Optimization of Tree-Based Machine Learning Models to Predict the Length of Hospital Stay Using Genetic Algorithm

Table 6

Mechanism of tree-based HPO with GA.

---

Input: Number of folds for cross-validationK, dataset D, number of generations , population size , crossover probability CP, and mutation probability MP

Output: the optimum value of tree-based models hyperparameter

 = 0

Initialize population randomly of size

While  < G do:

 =  + 1

 For p = 1 to do:

  Use the GA solutions for tree-based hyperparameters from the pth individual

  For k = 1 to K do:

   Divide D into K parts, 1 part as the testing set Test-S and () parts as the

   training set Train-S

   Train the tree-based model on the training set Train-S

   Predict the test set Test-S using the trained tree-based model

   Calculate the mean square errors (MSEs)

  End for

  Calculate the value of the fitness function based on calculated MSEs

 End for

 Select 2 individuals by the steady-state selection method

 Use the uniform crossover operator with the probability CP on selected individuals

 Use the mutation operator with the probability MP on a new individual

 Add the new individual to the population

End while

Return the optimum hyperparameter values of tree-based models

---