An accurate prediction of software effort has been the goal of successful software project management for more than thirty years. In order to achieve this goal, many software effort estimation methods have been proposed. Unfortunately, none of these methods developed thus far have been able to offer consistent prediction accuracy in all cases. In this paper, therefore, we integrate several software effort estimation methods and assign linear weights for combinations. It is noted that the weight assignment is based on the outcome of single methods. Seven public datasets and three criteria are used to evaluate the accuracy of our combinational models. Experimental results show that the proposed combination models are a useful method for improving estimation accuracy.