- Dataset from the data journalism website fivethirtyeight, listing college majors, associated demographic data, and their economic outcomes, especially income and employment
- Data includes only recent BA/BS graduates to get an up-to-date view of current economic outcomes
- Begins by constructing new variables, e.g., share of recent grads working part-time
- Exploratory data analysis with correlations, visualizations, and multiple linear regressions
- Three algorithms (K-Means, KNN, and Hierarchical Clustering) were used to build clusters of majors
- Loops with different validation methods found the optimal number of clusters
- Principal Component Analysis was applied to visualize the two successful models: K-Means and Hierarchical Clustering
- The PCA plot visualizing the K-Means clusters is as follows:
