论文部分内容阅读
提出一种建立在局部最优基础上的动态集成选择算法,并从理论上对算法进行了分析.该算法首先在多个局部特征空间上构造最优集成,然后使用动态集成选择技术对未知样本进行识别.局部空间上的集成构造问题被转换为一个单目标优化问题,并使用多种群遗传算法进行了求解.基于UCI数据集的实验表明,相对于现有的动态分类器选择算法和动态集成选择算法,新算法能够取得更高的识别率.同时,相对于现有的动态集成选择算法,新算法构造的集成规模更小,识别速度更快.
A dynamic integration selection algorithm based on local optimality is proposed and the algorithm is theoretically analyzed.The algorithm first constructs the optimal integration on several local feature spaces and then uses the dynamic integration selection technique to analyze the unknown samples The problem of integrated structure in local space is transformed into a single-objective optimization problem and solved by using multi-population genetic algorithm.Experiments based on UCI datasets show that compared with the existing dynamic classifier selection algorithm and dynamic integration Selection algorithm, the new algorithm can achieve a higher recognition rate.At the same time, compared with the existing dynamic integration selection algorithm, the new algorithm has a smaller integration scale and a faster identification speed.