Fuzzy Rule-Based Intelligent System for Predicting Hotel Occupancy

Accuracy and interpretability have always been significant issues in forecasting methods, and it is very important to have a balance between these issues when developing a system in tourism demand forecasting. There are various clustering algorithms used in many branches. The efficiency of the clustering technique is stipulated by the performance of the clustering results. Fuzzy c-means algorithm is highly efficient for unbiased clustering. In this thesis, fuzzy cmeans algorithm is applied on monthly number of guest arrivals in one of the hotels of North Cyprus over 40 months to find the optimal number of clusters in the analysis problem. Also, the fuzzy rule-based system model for hotel occupancy forecasting is developed, and in order to enhance the comprehensibility and accuracy of this model, Mamdani fuzzy rule-based system is used. Based on the values of root mean square error and mean absolute percentage error which are metrics for measuring forecast accuracy, it is defined that the forecasting model with 7 clusters and 4 inputs provides an optimal solution of the problem. Keywords: Forecasting, Time series, Fuzzy c-means clustering, Fuzzy rule-based system, Mamdani model