The present project proposes several research directions towards using machine learning techniques in the resolution of practical problems in software engineering. Since maintenance and software evolution problems are of utmost importance to the modern programmer, there exists a grown interest in the automation of as many processes in the lifecycle of software as possible, as well as the development of adequate mathematical models. The present project is interdisciplinary, contributing to the fundamental and applicative research in the domains of Applied Computational Intelligence and Software Engineering and offering solutions based on machine learning for the most important problems of Software Engineering. The problems tackled have a major practical importance since software developers face them daily, therefore the development of automated techniques which offer solutions to these problems would lead to more correct and error-free software. Since it is very hard to identify direct solutions, classification methods based on machine learning are extremely useful in solving the aforementioned problems. The theoretical results will be used to develop an integrated software system AMEL which will include all the computational techniques developed and will assist the developers in the maintenance and evolution stages of the software lifecycle.