Service recommendation method based on compatibility perception

The invention relates to a service recommendation method based on compatibility perception. The method comprises the following steps: extracting application program development demand features by using a Transform model, and then deeply mining the matching degree of candidate services and development intentions by using a deep neural network to obtain text preferences of an application program for each service; an application program service category network diagram is obtained through application program-service historical calling records, application program-service calling information is deeply mined through a KGAT model, application program features and service features are extracted, and therefore the structural preference of an application program for all services is obtained. According to the method, a weighted service correlation graph can be calculated through an application program service category network graph, a Node2Vec model is utilized to perform preliminary extraction on features of services