My Projects | 我的项目

Here are brief introdutions for each project I did and links of the full reports are provided.

Summary of projects｜项目目录

  1. NetEase Cloud Music Case Study 网易云音乐用户行为分析和推荐系统算法研究
  2. Google Ads Campaign 谷歌关键词广告投放商业实践与分析
  3. Google Antitrust Case 谷歌反垄断案例分析
  4. RFM Segmentation and Marketing Effectiveness Evaluation RFM 细分模型和多渠道营销效率分析

NetEase Cloud Music Case Study | 网易云音乐用户行为分析和推荐系统算法研究

Report I - User Analysis 报告1 - 用户分析
Report II - Prediction Model 报告2 - 预测模型
Report III - Recommender System 报告3 - 推荐系统

  • Major tasks:
    1. Analysis on user characteristics and preferences to differentiate inactive users from active users.
    2. Building prediction models, including Logistic Regression, SVM, Random Forest and Neural Networks, to identify user activity based on initial actions.
    3. Summarizing commonly used recommendation algorithms; Using K-prototypes for user clustering analysis, building recommendation systems based on the collaborative filtering algorithm(including UserCF, ItemCF, LFM) for each cluster, and discussing their advantages and disadvantages.
  • Data Volume:
    over 57 million impression data and 2 million user data

  • Tools:
    PostgreSQL, Python(Pandas, Numpy, Matplotlib, Scikit-learn)

  • 主要内容：
    1. 分析用户特征和偏好，区分活跃用户和非活跃用户。
    2. 建立预测模型，根据用户初始行为识别用户活跃度，包括逻辑回归模型、支持向量机、随机森林和神经网络。
    3. 常用推荐算法总结；使用K-prototypes对用户进行聚类分析，对每个聚类建立基于协同过滤算法的推荐系统（UserCF、ItemCF、LFM），并讨论其优缺点。
  • 数据量：
    超过5700万条视频曝光数据和200万用户数据。
  • 工具：
    PostgreSQL, Python(Pandas, Numpy, Matplotlib, Scikit-learn)。

Back to catalog 返回目录

Google Ads Campaign | 谷歌关键词广告投放商业实践与分析

Report I - A/B Test 报告1 - A/B测试
Report II - Keyword Advertising 报告2 - 关键词广告

  • Major tasks:
    1. Design and launch a website for product and service marketing by Weebly (check it), and use tag manager to monitor the site via Google Analytics.
    2. Conduct an A/B test to adjust website design and eventually get higher CTR.
    3. Make Google Paid/Keyword Advertising through the process of conducting competitor analysis, designing our advertisement, making bidding strategy and publishing our ads online.
    4. Track and adjust the advertising plan according to the effectiveness evaluatation based on the real data from Google Analytics.
  • Data Volume:
    Daily impression data and user data on our website in March and April 2021

  • Tools:
    Google Ads, Google Analytics, Excel, Python(Pandas, Numpy, Matplotlib)

  • 主要内容：
    1. 通过Weebly自己设计和发布了一个网页用于产品和服务宣传(查看网页)，使用Google Analytics中的标记管理器来监测网页实时数据。
    2. 通过A/B测试确定和调整网页设置，从而实现更高的CTR。
    3. 完成谷歌广告投放，包括竞争对手分析、广告设计、制定广告投标价格策略和广告发布。
    4. 通过对Google Analytics中的实时数据进行追踪和分析，评估广告投放的效率，调整广告策略。
  • 数据量：
    2021年3月和4月的网站每日广告曝光数据和用户数据。
  • 工具：
    Google Ads, Google Analytics, Excel, Python(Pandas, Numpy, Matplotlib)。

Back to catalog 返回目录

Google Antitrust Case | 谷歌反垄断案例分析

Full Report 完整报告

This project helps to gain a thorough understanding of digital marketing, the peculiar network and platform power inherent in this kind of marketing and the structural issues that are causing so many problems to so many businesses.

  • Major tasks:
    1. Literature review and research on the background, including overview of Google and its antitrust cases.
    2. Analysis on Google’s dominance, cross-market strengths and network power in different markets, such as mobile operating system market and Android’s default applications.
    3. Summarize the learning from other antitrust cases and emerging enforcement on data transparency and competition for fintech.
    4. Provide proposals and recommendations to Google, regulator, industry and consumer, including break-up of Google mobile access product/services, open tech standard and common data storage platform.

这个项目有助于深入了解数字营销，特别是其固有且独特的网络和平台力量，以及给众多企业带来诸多麻烦的结构性问题。

  • 主要内容：
    1. 文献收集和背景研究，包括谷歌及其反垄断案例综述。
    2. 分析谷歌在不同市场的主导地位以及它的跨市场优势和网络实力，比如在移动操作系统市场和Android的默认应用市场。
    3. 总结从其他反垄断案件、数据透明度和金融科技竞争方面的学习。
    4. 向谷歌、监管机构、行业和消费者提供对策和建议，包括谷歌移动系统接入产品和服务的拆分、开放技术标准和通用数据存储平台等。

Back to catalog 返回目录

RFM Segmentation and Marketing Effectiveness Evaluation | RFM 细分模型和多渠道营销效率分析

Full Report 完整报告

In this project, we analyse the dataset representing multi-channel sales campaigns and sales of a gifts company to gain insights into the effectiveness of the various direct-marketing channels – specifically catalog mailing v.s. email.

  • Major tasks:
    1. Proposing two interesting business analytical questions and study the data to give our answers and analysis.
    2. Segmenting the dataset using RFM dimensions with 5 quantiles. Estimating response rates for each RFM cell and made a decision on how many to mail. Validating our choice on a sample mailing to get ROI of the campaign.
    3. Improving the RFM model by choosing another dimension to add to target. Justifying our choice based on business/common-sense and data. Comparing the results.
  • Data Volume:
    over 100,000 customer records, 3.3 million marketing contact records, 240,000 order records and 610,000 line item records.
  • Tools:
    PostgreSQL, Python(Pandas, Numpy, Matplotlib), Excel

在这个项目中，我们分析了包含多渠道营销活动和某礼品公司销售情况的数据集，以深入了解不同营销渠道的有效性——特别是邮寄商品目录和电子邮件。

  • 主要内容：
    1. 提出两个有趣的商业分析问题，通过分析数据集回答问题并给出我们的分析。
    2. 使用RFM细分模型对用户进行划分，模拟营销情景，预测每个类别用户的回复率，找出使利润和 ROI 最优的目标营销客户群体。
    3. 优化 RFM 模型，从商业视角附加数据支撑增加新的细分维度，比较不同模型的营销效率。
  • 数据量：
    超过 10 万条用户信息数据、330 万条营销数据、24 万条订单数据、61 万条商品数据
  • 工具：
    PostgreSQL, Python(Pandas, Numpy, Matplotlib), Excel

Back to catalog 返回目录