The K Means clustering algorithm can not determine the initial cluster ing centers, which results in low accuracy and inability to reflect the interest ing hotspots. Here, algorithm based on clustering is proposed through applying Min M ax active learning strategy to ask the user for identifying the seed points. Several points are provided in small quantities of query for users to confirm the initial centers, and the weight is set in the recalculation of K Means centers, which reduces the number of iterations and improves the accu racy of clustering results. Moreover, the hot topics are obtained by applying th is algorithm to the micro blog clustering analysis.