Physical activity is one of the most important factors in leading a healthy life, which has increased the interest in the scientific community to evaluate methods and tools that can help people maintain an exercise routine, such as portable devices that can track the movements of the user and provide an appropriate feedback. Interest has also emerged in assessing the discrimination between physically active and inactive persons through the use of readily available data, which is the aim of this work. In this case, we used an auto-encoder to find the most outstanding characteristics of an anthropometric data set, in order to get the most representative attributes. Then use them to train an Artificial Neural Network (ANN), so that it could learn to identify between a physically active and a sedentary person. The ANN obtained 81% accuracy, 82% precision, 88% recall, 83% F1 score and 0.89 AUC. These results position the ANN as a viable model that could be used as a tool in scenarios such as customer profiling for different interested companies.