In the era of big data and digital information, blended learning has emerged as a prominent instructional paradigm. With the growing demand for business Japanese education in China, both the scale and scope of instruction are rapidly evolving. This trend has prompted increasing attention to the integration of online learning environments with traditional teaching models, particularly in terms of learning assessment. This study focuses on developing a scientifically grounded evaluation index system tailored to blended learning in business Japanese, employing algorithmic optimization and experimental validation as its core methodologies. Centered around high school learners, the research draws from educational psychology and learning theory to propose strategic guidance from institutional, instructional, and individual perspectives. Empirical analysis and model testing support the proposed system, with the optimized Nivre-based evaluation framework demonstrating strong adaptability to contemporary educational demands. The model achieves an accuracy exceeding 92% across various test datasets, verified through correlation analysis and reliability-validity metrics, indicating its robustness and applicability. The study provides a valuable reference for advancing blended teaching practices and enhancing the effectiveness of business Japanese education at the secondary level.