With the development of wireless communication, visible light communication (VLC) has become very promising technology owing to its many advantages. However, the nonlinear effect of VLC introduces many challenges for signal processing and deteriorates system performance. As machine learning has many advantages and significant potential for solving nonlinearity issues, the VLC that utilizes machine learning algorithms is bound to have tremendous research value. Existing research shows that traditional machine learning algorithms, such as K-means, DBSCAN, and support vector machine, perform well in pre-equalization, post-equalization, anti-system jitter, and phase correction. A deep neural network can further improve the performance of the VLC system because of its strong nonlinear fitting ability. In this article, we analyze the aforementioned methods and introduce their application to the signal processing in VLC. We hope this paper provides a reference for solving the nonlinearity problems related to machine learning in VLC.