AI - BASED SPECTRUM USAGE PREDICTION FOR 5G NETWORK

Abstract

The research focuses on developing an intelligent system to forecast spectrum utilization using machine learning techniques. With the rapid expansion of 5G technology, efficient spectrum management has become critical to maintaining optimal network performance. This project leverages historical data related to spectrum usage to build predictive models that assist in proactive resource allocation and network planning. The implementation begins with uploading and inspecting the dataset to identify structural details, missing values, and duplications. Pre-processing steps involve filling missing data using statistical imputation, removing duplicates, encoding categorical features such as ID, area, and device using label encoding, and standardizing the dataset to ensure consistency across features. Exploratory Data Analysis (EDA) is performed to understand the underlying relationships between features. Visualization techniques such as scatter plots and heatmaps are applied to reveal correlations, trends, and patterns, offering insights into the influence of different attributes on spectrum usage. The dataset is then split into independent features and the target variable, followed by division into training and testing subsets using an 80-20 split ratio. Two machine learning regression algorithms are employed for model building: K-Nearest Neighbors (KNN) Regressor and Decision Tree Regressor. The models are trained on the processed data and evaluated using standard regression metrics including Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and R² score. The KNN Regressor achieved a predictive accuracy of 66%, while the Decision Tree Regressor outperformed it with an accuracy of 91%. The final step involves deploying the trained Decision Tree Regressor on new test data to generate predictions. These predictions are then analyzed and stored for validation. Overall, the project demonstrates a reliable approach to spectrum usage prediction using machine learning, offering a practical solution for optimizing 5G network performance and enhancing resource utilization.