基于线性LTSA算法维数约减的软件缺陷预测研究

软件缺陷预测是软件开发过程中的一项重要技术,针对软件缺陷数据集的高维、小采样造成预测精度下降的问题,采用线性局部切空间排列算法对数据集降维处理,选用支持向量机作为基础分类器进行二值分类,建立软件缺陷预测模型,采用二维混淆矩阵评价模型的预测精度.实验结果表明,与其他模型相比,该模型可用较少的邻域点约简至更低的维度,不需要重新学习样本空间的流行几何结构,直接映射新的样本点,且预测时间耗费成本由13. 726 9 s降低至6. 217 s,给定参数区间寻优时间耗费由267. 442 1 s降低至165. 98 s,有效提高了软件缺陷预测的效率.

A software defect prediction model based on the LTSA dimension reduction algorithm

Software defect prediction is an important technology in the software development process. For a high-dimensional and small sampling of software defect data sets,the prediction accuracy is degraded. A linear partial-cutting spatial arrangement algorithm is used to reduce the dimensionality of data sets. The vector machine is used as the basic classifier for binary classification,and the software defect prediction model is established. The prediction accuracy of the model is evaluated by the two-dimensional confusion matrix. The experimental results show that compared with other models,this model can be reduced to a lower dimension with fewer neighborhood points,without re-learning the popular geometry of the sample space,directly mapping new sample points,and predicting the time cost. From 13. 7269 seconds to 6. 2178 seconds,the time cost for parameter optimization is reduced from267. 4421 seconds to 165. 98 seconds,effectively improving the efficiency of software defect prediction.