The composition service selection based on quality of service (QoS) is still challenging because the difficulty of obtaining user's requirements and the complex and uncertain environment make decision making hard. Firstly, a user weight model closer to user's expression habit to obtain the user's weight is presented. And then a novel composition Web service selection algorithm (CWSSA) based analytic hierarchy process (AHP) and technique for order preference by similarity to an ideal solution (TOPSIS) is presented to solve the problem of optimal service selection in the complex and uncertain environment t for the first time. It can convert user's pair-wise comparison matrix of QoS attributes to user weight and take QoS information with interval number into consideration. Other contributions of the paper include a user weight model, a QoS model with interval number and strategy for aggregating QoS. Finally, the benefits and effectiveness of our approach are demonstrated.