人工智能的不平衡数据集异常点抽样算法

Algorithm for Sampling Outliers in Imbalanced Data Sets of Artificial Intelligence

针对传统抽样算法在不平衡数据集抽样时存在抽样准确度较低,耗时较长等问题,提出人工智能的不平衡数据集异常点抽样算法。结合人工智能技术,修整弱分类器的不足;通过架构预测集合及其相应权值,获取抽样算法的决策表达式;根据合并得到的损失函数,获取逐点损失函数表达式;将损失函数通过二阶泰勒展开,获取理想化预设条件与权值更新结果;对初始种群实施遗传操作后,将抽样结果作为样本集,利用聚类算法得到分类中心代替原始种群;依据最佳异常点抽样结果完成自适应调整,并采用编码矩阵对所有聚类中心进行重构,通过类内与类间距离的求解,完成异常点的交叉操作,使异常点抽样算法得以实现。仿真结果表明,所提算法具有较好的自适应性与较高的异常点抽样精准度,有效降低了抽样的复杂度。

In order to solve the problems of low accuracy and high time consumption in traditional sampling algorithm,an algorithm for sampling abnormal points in imbalanced data sets based on artificial intelligence was put forward.Combined with artificial intelligence technology,we covered the shortage the shortcomings of weak classifiers.After that,we obtained the decision-making expression of sampling algorithm by constructing the prediction sets and weights.According to the loss function,we obtained the expression of point-by-point loss function.Then,we expanded the loss function by second-order Taylor expansion and thus to obtain the ideal preset condition and weight update result.After genetic operation for the initial population,we took the sampling results as the sample sets,and used the clustering algorithm to find out the classification center instead of the original population.Thus,we finished the adaptive adjustment by the best outliers sampling result.Meanwhile,we used the coding matrix to reconstruct all the clustering centers.Finally,the crossover operation of outliers was completed through the solution of intra-class distance and inter-class distance,so that the outliers sampling algorithm was achieved.Simulation results show that the proposed algorithm has good adaptability and high sampling accuracy of abnormal point.Meanwhile,this algorithm effectively reduces the sampling complexity.