My machine-learning projects (notebooks)
In this repository I will be uploading my machine learning projects (notebooks), the goal is to demonstrate and put my academic knowledge to solve real world problems. I have uploaded two notebooks that demonstrate regression and classification problems. The datasets and problems are from two of the recent Microsoft Capstone projects that I successfully completed.
First notebook (PredictStudentLoanRepayRate-XGBRegressor.ipynb):
The goal of the analysis is to predict the repayment rate for loans given to students at United States institutions of higher education.
The analysis carried out answers the question: just how good of an investment is higher education by predicting the percent of students that are actively repaying their loans within three years of graduating and helping students understand which institutions are good investments, and which ones leave them in debt.
A student's ability to repay loans depends on the job they get after graduation. This in turn depends on the school they went to, the degree they earned, the area they live in, their family circumstances, and how much money they borrowed. The analysis makes these predictions for each institution using information about the degrees the school offers, the financial makeup of the student population, the academic merits of the school, the graduation rates, and additional demographic information.
After exploring the data by calculating summary and descriptive statistics, and by creating visualizations of the data, several potential relationships between student loan characteristics and repayment rate were identified. After exploring the data, a regression model to predict repayment rate from its features was created. By modeling what percent of students repay their loans, we can better understand the properties of schools that make them better or worst investments.
Second notebook (NepalEarthquake-multi-class-classification.ipynb):
In April, 2015, a 7.8 magnitude earthquake with an epicenter in the Gorkha District of Nepal devastated the surrounding area, resulting in almost 9,000 deaths and 22,000 injuries. Some of these casualties happened in buildings that collapsed in the earthquake, and may have been preventable if they had withstood the initial ground motion or resulting aftershocks.
Using data on buildings in the affected area and how they were impacted by the earthquake, we'd like to model risk of damage. Accurate models of this kind help first responders plan their initial triage after an earthquake, and help governments direct scarce resources which may be available to mitigate risk before another earthquake happens.
The goal is to predict buildings at risk of earthquake damage. Based on aspects of building location and construction, my goal is to predict the level of damage to buildings caused by the 2015 Gorkha earthquake in Nepal.