relationship between Naïve Bayes and Bayesian networks
Bayesian networks are a probabilistic graphical model that uses Bayesian inference for probability computation, while Naïve Bayes is probabilistic classifiers based on the application of Bayes theorem.
The Bayes theorem incorporates strong naïve independence assumptions between its features. Jiang et al. (2016) maintained that the Bayesian Network is more complicated than the Naïve Bayes. However, they both perform well to an extent except Bayesian network datasets that perform worse than the Naïve Bayes. The Bayesian network has fifteen distinct attributes that are partially discarded during structured learning.
Process of developing a Bayesian model
The process of Bayesian network model development involves expert-driven identification of model variables. Gabbay and Rozenberg (2017) reported that the variables are considered to be estimating the risk of violent re-offense. This helps in alternative identification at first.
The second step is constructing a causal model structure based on the variables identified in the first step. This will be the key to the progress in the following steps.
The third step is to link the relevant data to model variables. Wu (2015) found that establishing the link between the variables is an essential stage in Bayesian development. It is after making these links that the algorithm is determined. Then, there is the performance of model parameterization and the expectation-maximization algorithm for dealing with missing data.
The fifth and last step is experts review on the resulting behavior of the model. It is at this stage that they may suggest further revisions where necessary in the network model.
Reference
Jiang, Q., Wang, W., Han, X., Zhang, S., Wang, X., & Wang, C. (2016, August). Deep feature weighting in Naive Bayes for Chinese text classification. In 2016 4th International Conference on Cloud Computing and Intelligence Systems (CCIS) (pp. 160-164). IEEE.
Gabbay, D. M., & Rozenberg, G. (2017). Reasoning Schemes, Expert Opinions and Critical Questions. Sex Offenders Case Study. IfCoLog Journal of Logics and Their Applications.
Wu, X., Liu, H., Zhang, L., Skibniewski, M. J., Deng, Q., & Teng, J. (2015). A dynamic Bayesian network based approach to safety decision support in tunnel construction. Reliability Engineering & System Safety, 134, 157-168.