欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
数据完备性是基于数据驱动的分析方法的一个重要的前提，不完备的数据意味着很可能会丢失重要的判别信息，从而影响分析结果的准确性。针对现实情况下客户数据特征不同以及数据的不完备性对传统的基于数据驱动的分析方法的不利影响，本文将机器学习领域中迁移分类的方法应用于客户流失预测，通过谱特征排列（spectralfeaturealignment，SFA）实现了跨领域属性的近似统一，并利用直推式支持向量机（transductivesupportvectormachines，TSVM）对客户数据进行分类从而识别忠诚客户和流失客户，使得预测模型的性能显著提高。在最后的实验部分，使用两个不同的数据集进行的迁移分类结果证明了本文提出模型的有效性。
英文摘要：
Data integrity is one of the most important preconditions of the traditional data-driven analysis method, because the disintegrate data often signifies the loss of important discriminative information and thus negatively affects the accuracy of the result. Based on this negative effect caused by different distributions under different circumstances and disintegrate data, this paper introduces the concept and approaches of transfer classification, and applies them in the prediction and analysis of customer chum. We approximately unify the cross-domain attributes via SFA （ Spectral Feature Alignment） method, and recognize the loyal customers and un-loyal ones through TSVM. The conclusions prove that transfer classification model can significantly improve the predictive capability. At the end of this paper, an experiment of transfer classification on two different data sets proves the effectivity of the proposed model.
同期刊论文项目
同项目期刊论文
Perturbation analysis for the normalized Laplacian matrices in the multiway spectral clustering meth
A Method for Multi-granularity Uncertain Linguistic Group Decision Making with Incomplete Weight Inf
Parallel identical machines scheduling with deteriorating jobs and total absolute differences penalt
Evaluation Knowledge management capacity of organization: A proportional 2-tuple linguistic represen
Special Knowledge Sharing Incentive Mechanism for Two Clients with Complementary Knowledge: A Princi
期刊信息
位置：