Config file has missing hyperparameter configuration and inefficient grid search ranges across regressors

[midaa1]

Description

While running the training and prediction pipeline, the Random Forest Regressor consistently fails during execution, causing the model evaluation stage to break and preventing full comparison across supported regressors.

Additionally, the current hyperparameter grids for several models are overly broad and include unstable or unnecessary ranges, which results in:

• Excessive GridSearchCV runtime

• Large search spaces with low-value configurations

• Slower experimentation cycles

• Inconsistent model ranking

Problem Observed

From the evaluation report:

• Random Forest Regressor throws a pipeline error because no hyperparameter grid is defined.

• Some models (e.g., SVR, SGD, ElasticNet) explore very large or impractical search spaces.

• Training time is high for some models without noticeable performance improvement.

Impact

• Incomplete benchmark comparison.

• Longer experimentation cycles.

• Higher computational cost.

• Harder to iterate quickly on model improvements.

Proposed Solution

• Add hyperparameter grid for Random Forest Regressor.

• Narrow and optimize parameter ranges for existing models.

• Remove unnecessary extreme values.

• Focus grids on empirically effective ranges based on prior runs.