 This paper proposes a novel deep learning classifier to select the best enhanced oil recovery technique from a set of potential candidates. The classifier was trained and tested on 735 EOR projects involving different types of reservoirs such as sandstone, unconsolidated sandstone, carbonate and conglomerate. The results showed that the proposed deep learning classifier achieved an accuracy of 96.82% for the training data, 84.31% for the validation data and 82.61% for the testing data. Additionally, the categorical cross entropy was found to be 0.1548, indicating that the classifier can accurately identify the best EOR technique for a given reservoir. This deep learning classifier is a powerful tool for selecting the most suitable EOR candidate for a given oil reservoir with limited field information. This article was authored by Rakesh Kumar Pande, Asghar Gandamkar, Pezod Vafri and others.