It's well known that human elements are deeply involved in collision of ships. Most of human errors or unsafe acts and their causational factors or performance shaping factors (PSFs) have been identified by varies methods such as accident statistics, expert judgment and accident investigation, but the relationship between them has not been depicted clearly yet, not to mention to describe it quantitatively. It is quite necessary to identify the relationship between them for reducing the negative influence of human elements on collision avoidance. In this paper, an attempt to establish the link between unsafe acts in collision of ships and performance shaping factors with the help of data mining technique is presented. For the purpose of this paper, 100 written collision accident reports are collected and analyzed. With the human elemet classification modeled by the authors, all data relating to human elements are sought out from these accident reports and put together forming a small database. With the help of data mining technique, the association rules between each kind of unsafe acts and performance shaping factors are discovered. There are mainly 6 kinds of unsafe acts and 9 kinds of performance shaping factors; each of the unsafe acts is influenced by several PSFs out of 9 PSFs respectively. For example, improper lookout is influenced by knowledge/experience/training, organization and administration, attention, constrained waters, visibility, manning, traffic density, equipment layout and capability of information management, while late action has a close relationship with organization and administration, knowledge/experience/training, constrained waters, and attention. The degree of support and credibility of the relationship between the unsafe acts and PSFs has also been calculated which provide a platform for further analysis of human elements in collision of ships.