This is my code file for a machine learning competition at UT Austin for undergraduate BAX and MIS majors.
In this competition, all of the students received a historical data set of NBA players and statistics associated with them (such as games played, minutes played, points, etc.). Using this data set, we had to output the most profitable investment strategy to create the optimal portfolio of player investments. The investments that we were making were based on the predictions of a classification model predicting whether or not a given player will be inducted into the Hall-of-fame (and the probability thereof). These players were profitable investments.
In this code file I worked with 5 different machine learning models: Decision Trees, Bagging Models, Random Forests, Multinomial Naive Bayes, and K Nearest Neighbors.
Additionally, I utilized feature engineering, grid search, comparing ROC AUC scores between models, correlation matrices, variable filtering, and train test splitting to produce the random forest model that produced the highest profit of $3,000,000+ winning first place.