Predictive models using big data

宏利目前拥有大量客户数据，包括过往客户行为、客户人口统计数据及渠道交易数据。他们希望建立来自宏利所有部门的多个数据科学实验。获取和分析与其保单持有人相关的结构化数据以及来自各种来源（包括社交媒体）的非结构化数据，可以帮助宏利评估为特定人员投保的风险，并相应地设定保单的保费。此外，大数据和分析还影响客户洞察、索赔管理和风险管理。随着客户喜好的改变，宏利必须开发更简单和更透明的产品。他们可以分析大数据以更好地预测客户行为，从而提高客户保留率并提高盈利能力。 宏利可以使用预测分析来应对欺诈性索赔和损失的增加。我们建议使用分类算法，如深度神经网络，朴素贝叶斯，SVM等，根据先前识别的关键属性（或简化特征）自动分类过去客户行为的根本原因。该等项目的目标是利用Hadoop、MongoDB等大数据工具，从宏利内部不同来源及格式提取数据。

Manulife currently has vast amount of customer data including past customer behaviour, customer demographics, and channel based transactions data. They would like to build multiple data science experiments coming from all divisions within Manulife. Capturing and analyzing structured data associated with their policyholders and unstructured data from various sources-including social media-can help Manulife evaluate the risks of insuring a particular person and set the premium for the policy accordingly. In addition, Big Data and analytics also affects customer insights, claims management, and risk management. As customer preferences change, Manulife has to develop simpler and more transparent products. They can analyze Big Data to better predict customer behavior so they can improve customer retention and become more profitable. Manulife can use predictive analytics to address the increase in fraudulent claims and losses.We propose the use of classification algorithms, such as deep neural network, Naïve Bayes, SVM, etc., on the key attributes (or reduced features) identified earlier to automatically classify the root causes of a past customer behavior. The objectives of the projects are to extract data from different sources and formats within Manulife by using big data tools such as Hadoop, MongoDB, etc.