The detection of radiological materials is key to ensuring a robust second line of defense in nuclear security. Neutron-capture prompt-gamma activation analysis (PGAA) is an efficient non-destructive radio-analytical technique in the measurement of elements that do not form neutron capture products with delayed gamma-ray emissions. PGAA is a useful for in-situ multi-elemental detection technique for radioactive materials across the entire Periodic Table, from hydrogen to uranium and can be used to develop low cost detectors for application in nuclear forensics. However, the unavailability of accurate and complete data posed a significant challenge in the qualitative and quantitative analysis of complicated capture-gamma spectra by means of PGAA. As a consequence of the various recommendations and coordinated effort, International atomic energy agency (IAEA) initiated the development of a database for Prompt Gamma-ray Neutron Activation Analysis in 1999. In this study, diverse machine learning algorithms are developed to classify the elements such as Cobalt, Caesium, Iridium, Uranium and Thorium based on the PGAA energy spectra (Eɣ) and effective capture cross section (σ). The classification algorithms employed in this work are based on K-nearest Neighbours (KNN), Artificial Neural Networks (ANN), Support Vector Machines (SVM), Classification Trees (CT), Random Forest (RF) and K-means clustering. The model performance is evaluated based on classification metrics namely accuracy, precision, recall and f1-score. The results are used to evaluate the classification performance of models developed using different machine learning algorithms and to enhance the understanding of the models that are best-suited for classifying elements using low dimensional data.