In the age of Web2.0, various network services became critical. Exchange of messages between entities in the network is so frequent that information explosion is quite common nowadays. Volume of Information passed is growing up rapidly. With the wide development of web applications, people need to learn how to filter the important messages; service providers have urgent need to trace the ever changing role of users. This research studies the detections of the user interaction scenario, based on the result from the test function in the on-line learning platform. The learning platform users are divided into two groups, teachers and students, based on their roles. Usually students sit for an on-line examination at the end of each learning activity. The teachers are in charge of helping students with their presentations, encouraging those with good grades, and helping the weaker ones to reach their potential. But in the one-to-many teaching method, a teacher needs to face many students and the resultant grade of an examination becomes a heap of fuzzy and difficult to comprehend numbers. Even though some mathematical tools can help the teachers analyze the data, it is still very difficult to provide appropriate response to each student. The purpose of this research motives building an examination system which combines context-awareness and auto-notification, and bring the advantages of digital examination. An inference engine is used to calculate linear regression of learning curve for each student, then review the old data, and transfer the analysis into the learning context. Then the feedback is given to the students under the various learning context or the teacher will get notification after it compile the analysis. Besides analyzing the past data, the linear regression result will be adjusted to fit the characteristics of learning curve and infer the personal goal of the student. If result is better than expected goal, students should be encouraged. On the other hand, the remediable actions will be administered. Those events can be scheduled by the manager of auto-notification system, published in the appropriate time, and achieve the goal of variety, personalization, and automation.