Tool for evaluating target performance using a comparative table

[infingardi]

What was changed

Added function targets_performance_comparative_table in _bibmon_tools.py

Objective

Evaluates the performance of all dependent variables (targets) in a regression model by computing key performance metrics:

• R² score: Measures how well the predictions align with actual data.
• Mean Absolute Error (MAE): Quantifies the average prediction error.
• False Alarm Rate (FAR): Represents the frequency of false positives.
• Fault Detection Rate (FDR): Indicates the accuracy in identifying true faults.

Importance

Selecting the best target variable is essential for improving regression model performance. It ensures more accurate predictions (higher R², lower MAE), better fault detection (higher FDR), and fewer false positives (lower FAR). This helps optimize model reliability, reduces noise from false alarms, and directs resources effectively, making the system more efficient and robust in practical applications.

Benefits

Quick way to find the best variable (target) to be considered in the Y set.

How to use

tab_pred, tab_detec = targets_performance_comparative_table(dataFault, 
                                         start_train=start_train,
                                         end_train=end_train, 
                                         end_validation=end_validation,
                                         end_test=end_test,
                                         tags= allTags,              
                                         model=model,
                                         metrics=mtr, 
                                         count_window_size=3, 
                                         count_limit=2,
                                         fault_start=fault_start,
                                         fault_end=fault_end)

# Print Prediction table
tab_pred

# Print Detection table
tab_detec