In this work a comprehensive multi-step machine learning data mining and data visualization framework is introduced. The different steps of the approach are: preprocessing, clustering, and visualization. A preprocessing based on a Robust Principal Component Analysis Neural Network for feature extraction of unevenly sampled data is used. Then a Probabilistic Principal Surfaces approach combined with an agglomerative procedure based on Fisher’s and Negentropy information is applied for clustering and labeling purposes. Furthermore, a Multi-Dimensional Scaling approach for a 2-dimensional data visualization of the clustered and labeled data is used. The method, which provides a user-friendly visualization interface in both 2 and 3 dimensions, can work on noisy data with missing points, and represents an automatic procedure to get, with no a priori assumptions, the number of clusters present in the data. Analysis and identification of genes periodically expressed in a human cancer cell line (HeLa) using cDNA microarrays is carried out as test case.