Improving the Accuracy of Room Occupancy Forecasts with Hybrid Models

Conventional prediction models used so far can often not capture complex patterns influenced by various dynamic factors such as time, weather, scheduled activities, and user behaviour. This study aims to predict the occupancy rate of rooms in a popular tourist destination. The Fuzzy Time Series method was chosen because of its flexibility and ability to work without strict statistical assumptions. The addition of Markov Chains has been shown to reduce the error rate, while SSA improves the model by decomposing the data into trend, seasonal, and residual components. This study found that the hybrid FTSMC-SSA method significantly outperformed the traditional method, with a Mean Absolute Percentage Error (MAPE). This shows that the developed hybrid model has significantly improved the accuracy of room occupancy forecasting compared to a single conventional model. This model can capture complex temporal and non-linear patterns in occupancy data by combining machine learning methods such as Random Forest and Long Short-Term Memory (LSTM) and statistical approaches such as ARIMA. The implications of this study are significant for facility management and space planning in various sectors, such as offices, educational institutions, hospitals, and shopping centres. With the increased accuracy of room occupancy forecasts through hybrid models, managers can make more informed decisions regarding space usage scheduling, automatic lighting and ventilation settings, and energy savings.