This paper discusses three-way decision problems on intuitionistic fuzzy information systems. Firstly, a similarity degree function is defined, which is composed of the fuzzy factor, the mean factor and the probability factor, and then a three-way decision model on intuitionistic fuzzy information systems is established, and it is pointed out that the model theoretically unifies all kinds of two-universe models. Secondly, considering the different evaluation values of objects in the universe, this paper proposes a partition measure based on evaluation values: the weighted information entropy, and it is proved that the finer the partition is, the greater the weighted information entropy is. Finally, based on the property of the weighted information entropy, a reasonable explanation of the optimal tri-partition is given, and a new thresholds solution method is proposed.