The following projects were completed during the MS-Business Analytisc program at Carlson School of Management.
This project explores Hillary Clinton's released email contents and contacts.
Using topic modeling, IBM's Waston API and R, the exploration consists of three parts: a)How topics trending overtime b)How sentiment of emails change on weekdays c)Frequent contacts of Hillary Clinton and the interaction.
- As a result of topic modeling using LDA, 7 topics were identified, and their trend overtime is plotted here.
- I detected the most frequent contacts Hillary emailed with, and visualized the interactions with this chord gram.
- To do sentiment analysis, IBM's Waston API was called. It turns out that on Monday, emails tend to be more negative than positive, whereas on Friday, positive emails increase.
This project optimized the bidding strategy on Search Engine using the data from AirFrance Case Study.
Using Excel, JMP, I built OLS models for each keyword on each advertisement platform to predict a reasonable rank as target position. The KPIs used to measure advertisement and bidding strategy effect include Click Through Rate, Conversion Rate, Net Revenue and ROA. As a result, according to the analysis, the ROA increased 40%.
This project analysis the change of English word usage following the method in the research paper Quantitative Analysis of Culture Using Millions of Digitized Books.
I extracted Google n-gram data on AWS public S3 bucket using Hive, to analyze the linguistic change of English words. I came to the same conclusion as the research paper, that the process of English verb past tense regularity is ongoing.
I built a data visualization page to show intuitivelly how the Chinese Tourism Industry developed over years (the contents is in Chinese). The charts were written using D3.js package while the main frame of the page was in HTML5 and CSS.
This project redesigned a normalized database into data warehouse.
I consolidated 11 highly normalized data files into a data warehouse with star schema structure. After designing the ER Model, I used SQL to creat tables and their relationship, then performed data ETL task to load the data from source files to SQL Server.
The project statistically evaluates the likelihood impact of unemployment rate on the campus crime rates.
I adopted a probabilistic method and built hypothesis testings to measure the effect of unemployment rate of the community on campus crime rate. The analysis was done using Python.