科技通报
科技通報
과기통보
BULLETIN OF SCIENCE AND TECHNOLOGY
2015年
4期
100-102
,共3页
E-Learning%信任节点%网络%个性化推荐
E-Learning%信任節點%網絡%箇性化推薦
E-Learning%신임절점%망락%개성화추천
E-Learning%node trust%network%personalized recommendation
提出一种基于E-Learning随机场学习的信任节点定位算法，采用SVM监督学习辅助策略降低新标记样例中可能的误标记样例数以减少新训练集的噪声，设计富信息策略训练学习器，基于KKT原理，把训练ELM的过程等同于解决下面的对偶优化问题。通过E-Learning随机场学习，充分利用算法对噪声具有鲁棒性，实现对了个性化推荐网络信任节点的定位算法改进。仿真结果表明，该算法能准确定位到个性化推荐网络的信任节点，为P2P商家推荐和信任评估提供数据基础，信任节点定位的准确性优越于传统算法。算法采用E-Learning随机场学习，学习过程中不需要调整大量参数，所以算法学习速度快，尤其适合处理大规模的不平衡数据集分类问题，提高了个性化推荐网络信任节点的定位准确性。
提齣一種基于E-Learning隨機場學習的信任節點定位算法，採用SVM鑑督學習輔助策略降低新標記樣例中可能的誤標記樣例數以減少新訓練集的譟聲，設計富信息策略訓練學習器，基于KKT原理，把訓練ELM的過程等同于解決下麵的對偶優化問題。通過E-Learning隨機場學習，充分利用算法對譟聲具有魯棒性，實現對瞭箇性化推薦網絡信任節點的定位算法改進。倣真結果錶明，該算法能準確定位到箇性化推薦網絡的信任節點，為P2P商傢推薦和信任評估提供數據基礎，信任節點定位的準確性優越于傳統算法。算法採用E-Learning隨機場學習，學習過程中不需要調整大量參數，所以算法學習速度快，尤其適閤處理大規模的不平衡數據集分類問題，提高瞭箇性化推薦網絡信任節點的定位準確性。
제출일충기우E-Learning수궤장학습적신임절점정위산법，채용SVM감독학습보조책략강저신표기양례중가능적오표기양례수이감소신훈련집적조성，설계부신식책략훈련학습기，기우KKT원리，파훈련ELM적과정등동우해결하면적대우우화문제。통과E-Learning수궤장학습，충분이용산법대조성구유로봉성，실현대료개성화추천망락신임절점적정위산법개진。방진결과표명，해산법능준학정위도개성화추천망락적신임절점，위P2P상가추천화신임평고제공수거기출，신임절점정위적준학성우월우전통산법。산법채용E-Learning수궤장학습，학습과정중불수요조정대량삼수，소이산법학습속도쾌，우기괄합처리대규모적불평형수거집분류문제，제고료개성화추천망락신임절점적정위준학성。
This paper proposed an algorithm for trust node localization based on E-Learning random field study, SVM super?vised learning auxiliary strategies to reduce the error marker sample hundreds of possible new labeled sample to reduce the noise of new training set with rich information, learning strategies training device, based on the principle of KKT, the pro?cess of training ELM is equivalent to the dual optimization problem solving the following. Through the E-Learning random field study, make full use of the algorithm is robust to the noise, the realization of the improved localization algorithm for personalized recommendation trust of network node. The simulation results show that, the algorithm can accurately locate the personalized recommendation trust node of the network, provide the data basis for P2P business recommendation and trust evaluation, and trust of node localization accuracy is superior to the traditional algorithm. The algorithm using the E-Learning random field of learning, the learning process does not need to adjust many parameters, so the algorithm of fast learning speed, especially suitable for processing large-scale data set of unbalanced classification problem, to improve the positioning accuracy of personalized recommendation trust of network node.
