The dependency of the empirical parameters in frequent patterns mining of Traditional Chinese Medicine (TCM) prescriptions should be reduced to improve the accuracy of mining results. Aiming at the characteristics of TCM prescription data, an efficient Top-Rank-k frequent patterns mining algorithm based on Weighted Undirected Graph (WUG) was proposed. The new algorithm can directly mining frequent k-itemset (k≥3) without mining 1-times and 2-times, and then quikly backtrack to the corresponding prescription of the frequent itemsets of core drugs combination. Besides, the compression mechanism of Dynamic Bit Vector (DBV) was used to store the edge weights in undirected graph to improve the spatial storage efficiency of the algorithm. Experiments were conducted on TCM prescription datasets, real datasets (Chess, Pumsb and Retail) and synthetic datasets (T10I4D100K and Test2K50KD1). The experimental results show that compared with iNTK (improved Node-list Top-Rank-K) and BTK (B-list Top-Rank-K), the proposed algorithm has better performance in terms of time and space, and it can be applied to other types of data sets.