A Bayesian Network for Symptom-diagnosis Data
Abstract
Bayesian network is an important expression tool of uncertain knowledge by integrating probability table and graph. This paper applies Bayesian network into clinical data for knowledge discovery and constructs a Bayesian network for symptom-diagnosis data with optimal structure. First of all, a Markov chain-Monte Carlo based Metropolis-Hastings sampling method is introduced to fill in the missing data; Then, a K2 algorithm is used to search for all possible Bayesian networks among the relationship between symptoms and diagnoses; Further, the BDE scoring function is used to determine the optimal network structure illustrating the relationship between symptoms and diagnoses. The results show that the network structure accuracy has greatly improved after filling the missing data with the method of Metropolis-Hastings sampling.
Keywords
Bayesian, Markov chain-Monte Carlo, Metropolis-Hastings sampling, K2 algorithm, BDE scoring function.Text
DOI
10.12783/dtcse/cmsms2018/25238
10.12783/dtcse/cmsms2018/25238
Refbacks
- There are currently no refbacks.