The recommendation system’s task is to predict what a user would like to buy in his/her next shopping period given in shopping history. Many state-of-the-art recommendation systems have relied mainly on user explicit feedback to deliver recommendations to the end-users since implicit feedback does not directly reflect user feedback. A set of items (multiple) that can be recommended for an user is called next basket recommendation. Most of the next basket recommendation systems treat user sequential behavior and common interests differently. These types of systems are only capable to capture local sequential behavior between adjacent baskets. In this study, an effort was made to close the gap between the hybrid approach’s locally and globally collected item sequences. This paper proposes a hybrid recommendation system that deals with both issues, achieving better results. The model consists of an autoencoder to extract rich contextual information from the user implicit feedback and a recurrent neural network to learn a dynamic representation of the user and capture global sequential features between baskets. The proposed approach has been tested on two bench mark datasets and the results indicate that it is more effective than the state-of-the-art recommendation systems.