International Conference on Applied and Pure Sciences (ICAPS)
Permanent URI for this communityhttp://repository.kln.ac.lk/handle/123456789/21779
Browse
Item Travel recommendation system for Sri Lankan tourist based on Hybrid- Neural Collaborative Filtering(Faculty of Science, University of Kelaniya Sri Lanka, 2024) Abeysekare, E. M. N.; Jayetileke, H. L.Tourism in Sri Lanka plays a vital part in the economy by attracting tourists with its beautiful beaches, historical places, and its diversity of wildlife. The country has a mix of cultural heritage places, pretty landscapes and places which offer opportunities for adventure tourism such as surfing, hiking, etc. With the development of technology, travelers are left with a large number of options making it difficult to obtain personalized recommendations according to their preferences. Therefore, developing a travel recommendation system is of great importance in providing customized recommendations by increasing personalization and saving time. As of now there is no specific information available about a travel recommendation system operating exclusively in Sri Lanka. This research introduces a novel travel recommendation system for Sri Lankan tourists in two phases. In phase one Agglomerative Nesting (AGNES) method is used to generate the travel recommendations followed by genetic algorithm to obtain the optimal route for a given location based on the nationality of the tourists. And in phase two Neural Collaborative Filtering (NCF) technique is implemented. The data required to generate ratings for this study is provided by Sri Lanka Tourism Development Authority which includes 5,000 data points covering demographic details such as nationality, age, travel preferences, and ratings of various tourist locations and activities. In this study, identified difficulties in existing traditional recommendation systems such as cold start problem, data sparsity, and scalability are addressed by using demographic information to overcome cold start problem, using matrix factorization method to solve data sparsity problem and develops a modified system. Then NCF method is implemented to generate more accurate recommendation technique by increasing the performance of modified recommendation system. The traditional recommendation system was implemented by hierarchical clustering without addressing any existing issues and used as a baseline method to compare with other two models. Five distinct tourist groups are identified using AGNES clustering algorithm based on preference patterns. The clustering tendency of the data was validated using Hopkins Statistic (0.7836). This value implies strong clustering potential. In order to enhance the accuracy and performance of the system NCF is used which finally results in a better performance with the lowest Root Mean Square Error (RMSE - 0.1547) and Mean Absolute Error (MAE - 0.1397) compared to traditional and modified systems. Through the use of AGNES and NCF approaches in travel recommendation system, this study significantly improved the performance and accuracy of travel recommendations for tourists. These implemented results can be adapted in to travel app, which gives personalized recommendations and shortest path for the tourists optimizing the travel experience in Sri Lanka as a future research recommendation.