Tree-Based Machine Learning Models with Optuna in Predicting Impedance Values for Circuit Analysis.

Area of Science:

• Electrical Engineering
• Computer Science
• Materials Science

Background:

• Printed circuit board (PCB) transmission characteristics are vital for signal integrity.
• Impedance matching is critical in high-speed PCB circuit design.
• PCB impedance is influenced by production processes, requiring designer-manufacturer collaboration.

Purpose of the Study:

• To forecast target impedance values in PCBs using machine learning.
• To evaluate the effectiveness of tree-based models with hyperparameter optimization.
• To enhance impedance matching accuracy for improved signal integrity.

Main Methods:

• Applied five tree-based machine learning models: Decision Tree (DT), Random Forest (RF), XGBoost, CatBoost, and LightGBM.
• Utilized the Optuna algorithm for hyperparameter tuning of the forecasting models.
• Assessed forecasting accuracy using Mean Absolute Percentage Error (MAPE), Root Mean Squared Error (RMSE), and Coefficient of Determination (R2).

Main Results:

• All five machine learning models with Optuna achieved satisfactory forecasting accuracy.
• The LightGBM model, optimized with Optuna, outperformed other models in impedance prediction.
• Hyperparameter tuning with Optuna significantly increased the accuracy of impedance matching.

Conclusions:

• Tree-based machine learning techniques combined with Optuna are effective for predicting PCB impedance.
• This approach offers a viable and promising alternative for circuit analysis and design.
• Optimized impedance prediction contributes to enhanced signal integrity in electronic systems.