Wavelet-based attribute noise detection

Data quality is a critically important issue when inferring knowledge from datasets. The identification of noisy attributes which can easily corrupt and curtail valuable knowledge and information from a dataset, can be invaluable to analysts. We present a novel detection method to identify noisy attributes in datasets of software metrics using multi-resolution transformations based on Discrete Wavelets. The proposed method has been applied to supervised datasets of scientific full-scale data from NASA's Software Metric Data Program (MDP). Empirical results have been favorably compared to those obtained from the robust Pairwise Attribute Noise Detection Algorithm (PANDA) using the same datasets. These results were verified with several case studies that included injecting known simulated noise into a specific attribute of a dataset with no Class noise.