1. Scratched datasets from webpages by parsing the html source code, cleansed and sorted data in R.2. Operated tests of stationary and randomness to analyze properties of data, constructed ARIMA models to predict future rates of return, realized the numerical results for 10 days in SAS with error less than 0.5%.3. Trained a digit-classification machine learning model based on multi-layer neural networks with convolutional layers to recognize the credit card number in an image of credit card in Python with accuracy more than 98%.