As an overt expression of internal mental processes, discourses have become one main data source for the research of interactive learning. To deeply explore behavioral regularities among interactions, this article firstly adopts the content analysis method to summarize students' engagement patterns within a course forum in a small private online course (SPOC) system. Secondly, through sentiment word matching and sentiment density calculation, the authors characterize the evolution trends of collective positive and negative sentiments, and compare sentiment strengths of different achieving students. The analytical result shows that there is a significant correlation between most engagement patterns and academic achievements, and high-achieving group seems more active than low-achieving group in terms of interactive, register, question, viewpoint and thematic postings. Besides, both high and middle-achieving students are superior to low-achieving students on positive sentiment. But, there is no significant difference among high-, middle- and low-achieving students on negative sentiments.