Scholarlink

The increased student population at Kwame Nkrumah University of Science and Technology (KNUST) has resulted in significant congestion and slow internet connectivity on campus. We propose a robust model for predicting Wi-Fi demand on the KNUST campus to address this issue. Using time series data collected from the University Information Technology Services (UITS), we evaluate three forecasting models—Auto-Regressive Integrated Moving Average (ARIMA), Gradient Boosting Regressor (GBR), and Extreme Gradient Boosting (XGBoost)—to determine the optimal approach for predicting Wi-Fi traffic demand in both faculty and hall areas. The results indicate that the GBR model performed best in the faculty area, achieving a Mean Absolute Percentage Error (MAPE) of 18.73%, while the XGBoost model excelled in the hall and commercial areas with a MAPE of 17.95%.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Campus WiFi Demand Prediction: A Case Study on KNUST Campus

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Justice Owusu Agyemang,
  • Manhyia Gideon Aboagye,
  • Awumee Gabriel Selorm,
  • Somuah Obed,
  • Addai Clinton,
  • Henyo Chester Elikplim

摘要

The increased student population at Kwame Nkrumah University of Science and Technology (KNUST) has resulted in significant congestion and slow internet connectivity on campus. We propose a robust model for predicting Wi-Fi demand on the KNUST campus to address this issue. Using time series data collected from the University Information Technology Services (UITS), we evaluate three forecasting models—Auto-Regressive Integrated Moving Average (ARIMA), Gradient Boosting Regressor (GBR), and Extreme Gradient Boosting (XGBoost)—to determine the optimal approach for predicting Wi-Fi traffic demand in both faculty and hall areas. The results indicate that the GBR model performed best in the faculty area, achieving a Mean Absolute Percentage Error (MAPE) of 18.73%, while the XGBoost model excelled in the hall and commercial areas with a MAPE of 17.95%.