This dissertation presents EasyLearn, a framework to support the development of supervised learning applications. EasyLearn dfines an intermediate layer, which is easy to configure and understand, between the application and WEKA, a machine learning framework created by the University of Waikato. All classifiers and filters implemented by WEKA can be easily encapsulated to be used by EasyLearn. EasyLearn receives as input a set of configuration files in XML format containing the definition of the processing flow to be executed, in addition to the data source to be classified, regardless of format. Its output is customizable and can be configured to produce classification accuracy reports, the classified data source, or the trained classification model. The architecture of EasyLearn was defined after a detailed analysis of the classification process, which identified a set of common activities among the three analyzed processes (learning, evaluation and classification). Through this insight and taking the object-oriented languages as inspiration, a framework was created which is able to support the classification processes and its variations, and which also allows reusing settings by implementing inheritance and polymorphism in their configuration files. This dissertation also illustrates the use of the created framework presenting a full case study about e-commerce product classification, including corpus creation, attribute engineering and result analysis.