With the rapid growth of e-commerce and user-generated content online, the increasing product online reviews have significant influence on both buyers and sellers. However, among the thousands of online reviews, only the reviews of high-quality matters to the market, thus quality reviews detection rises in response to the requirement of retrieving authentic feedbacks from consumers. In this paper, a state-of-the-art ensemble model, gradient boosting decision trees (GBDT), is applied to select useful features for quality evaluation of online reviews. Firstly, four types of features are extracted based on information adoption theory. Then, the GBDT model is adopted to select useful features for quality reviews detection. At last, comparative experiments are conducted through online reviews of searching goods, based on two baseline models such as Decision Tree and Logistic Regression, and the results show that GBDT model achieves a better performance in detecting reviews of high-quality. This research indicates that product attributes, reviewer characteristics and objectiveness of reviews are key ingredients in high quality reviews.