Dataset and Machine-Learning Code for a Truncated-Ground Octagonal Annular-Ring MIMO Antenna

Description

This dataset contains the simulated electromagnetic responses and machine-learning code associated with a two-port truncated-ground octagonal annular-ring MIMO antenna. The electromagnetic data were generated using full-wave simulations in CST Microwave Studio over the frequency range of 6–18 GHz. A total of 87 distinct antenna geometries were evaluated at 501 frequency points, resulting in 43,587 observations. The dataset includes frequency and five geometrical parameters: feed length (Lf), inner radius (R1), outer radius (R2), feed width (Wf), and ground length (Lg), together with the corresponding simulated S11 and S21 responses. The accompanying Jupyter Notebook contains the Python code used for machine-learning model development, geometry-grouped cross-validation, performance evaluation, and prediction for unseen antenna geometries. Extra Trees, Random Forest, Histogram Gradient Boosting, and XGBoost regression models are evaluated. The repository also includes a README file describing the dataset variables, validation procedure, software environment, and steps required to reproduce the machine-learning analysis.

Steps to reproduce

1. Download MIMO_Antenna_ML_Dataset.xlsx and ML_Analysis_Code.ipynb. 2. Place both files in the same working directory or update the dataset path in the notebook accordingly. 3. Install the required Python packages, including scikit-learn and XGBoost. 4. Run the Jupyter Notebook cells sequentially. 5. The notebook performs geometry-grouped cross-validation using Design_ID as the grouping variable and evaluates the machine-learning models for S11 and S21 prediction. 6. The antenna geometries D07 and D11 are excluded from model development and used for external testing.

Institutions

• Parul University

  Gujarat, Vadodara

Categories

Licence