This paper presents a practical knowledge discovery approach to software quality and resource allocation that incorporates recent advances in rough set theory, parameterized approximation spaces and rough neural computing. In addition, this research utilizes the results of recent studies of software quality measurement and prediction. A software quality measure quantifies the extent to which some specific attribute is present in a system. Such measurements are considered in the context of rough sets. . It has been shown rough sets work well in coping with the uncertainty in making decisions based on software engineering data. The thrust of this research is to provide a framework for making resource allocation decisions based on evaluation of various measurements of the complexity of software. Knowledge about software quality is gained during preprocessing during which, software measurements are analyzed using discretization techniques, genetic algorithms in deriving reducts, and in the derivation of training and testing sets, especially in the context of the Rough Sets Exploration System (RSES) developed by the logic group at the Institute of Mathematics at Warsaw University. The results of preprocessing provide a basis for rough neural computing and resource allocation decisions. Experiments have shown that both RSES and rough neural network models are effective in classifying software modules. The contribution of this paper is a presentation of a rough set based framework for making decisions about software quality and resource allocation.