Title:Assessing Retail Employee Risk Through Unsupervised Learning Techniques

Abstract:The retail industry is a data-rich environment, especially in the case of risk management. However, much of this risk management data is not used to its full potential and there exists a need for more automated analytic tools. Supervised learning solutions are not enticing due to the difficulty of providing meaningful labels for these types of dataset. In this project, we explore the utility of three unsupervised learning techniques: Principal Component Analysis (PCA), t-Distributed Stochastic Neighbor Embedding (t-SNE), and Self-Organizing Feature Maps (SOFM) on an unlabeled real world dataset. Specifically, we use these techniques to identify features that correlate to the amount of risk a retail cashier poses to the store. We show that while PCA and t-SNE allow clustering and visualization of cashier data, they are not sufficient for detailed risk management analysis. Also, we expand the reach of the SOFM algorithm in real world applications by showing that it provides better visualization than PCA and t-SNE, and allows for detailed analysis of this dataset.