Inconsistent outcomes from rival prediction algorithms have been discovered to be a concern in software engineering. The complexity and quality of software requirements have significantly changed in recent years, and customers have significantly raised their expectations in terms of the cost turnaround time, and quality of software solutions. Nevertheless, accurate prediction is essential. Additionally, these factors could directly clash with one another, making their resolution dependent on the efficient creation of software utilizing reliable software engineering techniques. In this thesis, two models were created by integrating the honey badger algorithm (HBA), a recentpopulation-based strategy, with the backpropagation neural network (BPNN) in order to improve prediction accuracy and overcome BPNN's drawbacks. They were applied on 18 software engineering datasets The first proposed method is applying the original HBA with BPNN, and the second one is called HBA-OBL, which combines the HBA with opposition-based learning (OBL) to enhance the global search capabilities of HBA. The two proposed approaches, HBA and HBA-OBL, aim to produce the best possible weight for BPNN that can be input into the prediction model in order to solve software fault prediction problems.Based on the results of accuracy, data distribution, convergence speed and the significant test, the hybrid approach HBA-OBL has the strength to enhance the weights of the BPNN rather than the original HBA method. As well as, the suggested methods' performance efficiency has been assessed and compared to recent approach, namely SSA-BPNN. In terms of accuracy rate, the proposed HBA-OBL approach was shown to be significantly superior. It surpassed the SSA-BPNN in 13 out of 18 datasets with average accuracy of 0.90.