Data Project
The combination of econometrics, machine learning, and fully-connected neural netwrok deep learning toolbox enable us to figure out what happens behind the database.
U.S. Census Data Analysis
Used statistical tools in Python to process, clean, summarize, and visualize data from IPUMS America Community Survey (ACS) and derived conclusions related to the census attributes about the U.S. population.
Found the relation between local (county-level) single rate and income inequality according to the county-level demographic characteristics constructed by NumPy & Pandas data analysis tools.
Identified the international trade exposure's influence on the local unemployment rate by combining the local demographic characteristics and UN Comtrade data.
Predicted the discrete characteristics among the agents in the dataset by Machine Learning tools (Logistic and KNN) with an average accuracy rate of over 80%.
Java-Based Advanced Undergraduate Course Advising System
Used Java-based object-oriented programming method to develop a system that advised students on advanced course selection based on their attributes.
Built the machine-human interaction system to simulate the students' advising process for advanced undergraduate course registration.
Developed advanced general interface for future system improvement, combining with other higher education-related functions.
The Effect of Dam Construction on Local Labor Market, Evidence from the U.S.
Figured out the positive effect of dam construction on local manufacturing industry employment.
The dams with a power station would contribute more to the local labor market than dams without a power station.
Identified the time-fixed effect also influenced the contribution of dam construction to the local labor market. Especially during the past 20 years, building a dam would improve the local employment rate much more than previously.
The Effect of Critics' Reviews on Movies' Weekly Earnings Over Time
Used a non-linear function to estimate the impact of critics. Review overtime on movies' weekly box office earnings.
Simulated movies' weekly box office earnings using different non-linear models in Matlab.
Concluded that different scores of critics' reviews positively influenced the peak value of movies' weekly earnings and negatively influenced the decreasing velocity of movies' weekly earnings after arriving at the peak value.
One Special Mathematical Model and its Simulation of Stock-Option
Solved the Black-Scholes-Merton equation of the option price under the assumption that implied volatility linearly increased with time.
Simulated the model of stock price and stock-option price and found that the stock option price fluctuated more strongly as time went on.
Compare the real price of stock options with the estimated price using the model under different time delay situations to test the model I constructed.