Web service composition refers to the aggregation of web services for producing composite solutions in order to satisfy user requirements which can't be satisfied by atomic services. It is an essential challenge to find the most reliable and trustable complex services in each composition process. Many of current composition approaches use QoS (Quality of Service) values to select among different composition candidates. However, QoS values can't be trusted all the time since some service providers may promote their services by publishing wrong QoS values. Social network analysis techniques such as PageRank have been used successfully in finding the trustable and authoritative web resources. We believe that these techniques can also be used to improve the service composition process. We have developed a modified PageRank algorithm called Service Rank in order to find the importance level of each service in a composition based on its connectivity and invocation history. This can be accomplished by assigning higher weights to links which have more number of invocations, more up-to-date invocation time and contract signing time, and longer contract durations. Eventually the Service Rank score will be combined with the QoS score for composition ranking. Preliminary results from our experiments have proved the effectiveness of this method. As a consequence users can be more satisfied with the service composition result.