Symbolic Regression Mini-Application within Jupyter Notebook

Objective: to create a general-purpose piece of software which allows for easy point and shoot model development which is highly interpretable. Example open-source library’s which can help with tackling the project we are after: • [login to view URL] • [login to view URL] Steps we expect the software (which by the way should happen inside of a Jupyter Notebook – nothing fancy): Enter Data • Think CSV or Tabular view of data Prepare Data • Basic stuff • Handle missing data (offer median or mean of column) • Categorical needs to be converted to dummy variable (1 if present, else 0) Set Target • Search for a formula f( ) that satisfies the equation • Primary options o See Building Blocks section at the bottom o Check the boxes next to building blocks you want to be included in the mix. • Error metric o See error metrics we need to show Start Search Figure 1-Example of performance statistics updating during search Figure 2-Example project log of all the various solutions being tested and when. Figure 3-Example of achieving objective over time. View Results Figure 4-Example of formula output Figure 5-Example of stack ranked solutions by lower error rate Figure 6-Example of plotted x and y values for train, validation, and solution sets. Figure 7-Example output of statistics on train and validation data. Figure 8-Example of Evaluation Metric as compared to the complexity of formula. *** Please See Attached Document for More Details ***