The invention discloses a
disease transformation prediction method and
system for parkinson's
disease predecessor symptoms, and the method mainly comprises the steps: constructing a
Bayesian network model containing multi-
modal characteristic variables, and carrying out the
data processing, dimension reduction
processing and
feature fusion based on the multi-
modal characteristic data of a target individual, thereby achieving the prediction of parkinson's
disease predecessor symptoms. The method comprises the following steps of: firstly, extracting a
Bayesian network model, then converting various characteristic data into the disease obtaining probability of the Parkinson's disease of a target individual, effectively capturing a complex non-
linear relationship between a precursor symptom and an onset risk of the Parkinson's disease, and compared with a traditional integral
algorithm, the
Bayesian network model disclosed by the invention can comprehensively consider the interaction effect of various biomarkers and clinical parameters, so that the method has a good application prospect. Particularly, through
network structure learning guided by a
knowledge graph, implicit distinguishing of proactive subtypes is realized.