At present online shopping is very popular as it is very convenient for the customers. However, selecting smartphones from online shops is bit difficult only from the pictures and a short description about the item, and hence, the customers refer user reviews and star rating. Since user reviews are represented in human languages, sometimes the real semantic of the reviews and satisfaction of the customers are different than what the star rating shows. Also, reading all the re-views are not possible as typically, a smartphone gets thousands of reviews in popular online shopping platform like Amazon. Hence, this work aims to develop a recommended system for smartphones based on aspects of the phones such as screen size, resolution, camera quality, battery life etc. reviewed by users. To that end we apply hybrid approach, which includes three lexicon-based methods and three machine learning modals to analyze specific aspects of user reviews and classify the reviews into six categories--best, better, good or somewhat for positive comments and for negative comments bad or not recommended--. The lexicon-based tool called AFINN together with Random Forest prediction model provides the best classification F1-score 0.95. This system can be customized according to the required aspects of smartphones and the classification of reviews can be done accordingly.