Abstract: In view of the lack of pertinence in teaching plan and timeliness in teaching effect valuation，an early warning model of interactive teaching system is proposed based on rule detection and outlier detection． Analytical algorithms based on the learning process and academic performances are designed for automatic intervention and active intervention based on learning behavior analysis． After information visualization，the result shows that all parties involved can acquire early warning information intuitively and efficiently，which can promote the improvement of learning effect and learning success rate．

Key words: educational data mining, learning analytics, school precaution, outlier detection, similarity sorting