多智能体系统(MAS)在企业内部信用评估中的应用研究

研究、实现了企业内部信用评估多智能体系统IECRMAS（Inner-Enterprise Credit Rating System）.与传统决策支持系统不同的是,IECRMAS不仅关注客体信息利用,而且研究了如何提高评估主体和评估团队的理性,以及促进共享团队形成的评估机制.基于MAS技术的信用评估系统具有分布式、半自治的、可观测的、可控的、可动态配置等特点.该系统的主要创新之处在于;（1）基于企业信用交易过程的信用评估流程规范化;（2）面向评估任务的评估主体动态分组和共享团队培育机制;（3）以Agent辅助评估师个人构成理性评估个体.IECRMAS在多家单位进行了应用,取得了良好的社会效益和经济效益.     

Fragmentation Mechanism of Trans—α—Aryl—β—enamino Esters

Electron impact-induced fragmentation mechanism of Trans-α-Aryl-β-enamino esters were investigated using mass-analyzed ion kinetic energy (MIKE) spectrometry and high resolution accurate mass data It was found that the main characteristic fragmentations of compounds studied were:an odd electron ion M^ -EtOH was formed by losing a neutral molecule of ethanol;and the skeletal rearrangements took place;and the ring opening reaction happened after losing a carbon monoxide;and the typical McLafferty rearrangement underwent in ester group.The cycliztion reation caused by losing neutral molecule of TsNH2 due to the ortho-effects of substituted group of gromatic ring was also observed.     

用LDA Boosting算法进行客户流失预测

本文提出一种LDA boost(Linear Discriminant Analysis boost)分类方法,该算法能有效利用样本的所有特征,并且能够从高维特征空间里提取并组合优化出最具有判别能力的低维特征,使得样本类间离散度和类内离散度的比值最大,从而不会产生过度学习,大大提高算法效率。该算法有效性在某商业银行的客户流失预测过程的真实数据集中得到了验证。与其他同类算法,如人工神经网络、决策树、支持向量机等运算结果相比,该方法可以显著提高运算精度。同时,LDAboosting与其他boosting算法相比,也具有显著的优越性。