Travel Recommendation System for Domestic Travelers in Sri Lanka

Sri Lanka is one of the most attractive tourist destinations with a large number of tourist places. Choosing a destination for traveling varies from person to person depending on their background. Hence, the aim of this study is to identify the factors affecting the selection of tour destinations in Sri Lanka. To that end, we designed a model, which recommends suitable travel destinations in Sri Lanka based on travelers' demographic profiles and past travel experiences. The traveler’s profiles and travel information were collected from a sample of 600 domestic travelers randomly picked from the society through a self-administrated close-ended questionnaire. The convenience sampling method was used to select the representative sample. All questionnaire data was prepossessed by removing noisy data and encoded into discrete values. The attributes of the datasets are created using travelers’ profiles and the toured destination is set as the target attribute. The J48 decision tree learner implemented in the WEKA data mining library was trained to predict the potential tour destinations based on the traveler’s profiles travel history information. The classification accuracy of the model was measured using Area Under Curve (AUC) of Receiver Operative Characteristics Curves (ROC) as the initial class distribution is imbalanced. The results show that the personal factors of travelers such as travel mode, gender, age, marital status, education level, average income level, and hobbies are potential indicators of the travelers’ next travel destinations. This finding is very useful for the service providers in the tourism industry to customize their services based on the target groups.