Abstract:
Tourism is a significant source of income for countries and regions, and with
the advancement of technology, everything is now interconnected, generating
massive amounts of data. Recommender systems are one way to utilize this generated
big data. However, currently, Ethiopian Tourism Institutions do not have
a system to manage tourist sites, analyze customer preferences and behavior,
or filter based on their interests. Therefore, the purpose of this research is to
develop a collaborative-based recommender system for Ethiopian tourism sites.
This study applies both user-based collaborative filtering using a cosine similarity
algorithm and model-based collaborative filtering using Singular Value
Decomposition(SVD), Non-negative matrix factorization(NMF), and K-nearest
Neighbors(KNNBasic) to analyze the data collected from the Amhara tourism
office, which describes the Ethiopian tourism site. After the analysis, the results
show that cosine similarity has the lowest Root Mean Square Error(RMSE)
score.The methodology followed in this research is the design science approach,
and the artifact recommender system is developed using the flask framework.
After demonstrating the recommender system, domain experts evaluate it,
achieving a promising result of 84.2%. The scarcity of data is a significant
challenge, and thus, limited attributes for clustering are selected. Using more
attributes to make the similarity of tourists will improve the development of the
tourist site recommender system, which is left for future researchers to consider.
