学术报告简介
1. 袁先智教授 (同济大学 & BBD公司)
题目: 大数据框架下的金融解决方案：信用风险评理论与实践结合的突破
摘要: 在大数据框架下，基于企业全息画像框架下的公司商务行为的基因（DNA）刻画， 金融网络结构的KPI赋予的金融信息， 以及对应的“中心点， 桥梁点， 码头点”等核心概念，我将和大家分享如何利用线上和线下信息对小微企业进行有效的信用评级，从而全面解决困扰全球金融界对无抵押、 无担保情况下金融机构对小微企业进行贷款前， 中和后面动态分析管理的业界问题， 希望这能为中国的小微金融服务带来革命性的变化。
以“构建基于大数据的小微企业信用风险评级体系”作为大数据应用的切入点，构建了大数据平台和对应的小微企业信用评估体系， 是具有原创性和自主知识产权的业界新产品。在具体的实践中，通过增加对小微企业评价的维度，开拓不同的数据来源，形成三位一体的“大数据”体系，即整合“企业资产数据”、“政府数据”、“企业行为数据”；通过对小微企业信贷历史样本的分析研究，实现小微企业的画像，开发出小微企业的风控模型。该项目旨在创新小微企业的风控措施，突破传统信贷文化，改造小微企业信用评价模型和信贷业务流程，为小微企业设定更加合理的授信准入门槛，有力推进我国普惠小微企业的金融服务， 为建立广泛意义下的消费金融打好基础。
本结果是我们在大数据框架下的信用风险评理论与实践结合的原创性突破。
2. 朱利平教授 (中国人民大学统计与大数据研究院)
题目: Measuring and Testing for Interval Quantile Dependence
摘要: We introduce the notion of interval quantile independence which generalizes
the notions of statistical independence and quantile independence. We suggest an index to measure and test departure from interval quantile independence. The proposed index is invariant to monotone transformations, nonnegative and equals zero if and only if the interval quantile independence holds true. We suggest a moment estimate to implement the test. The resultant estimator is root-$n$-consistent if the index is positive and $n$-consistent otherwise, leading to a consistent test of interval quantile independence. The asymptotic distribution of the moment estimator is free of parent
distribution, which facilitates to decide the critical values for tests of interval quantile independence.
3. 王学钦教授 (中山大学数学学院)
题目: Ball Covariance: a Generic Measure of Dependence in Banach Space
摘要: Technological advances in science and engineering have led to the routine
collection of large and complex data objects, while dependence structure among these objects is often of great interest. However, many existing measures of dependence such as correlation coefficients and mutual information cannot characterize complex relationships among those complex objects (e.g., different brain subcortical structures), which often reside in some Banach spaces, due to their restrictive Hilbert space assumption. To overcome the limitations of these existing measures, we propose Ball Covariance as a generic measure of the dependence between two random objects in possibly different Banach spaces. Our Ball Covariance possesses the following attractive properties: (i) It is model-free and makes less restrictive data assumptions; (ii) It is nonnegative and equal to zero if and only if two random objects are independent in a separable Banach space under some mild conditions; (iii) Intuitive empirical Ball Covariance is feasible and can be used as a test statistic of statistical independence. We use both theoretical and numerical results to reveal the potential power of the Ball Covariance in detecting dependence. More importantly, we analyze a real example to demonstrate the practicability of Ball Covariance in the complex dependence detection.
4. 田国梁教授 (南方科技大学数学系统计学)
题目: Some Recent Advances in Computational Statistics (计算统计中某些算法
的新进展)
摘要: This talk consists of five parts: (1) We present a literature review on the
advantages and disadvantages of the Newton-Raphson algorithm and its many variants such as the Fisher scoring algorithm, Gauss-Newton method, Quasi-Newton method, which are commonly used in computational statistics. (2) By exploring the merits and defects of the EM-type algorithms (i.e., Q-based EM) for dealing with incomplete data, we introduce a class of new SR-based EM algorithms and their applications. (3) We introduce a recent advance on MM algorithms; i.e., the assembly and decomposition (AD) approach. (4) We discuss new advances on the acceptance-rejection method. (5)
Finally, we introduce a new statistical algorithm; i.e., the mode-sharing algorithm.
5. 郑术蓉教授 (东北师范大学数学与统计学院)
题目: Testing the Linear Structure of High Dimensional Correlation Matrix
摘要: The linear structure of correlation matrix includes the identity matrix,
banded matrix and equi-correlation structure etc.
Many people have studied testing the identity of high dimensional correlation matrix. But few literatures are found for testing the general high dimensional correlation matrix.
This paper will develop a method to test the linear structure of high dimensional correlation matrix. Simulation studies are conducted for the illustration of our testing method.
A real example is also given.
6. 傅博教授 (复旦大学大数据学院)
题目: Causal Analysis of Administrative Data and Methodological Challenges (行
政数据的统计因果推断及其方法挑战)
摘要: Combination therapy is the hallmark of therapies for cancer, viral or
microbial infections, hypertension and other diseases involving complex biological networks. We discuss the impact of integrated genomics Big Data on drug development using this approach. Then we focus on the modeling approach to reduce the problem into one that laboratory can handle. Because drug-effect is dose-dependent, multiple doses of an individual drug need to be examined, yielding rapidly increasing number of combinations.Finding proper combinations for drug developments is a challenging high dimensional statistical modeling problem. We proposed a novel two-stage procedure starting with an initial selection by utilizing an in silico model built upon experimental data of single drugs to obtain global sensitivities indices using functional principal components and currentsystem biology information to obtain maximum likelihood estimate. Simulations show that the procedure performs well in identifying promising drug combinations. The research is in collaboration with Hengzhen Huang and Hongbin Fang.
7. 王启华研究员(中国科学院数学与系统科学研究院应用数学研究所)
学术报告简介
