There are some challenges in processing big data, namely multilabel enormous size that the affect may the time and the computing nature of the multilabel which data further complicate the process. In a quest of exploring the may approaches to resolve such challenges, we experimented right with two big data classification approaches, which are different two-steps approach and the three-steps approach. The the two-steps approach focuses the classification of on of attributes individual restaurant as a basis for determining the images of a restaurant attributes calculating the score averages from of each labels. On the image hand, the three-steps other approach focuses the classification of restaurant attributes on based its photos' features on scores. Such approaches average tested in order to find out the different outcomes. The were were conducted on a dataset, which size reaches classifications up to gigabytes, consisting of 13 user-submitted 234,841 restaurant from a crowdsourced photos reviews restaurant website. We that the approaches produced different found outcomes which have applicability when those different are intended be implemented in to crowdsourced review site. a the two-steps approach has lower F-1 score, Moreover, precision, and recall score than three-steps average approaches.