Abstract: 【Purpose/significance】The rapid development of information technology and Internet technology has brought the education industry into the era of big data.As learners' learning behavior in the whole learning process can be recorded completely,education big data provides new ideas for the deep reform of education evaluation【. Method/process】From the perspective of learner behavior charac⁃teristics,this paper designs a learner comprehensive evaluation model driven by multi-source process data.In this model.the streaming data clustering algorithm is used to process the continuous influx of learners' data.generate or update learners' portraits in time.then analyze learners' learning behavior based on learners' portraits,and build a comprehensive evaluation model for learners.so as to com⁃
plete the real-time feedback of learners' learning performance【. Result/conclusion】The model can comprehensively evaluate the learn⁃ing process of learners and give feedback in time, which is helpful to the development of teaching evaluation, enriches the existing teaching evaluation system.and provides the basis and reference direction for realizing the virtuous cycle of teaching evaluation and op⁃timization.【Innovation/limitation】In this paper.,a process data-driven dynamic comprehensive evaluation model for learners is pro⁃posed.and the research-based model will be applied to the actual data.