Local universities serve as critical incubators for national innovative talent and primary drivers of regional technological advancement. Higher education institutions face many challenges in innovation and entrepreneurship education. This study analyzes methodological approaches for assessing student innovation capabilities. Employing deep belief networks and multi-objective decision algorithms within educational contexts, the research examines the distinctive characteristics of wniversity students from different disciplines within this educational framework and explores how their engagement contributes to the enhancement of both innovation/entrepreneurship capabilities and academic performance. Survey questions were designed across four dimensions: internal/external characteristics and subjective/objective factors. Longitudinal data were collected from students across various academic years. A deep belief learning network was utilized to establish relationships between students’ learning characteristics and their acquired innovation/entrepreneurship credits and grade point averages (GPAs). Using strategies identified in the questionnaire as control variables, a multi-objective decision-making approach was applied to derive optimal cultivation strategies tailored to diverse student profiles. The findings demonstrate that optimizing the allocation of institutional innovation education resources and promoting the organic integration of curricula, projects, and competitions significantly enhances students’ innovation capabilities while concurrently improving learning efficiency.