Research On Software Defect Prediction Method Based On Integrated Incremental Learning

Software defects are common problems in the software development process,which may lead to system failures,functional errors and a decline in user experience.In order to improve software quality and efficiency,software defect prediction has become an important research field.The goal of software defect prediction is to analyze the characteristics and historical data of software projects,predict potential defects,and take measures to repair or improve them in advance.There are some limitations to traditional software defect prediction methods: when creating new data,it is necessary to rebuild a prediction model,which will consume time and cost,and also cause resource waste;The implicit information in the software data stream may change over time,making it difficult to accurately classify software data with concept drift.(1)A class imbalance mitigation algorithm is proposed to obtain software data stream based on time series,and combine oversampling technology with cost sensitive technology to improve the search range of prediction model for potential defect data.Through experimental comparison,it has been shown that This algorithm is effective of the algorithm and performs excellently in the recall index.(2)The selection method of base classifier for differential sexual selection is proposed.The method of input data perturbation is used to make the base classifier obtain the maximum difference,and the threshold value of the base classifier of the integrated model is set.Thus,from the perspective of integrating model classifiers that are good but different,control the scale of the integrated model can be achieved.(3)Propose a similarity matching algorithm for the base classifier,which achieves redundant processing of software data through the weight learning strategy of the base classifier and the replacement of the base model to maintain consistent algorithm scale.Use the similarity matching strategy to assign weights to the base classifier to obtain the optimal classification results.In summary,the incremental learning method is applied to software defect prediction technology to design a software defect prediction model that can handle realtime software data adaptively,improve generalization ability,reduce system resource consumption,and improve detection efficiency.Figure 15;Table 12;Reference 56...