Aiming at the fact that the traditional method ignores the temporal information and correlation contained in the event time series in the process of prediction, a network hotspot event heat prediction model based on the EKSC algorithm is proposed. The framework will distinguish the content information and relationship information of social networks, fetch, and store them in a targeted manner, and form a feedback mechanism between fetching and storing, which supports developers to expand business-related dynamic fetching strategies. The practice teaching mode is constructed from the aspects of the practical teaching link and the evaluation system, and a practical teaching course system of "one main line, two channels and three stages" is proposed, and it is recommended to combine different learning stages and courses to use inquiry learning.