Early Software Quality Prediction Based on Software Requirements Specification Using Fuzzy Inference System

Software Requirements Specification (SRS) is the key fundamental document formally listing down the customer expectations from the software to be built. Any weakness or fault injected at this stage in the requirements is expected to ripple towards the following phases of software development life cycle resulting in development of a software system of poor quality. Software quality prediction promises to raise alarms about the quality of the end product at earlier stages. It becomes more challenging as we move earlier in stages because of limited information is available at earlier stages. Therefore little effort has been put in literature to predict software quality at SRS stage. This position paper presents a novel approach of prediction of software quality using SRS. SRS document is converted into a graph and different parameters including readability index, complexity, size and an estimation of coupling are extracted. These parameters are fed into a Fuzzy Inferencing System (FIS) to predict the quality of the end product. The proposed model has been evaluated on a sample of student projects and has shown reasonable performance.