A RECOMMENDER MODEL FOR SEASON-BASED ACCOMMODATION SERVICES USING THE RANDOM FOREST ALGORITHM

Abstract

Tourism is a crucial industry sector which represents in some countries the highest contributor to their total gross domestic product (GDP). The technological advancements in recent decades shifted the tourism sector into another level of interest which made more than half of travel reservations of accommodations over the internet. Tourism recommender models which rely on machine learning techniques, were one of the breakthroughs in recent advancements which encouraged more people to make online reservations of accommodation services based on personal preferences. However, none of research works in the literature of tourism recommender models proposed recommending accommodation services based on the weather characteristics of seasons which are very impactful to the tourism recommendation. This research proposed experimental research design to implement season-based recommender model for the accommodation services using feature-based classification recommender. The proposed recommender model utilised natural language processing (NLP) techniques for feature extraction and the ensemble learning of Random Forest algorithm for the classification of the extracted features. The implementation of the proposed recommender model relied on the hotel reviews of the Datafiniti’s dataset. The researcher exploited the dataset to extract textual features using three NLP techniques which were N-gram modelling, the term-frequency-inverse document frequency (TFIDF) vectorisation, and the bag-of-words (BOW) representations. Subsequently, the researcher leveraged the Random Forest Algorithm to establish four predictors for each season for the extracted features. The results showed that the recommender model was capable to make season-based recommendation and the evaluation of the ground-truth indicated promising results. However, the model sightly performed poor in the generalisation of test data, as a consequence, the researcher proposed some improvements in future works to obtain more accurate prediction of accommodation services.