In this study, the fluctuation patterns of the long-term groundwater levels at 87 confined and 13 unconfined water observation wells in 42 sites in Tokyo Metropolis were investigated using Self-Organizing Maps (SOM). The SOM application classified the yearly groundwater level fluctuation patterns into eight clusters showing clearly distinguishable patterns which were then divided into 3 large groups. Consequently, the results show that the fluctuation patterns of natural groundwater levels were classified into Group-1. Most of the wells in the Tama region which showed regular fluctuation patterns caused by pumping were classified into Group-3. However, the majority of the wells were moved to Group-2 due to the pumping of drinking water being suspended or decreased. These results show that the SOM analysis was successful in extracting the unique long-term fluctuations of groundwater levels.