TOPSIS (Technique for Order Preference by Similarity to Ideal Solution) is a very common method for Multiple Attribute Decision Making (MADM) problem in crisp as well as uncertain environment. The interval trapezoidal neutrosophic number can handle incomplete, indeterminate and inconsistent information which are generally occurred in uncertain environment. In this paper, we propose TOPSIS method for MADM, where the rating values of the attributes are interval trapezoidal neutrosophic numbers and the weight information of the attributes are known or partially known or completely unknown. We develop optimization models to obtain weights of the attributes with the help of maximum deviation strategy for partially known and completely unknown cases. Finally, we provide a numerical example to illustrate the proposed approach and make a comparative analysis.