Conjecturing-Based Discovery of Patterns in Data
This work provides a new method for discovering underlying feature relationships in data, which could be beneficial for researchers in various domains seeking to understand complex datasets.
This paper introduces a conjecturing machine that identifies feature relationships as nonlinear bounds for numerical features and boolean expressions for categorical features. The framework successfully recovers known relationships and suggests risk factors for COVID-19 outcomes that are confirmed in medical literature.
We propose the use of a conjecturing machine that suggests feature relationships in the form of bounds involving nonlinear terms for numerical features and boolean expressions for categorical features. The proposed Conjecturing framework recovers known nonlinear and boolean relationships among features from data. In both settings, true underlying relationships are revealed. We then compare the method to a previously-proposed framework for symbolic regression on the ability to recover equations that are satisfied among features in a dataset. The framework is then applied to patient-level data regarding COVID-19 outcomes to suggest possible risk factors that are confirmed in the medical literature.