Software defect prediction (SDP) is beneficial for ensuring the quality of software products by detecting modules with high chances of defects in the course of development. Using machine learning classification techniques, SDP assists organizations in enhancing the efficiency and cost-effectiveness of testing processes. However, despite significant advancements, some constraints still exist that prevent the practical application of these models. Most current approaches provide binary predictions, merely indicating whether a module is defective while failing to deliver more detailed insights such as the status of the defect, priority, or root cause (multioutput classification). This lack of granularity limits the actionable value of predictions for developers who need more information to prioritize and address defects effectively. This survey studies the effectiveness of various datasets in the field of research in SDP and the potential in terms of predicting software faults using different machine learning, deep learning, and soft computing techniques while contemplating the issues of transparency and operational ability over different classes of software projects.