Abstract Wellbore stability is one determining factor to the success of drilling operations. A multitude of wellbore stability issues may still be encountered, despite all previous knowledge and experience, when drilling a certain formation. However, the use of the large amount of data collected during drilling can provide valuable insight. This collected data, in combination with artificial intelligence models, can be used to predict wellbore stability in new drilling operations. Because wellbore stability is such a common and expensive issue to the industry, many papers in the literature have presented a multitude of techniques and algorithms to predict wellbore stability. The predictions made in the literature have all provided relatively good accuracy; the most common methods that have been presented include the use of multivariate statistics, linear methods, and artificial neural networks. This paper presents two approaches to solve the wellbore instability prediction problem. First, principal component analysis (PCA) and linear discriminant analysis (LDA) are used in an attempt to reduce the dimensionality of the dataset. The results indicate that the dataset dimensionality could not be reduced. Second, nonlinear artificial intelligence models were implemented, including the artificial neural network (ANN), specifically the Bayesian regularization neural network (BRNN), and support vector machine (SVM). After building these two models and checking their accuracy, a new feature was implemented to filter the input data. Empirical mode decomposition (EMD) was adopted, which is commonly used in the field of signal processing to eliminate noise and variation in the data. This study evaluates the effect of EMD on the prediction and classification accuracy for wellbore stability. The study results are promising; by using EMD, accuracy was increased by 15% for both the ANN and the SVM models.