In this paper, we introduce an approach to predict bad smells from software design model using machine learning techniques. We use number of bad-smells and software design model metrics as data sets. The data sets are used as a learning set and testing set to predict bad-smells with 7 machine learning algorithms. We compare the performance of each machine learning algorithm by using several statistical significance tests such as prediction accuracy, hypothesis test, sensitivity and specificity, and predictive value of tests.