A research on modeling the human activity preference was conducted in this thesis. The dataset being used in this research is a historical Foursquare dataset containing check-ins made throughout the period of ten and a half months in Tokyo, Japan. Thesis claims that the decision on our activities are influenced by so called decision segments. In order to model the human activity preference, multilayer perceptron machine learning model is used. Two modeling approaches are tested trying to capture decision segments by using different sets of features. The models are evaluated and results of the research are reported.