Algorithm 1: Feature Selection using RFE

Input: Feature matrix $\mathit{x}$, target vector y, desired number of features $\mathit{k}$ Output: Selected feature subset $\mathit{X}\text{\_}\mathit{s}\mathit{e}\mathit{l}\mathit{e}\mathit{c}\mathit{t}\mathit{e}\mathit{d}$ 1. Initialize feature set $\mathbf{F}=$ all features in $\mathit{X}$ 2. While the number of features in $\mathbf{F}>\mathbf{k}$:       a. Train model M on $\mathit{X}[:,\mathit{F}]$ to predict $\mathit{y}$       b. Compute feature importance scores ${\mathit{T}}_{\mathit{j}}$ for each feature in $\mathbf{F}$       c. Identify feature $\mathbf{F}\text{\_}\mathbf{m}\mathbf{i}\mathbf{n}$ with the lowest importance       d. Remove $\mathbf{F}\text{\_}\mathbf{m}\mathbf{i}\mathbf{n}$ from $\mathbf{F}$ 3. Return $\mathit{X}\text{\_}\mathit{s}\mathit{e}\mathit{l}\mathit{e}\mathit{c}\mathit{t}\mathit{e}\mathit{d}$ = $\mathit{X}[:,\mathit{F}]$