Abstract: In the context of deepening educational evaluation reform, how to leverage artificial intelligence to intelligently identify and provide precise feedback on online learning processes has become a key issue in enhancing teaching quality. To address this, this study proposes an intelligent evaluation model that integrates learning behavior trajectories and learning outcomes data, to support the identification and stratified diagnosis of students' learning states. The model, based on the Dynamic Time Warping(DTW) measures the temporal similarity of learning behaviors, and bidirectional fusion modeling is achieved by combining K-Means and K-Medoids clustering. It utilizes learning logs and academic performance data from 595 college students to construct a dual-dimensional evaluation framework covering both learning processes and outcomes. The analysis identifies five typical learner profiles and identifies representative behavioral paths through lag sequence analysis. Results reveal a significant positive correlation between the depth of the learning process and the level of learning outcomes. The model demonstrates strong recognition and diagnostic capabilities, which can support personalized teaching interventions and stratified guidance. This study not only validates the application value of artificial intelligence methods in process assessment but also provides theoretical and methodological support for enhancing teaching quality and innovating evaluation systems under the context of building China into a leading country in education.