Data science has been widely applied to a lot of disciplines in recent years. However, few relevant works have been done in the field of geology. Therefore, this thesis aims to discuss the application and potential of data science in petrology and geochemistry. The most common data in petrology and geochemistry are rock images and geochemical compositions of rocks. The color and texture of igneous rocks are closely related to their diagenetic process while geochemical compositions can be used to decipher the source of rocks and the physical and chemical processes they undergo before they solidify. The development of large geochemical databases such as EarthChem and GeoRoc provides the geochemists with convenience for accessing the big data and studying the origin and evolution of Earth. However, the lack of a large rock image database hinders the potential of using data science to better understand the formation of rocks. Therefore, in this thesis, I propose an image processing workflow that allows geologists to easily acquire and report standardized rock images with smartphones, contributing to the buildup of the rock image database. I also use image analysis methods to discuss the origin of orbicular granitoids from Mount Magnet, western Australia. Moreover, I use machine learning tools to explore the relationship between heat-producing elements and major elements of the igneous rocks, providing the implication on the thermal state of the lower crust during the Archean.