Optimize hyperparameter search space and add Random Forest configuration to model pipeline

[midaa1]

Summary

This PR refactors the hyperparameter configuration used by the regression model pipeline to improve stability, reduce search time, and fix missing support for Random Forest evaluation.

Key goals:

• Add missing Random Forest hyperparameter grid.

• Reduce computational overhead from overly broad search ranges.

• Focus tuning on values that performed well in prior experiments.

• Improve overall pipeline reliability and comparability.

Changes Made

1️⃣ Added Random Forest Regressor configuration

• Introduced a GridSearchCV parameter grid:

  "Random Forest Regressor": {
      "param_grid": {
          "randomforestregressor__n_estimators": [100],
          "randomforestregressor__max_depth": [10],
          "randomforestregressor__min_samples_split": [2, 5, 10],
      }
  }```
  

Result:

Pipeline no longer fails.

Random Forest now participates in evaluation and ranking.

2️⃣ Hyperparameter space optimization

• Lasso Regression

  • Removed very small alphas.

  • Focused on stronger regularization values that previously performed better.

• Ridge Regression

  • Reduced alpha search range.

  • Added intermediate values (e.g., 0.005) for finer tuning.

• Elastic Net

  • Removed unstable edge values (0.0, extreme large values).

  • Focused on practical alpha and l1_ratio combinations.

• SGD Regressor

  • Reduced parameter permutations.

  • Fixed max iterations for consistency.

  • Narrowed alpha and eta0 ranges.

• Support Vector Regressor

  • Restricted to rbf kernel (best empirical performer).

3️⃣ Performance Improvements

• Observed after modifications:

• Random Forest successfully evaluated.

• Significant reduction in search time for SGD and other models.

• Better-performing parameter combinations discovered for SVR.

• Full performance summary now includes all models.

• Expanded higher C and gamma ranges around observed optimum.

Explanation of what has done :

• video to explain : https://drive.google.com/drive/folders/1I9Rk6oKRqAvuEIfg5Kjwg0fPdNszXbbr?usp=sharing

This if the fix of #69

[midaa1]

@jvJUCA @marcgc21 @KarinePistili