal influences (Perkusich et al., 2017). Bayesian networks allow for the combining of historical data with expert opinion.