The Machine Learning “Advent Calendar” Day 6: Decision Tree Regressor

A Decision Tree Regressor model works by recursively partitioning data to minimize prediction error. The optimal split for a feature is found by testing all possible split points, which are typically the midpoints between consecutive sorted values. For each potential split, the data is divided into two groups, and the weighted Mean Squared Error (MSE) is calculated across both groups to find the split that results in the lowest error. This process, demonstrated using Excel, can be extended to handle multiple continuous or categorical features by evaluating all possible splits across all features and selecting the global best.