Prediction of taxi requests have lately drawn increased much attention in research owing to its high applicability in massive intelligent transport systems. Most previous methods, on the other hand, focused solely on predicting taxi demand in origin areas, ignoring the analysis of target passengers' special circumstances. We believe it is inefficient to assign cabs to all areas in advance solely based on taxi origin request. This work studies a critical and fascinating task known as the taxi origindestination demand prediction, whose purpose is to forecast future cab requests across pairs in all areas. Determining the process to collect contextual data effectively in order to learn demand patterns. A novel methodology known as the Deep Neural Network (DNN) with Deep learning-based models is focused in this paper, which outperforms traditional machine learning methods in a variety of classification tasks, including origin and destination views. Extensive tests and analyses on a broad dataset clearly show that our DNN outperforms several other methodologies from literature. © 2023 Elsevier B.V., All rights reserved.