The prediction of human activity is the process of inferring forthcoming activity labels based on past and current actions. The ability to accurately predict future actions is a crucial aspect of many real-world applications, ranging from autonomous driving to human-robot interactions. The ideal prediction system should have the ability to predict future actions with high accuracy. In this study, we investigate the problem of human activity prediction using Hidden Semi-Markov Models (HSMMs), which are probabilistic models widely used in activity recognition and language processing. To evaluate the prediction capabilities of HSMMs, we conducted experiments on labeled activity sequences from an existing dataset representing human actions in a industrial setting. Our results indicate that HSMMs can be used for human activity prediction