Part time internship in Accenture Technology Lab, Beijing
Supervisor: Manager Andy Guo Ma
I developed a online demo, which could record speakers' sound in browser from several different people and be able to distinguish speakers' sound. In the project, I use openear lib to get the features of those sounds and svm machine learning algorithm to make prediction. Lamp(Linux, apache, mysql, php) technology was used to deploy the website. Other job duties during the internship include focused crawling of web pages, doing research about pedestrian and face recognition, as well as daily system maintenance.
A self initiated project - crawling transaction data from a online P2P lending website and analyzing
Being interested in the area of online P2P lending, I have crawled all the 400 thousand
transaction data pages of the private P2P lending website www.ppdai.com. I then extracted useful
information using python’s urllib and beautifulsoup library, after that I stored all the data
into a data warehouse. Further, I used R language to model the data and I ran several
different models, including SVM and Decision Tree. I used user profile and transaction
profile as features, and I use different models to predict whether a user will pay the
money back in a certain transaction. The final result has about 0.85 precision rate and
0.91 F-score. I even design a program that helps to guide the investment procedure, and
I invest some of my own money on that website according to my analysis. The annual earning rate
according to my analysis is more than 10 percent.