Organic molecules are known to adopt various crystal habits such as one-dimensional needles, two-dimensional plates, and three-dimensional blocks during crystallization. Among them, the preferred crystal habit for organic semiconductors used in organic transistors is a plate shape, which is easy to form thin films on a substrate. To date, however, it has been difficult to predict crystal habits without experiments or crystal structure information. Therefore, the purpose of this study is to predict the crystal habits from the molecular structure of organic semiconductors using machine learning.