Predictive models for solar weather and activity rely on large amounts of data from extensive timescales, from one year to multiple 11-year-long solar cycles. I worked with Slava Sadykov at Georgia State University to support the development of more accurate prediction models for solar activity. Using magnetograms from the operational overlap between the Solar Dynamics Observatory (SDO/HMI) and Solar and Heliospheric Observatory (SOHO/MDI), I established machine learning models to enable the creation of a homogenous magnetic dataset spanning from 1994-present. These regression models resolved differences in instrument resolution, operational limitations, and observatory location. I also researched algorithm refinement and effectiveness on active regions and the quiet Sun.