Linked data has been officially approved as one of the best useful sources for background information that publishing and connecting structured data on the Web. Therefore, Machine Learning community should respond with a clear package of methods and best practices to bring this type of data into the field. Relatively little attention has been paid in the literature to the possible union of both linked data and Machine Learning. In order to overcome this carelessness and achieve this joining, we should focus little more on RDF data. The main purpose of this article, is to present a typical Machine Learning Pipeline on RDF data and an improved version of RDF clustering algorithm based on Candidate Description by using different similarity measures and agglomerative hierarchical techniques.