During the process of construction investment decision, quite a lot of indexes may be involved, but not all indexes are indispensable to any projects. There are many redundant indexes. In order to improve the efficiency and reliability for evaluation, this article first introduces data mining technique and rough sets, and then applies the rough sets conducting attribute reduction to evaluation indexes. So the major indexes influencing construction investment decision can be found out. After then entropy method is introduced for deciding weights of these indexes. Lastly, on this basis scheme optimization would be carried on, and a satisfactory project construction investment can be gotten.