基于免疫粒子群优化的最小属性约简算法  

Minimal attribute reduction algorithm based on particle swarm optimization with immunity

作　　者：廖建坤[1] 叶东毅[1] 

机构地区：[1]福州大学数学与计算机科学学院,福建福州350002

出　　处：《计算机应用》2007年第3期550-552,555,共4页journal of Computer Applications

基　　金：福建省自然科学基金项目(2006J0029);福建省高新技术开发重点项目(2005H028)

摘　　要：把求决策表最小属性约简问题归结为一个0-1组合优化问题,为该问题定义了合理的粒子适应度函数,提出了一种把免疫接种、免疫测试机制与二进制粒子群算法相结合的混合算法用于求解该问题。对UCI数据表的实验结果表明该算法在获得更优解的同时,仍具有较快的运算速度。多种算法的比较结果表明了该算法的有效性和可行性。The problem of finding a minimum attribute reduction of a decision table was converted into a 0-1 combinatorial optimization problem. A proper fitness function of a particle for this problem was defined. A binary particle swarm optimization algorithm with immunity （BPSOI） was proposed to deal with the problem. Experimental results based on UCI repositories show that the proposed algorithm can quickly converge to a better solution in less runtime and within fewer generations. Its effectiveness and feasibility are also verified in the comparison with several other algorithms.

关 键 词：粗糙集 属性约简 二进制粒子群优化 免疫 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论] TP182[自动化与计算机技术—计算机科学与技术]