Scholarlink

<p>In highly competitive hotel markets, accurate demand forecasting plays a critical role in supporting effective revenue and pricing management decisions. This study conducts a comparative analysis between a machine learning approach that is Random Forest optimized with a Genetic Algorithm (RF–GA) and a stochastic method, the Kalman Filter (KF), to forecast hotel occupancy rates. Using monthly operational data from a three-star hotel in Surabaya, Indonesia, covering the period from April 2018 to May 2024, both models were evaluated based on forecasting accuracy measured by Root Mean Square Error (RMSE). The results indicate that the RF–GA model achieved substantially lower forecasting error (RMSE = 0.0320) compared to the Kalman Filter (RMSE = 0.5602), demonstrating that machine learning techniques can more effectively capture nonlinear demand patterns within hospitality data. From a managerial perspective, more accurate occupancy forecasts enable hotel revenue managers to optimize room pricing, anticipate demand fluctuations, and plan resources more efficiently. This research contributes to the revenue management literature by bridging computational intelligence with practical forecasting strategies in the hospitality sector, particularly within emerging market contexts.</p>

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

A comparative analysis of machine learning and stochastic methods for hotel occupancy forecasting

  • 期刊论文

  • 发表日期：

    2025年12月4日

查看全文
  • Mohamad Yusak Anshori,
  • Teguh Herlambang,
  • Denpharanto Agung Krisprimandoyo,
  • Zuraini Othman,
  • Mohd Sanusi Azmi

摘要

In highly competitive hotel markets, accurate demand forecasting plays a critical role in supporting effective revenue and pricing management decisions. This study conducts a comparative analysis between a machine learning approach that is Random Forest optimized with a Genetic Algorithm (RF–GA) and a stochastic method, the Kalman Filter (KF), to forecast hotel occupancy rates. Using monthly operational data from a three-star hotel in Surabaya, Indonesia, covering the period from April 2018 to May 2024, both models were evaluated based on forecasting accuracy measured by Root Mean Square Error (RMSE). The results indicate that the RF–GA model achieved substantially lower forecasting error (RMSE = 0.0320) compared to the Kalman Filter (RMSE = 0.5602), demonstrating that machine learning techniques can more effectively capture nonlinear demand patterns within hospitality data. From a managerial perspective, more accurate occupancy forecasts enable hotel revenue managers to optimize room pricing, anticipate demand fluctuations, and plan resources more efficiently. This research contributes to the revenue management literature by bridging computational intelligence with practical forecasting strategies in the hospitality sector, particularly within emerging market contexts.