Learning Bayesian networks from demographic and health survey data

Child mortality from preventable diseases such as pneumonia and diarrhoea in low middle-income countries remains a serious global challenge. We combine knowledge with available Demographic Health Survey (DHS) data India, to construct Causal Bayesian Networks (CBNs) investigate the factors associated childhood diarrhoea. make use of freeware tools learn graphical structure DHS score-based, constraint-based, hybrid learning algorithms. effect missing values, sample size, knowledge-based constraints on each algorithms assess their accuracy multiple scoring functions. Weaknesses survey methodology available, well variability CBNs generated by different algorithms, mean that it is not possible definitive CBN data. However, are found be useful reducing variation graphs produced produce which more reflective likely influential relationships Furthermore, valuable insights gained into performance characteristics Two score-based particular, TABU FGES, demonstrate many desirable qualities; (a) sufficient data, they graph similar reference graph, (b) relatively insensitive (c) behave constraints. The results provide basis for further investigation deeper understanding behaviour when applied real-world settings.