Approximate Model Checking based on Deep Forest

Abstract

Some classical Machine Learning (ML) algorithms have been applied to predict model checking results, while a data set contains several thousands of samples. However, the power of prediction will reduce sharply when the scale of dataset is bigger. To this end, some Deep Learning (DL) algorithms are employed in this study. First, a part of samples are inputted to a DL algorithm. Second, the obtained DL model can be used to predict model checking results. Our experiments demonstrate that Deep Forest (DF) has the better performance when one million samples are used, compared with the classical ML algorithms and the deep learning based on deep neural network.