Introduction: Rapid identification of cocoa bean varieties is vital for the authentication in cocoa trade. This paper examines the use of Near Infrared (NIR) Spectroscopy for nondestructive identification of cocoa bean cultivars. Methods: In this study, five cocoa bean cultivars (IMC85 x IMC47, PA7 x PA150, PA150 x Pound7, Pd10 x Pd15 and T63/967 x T65/238) were scanned in the NIR range of 10000-4000 cm-1. Linear discriminant analysis (LDA) and Support vector machine (SVM) algorithms were performed comparatively to build discrimination models based on principal component analysis (PCA). The models were optimized by cross validation to ensure their stability. Results: The performance of SVM model was superior to LDA model. The optimal SVM model was achieved with five principal components (PCs) and an identification rate of 100% in both training set and prediction set. Conclusions: The results proved that NIR spectroscopy technology with SVM algorithm can provide quick and reliable nondestructive analytical tool for the identification of cocoa bean cultivars and this would aid in quality control of cocoa bean.