The goal of software fault prediction (SFP) is to find faulty parts in software systems before they lead to production failures, improving software quality and saving maintenance expenses. SFP improves the testing and debugging process by using historical data to make it more focused and efficient. Class imbalance, where non-faulty instances greatly outnumber faulty ones, is a substantial obstacle to SFP research. A high number of false positives and negatives might result from this imbalance, which can negatively impact machine learning model performance. Class imbalance must be addressed in order to guarantee that the predictive models can precisely detect software errors. This research looks at the effects of class imbalance and assesses how well different machine learning models perform in the context of SFP. The results show that although the models achieve high accuracy, there are a lot of false positives and negatives due to their limited precision and recall for incorrect classifications. This emphasizes the need for better methods to deal with class imbalance and raise SFP models efficacy.