In the mid-latitude Northwest Pacific Ocean, subtropical and subarctic waters meet, intermingle, and mix to form multiple ocean domains characterized by a variety of unique oceanographic structures. The majority of previous studies on the characteristics and distribution of these ocean structures were based on existing definitions of the structure and water mass of interest derived from spatiotemporally limited data. With the massive amount of data accumulated by Argo and a small spatiotemporal bias, data-driven scientific methods may be able to objectively recapture the ocean structure. In this study, we used unsupervised clustering to analyze the temperature and salinity profiles from Argo float in the mid-latitude Northwest Pacific Ocean, and the results were compared to previous knowledge of ocean structure. The results showed that classes were distributed to form specific regions and each class has different oceanographic features that generally correspond to the previously described regional divisions. A striking advantage of this new method is that it quantifies the relative abundance of each class of profiles on a 1° grid. Furthermore, we discovered that the dynamic state of Kuroshio Extension affects the distribution of some classes and the percentage of profiles that are robust to clustering. This suggests that the stability of the Kuroshio Extension flow path has an effect on the distribution of ocean structure. These findings suggest that unsupervised clustering is useful in the analysis of oceanographic structures, allowing us to investigate oceanographic structures in areas that have not previously been extensively studied.