Abstract: Frequent items detection has been an important feature of many applications, but it is a new area of research for emerging uncertain databases. A new definition of frequent items detection for uncertain data is proposed, thereby forming the basis for two efficient filtering rules that can significantly reduce the number of items to be detected. Furthermore, an efficient algorithm UFI is proposed to detect frequent items on uncertain databases. The UFI algorithm locates the recursive rule in the probability computation and greatly improves the efficiency of single data detection. These proposed methods can efficiently narrow the field of candidates and reduce corresponding searching space, thereby improving performance of query processing of uncertain data.