Workflow from Scientific Research

Machine learning workflow. The input data consisted of instances (samples) with labels (phenotypes) and values of features (eGenes). Instances were first split into training and testing sets. The training set was further split into a training subset (90%) and validation subset (10%) in a 5-fold cross-validation scheme. After tuning the model parameters, the optimal model was used to provide performance metrics on the basis of PCC r value between the predicted and actual values in each environment, predict labels in the testing set for model evaluation purposes, and obtain feature importance scores.

DOI