An Empirical Test of Stacked Autoencoder as Recommendation Model

Abstract

The Autoencoder method can be used for multiple scenarios, as it is very variable.In this case, the method is used for suggesting actions. This paper describes the theoretical aspects of the recommendation models. The next section describes the use of the autoencoder.There will be provided two experiments. The first one, recommendations for movie lens dataset with process of transformation datasets and proper model dimension setting included. The results of the first experiment are favorable. The second experiment for suggesting actions from custom workflow inspired dataset. In this experiment, two ways of preparing a dataset for a model will be tested. Unfortunately, results are worse than model example. The article concludes with a discussion of the results achieved in comparison with other authors where results with movie datasets.