Characteristic-based factor models have emerged as a workhorse tool in empirical finance. These factor models map expected stock returns via loadings/exposures to common risk factors. Risk factors are built using characteristic sorted portfolios aggregated to so-called "market anomalies". Extensive research has been conducted to uncover an arsenal of new market anomalies. For example, Hou et al. (2020) examine a dataset of 452 market anomalies. Market anomalies are typically constructed using absolute stock characteristic measures like the market capitalization/size or book-to-market ratio of single stocks. This project combines financial research with data science and machine learning literature to create local/relative measures to construct market anomalies. I propose applying modern unsupervised machine learning tools to uncover local stock return patterns related to underlying characteristics. Early results of the investigation revealed that one example local measure, namely: 'local factor loading uncertainty' is priced in the cross-section of stock returns. This result stands in alignment with the findings of Armstrong et al. (2013) and Hollstein et al. (2020). I strive to utilize this striking finding and deduct implications for asset pricing research. In the finalization of this project, in addition to a scientific paper, I plan to provide a unique software toolbox in Python that implements visual interpretations of the findings.