Graduation Year


Date of Submission


Document Type

Campus Only Senior Thesis

Degree Name

Bachelor of Arts



Reader 1

Professor Lisa Meulbroek

Terms of Use & License Information

Terms of Use for work posted in Scholarship@Claremont.


To assess whether or not fundamental analysis can be improved with more advanced, non-linear statistical techniques like neural networks over linear techniques like lasso and least squares, I construct estimates of market capitalization utilizing the different methodologies proposed and utilize their deviation from actual market values to construct portfolios which I test for significance. Where several different neural-network derived portfolios were able to generate statistically significant risk-adjusted returns, least squares and lasso regression were not under any scenario. This leads me to the conclusion that neural networks are in fact a superior means of conducting fundamental analysis and that the relationship between fundamental values and a firm’s subsequent returns is complex and best explained by a model which is capable of modeling these non-linear relationships.

This thesis is restricted to the Claremont Colleges current faculty, students, and staff.