We present a data-driven investigation leveraging high-throughput density functional theory calculations and machine learning to expedite the discovery of van der Waals (vdW) materials with nonlinear optical properties. Using the Computational 2D Materials Database, we analyze data from 345 noncentrosymmetric, nonmagnetic semiconductor monolayers, focusing on their second-order susceptibility tensors across multiple energy ranges suitable for various laser applications. By applying data mining techniques to extract key features from second harmonic generation spectra and employing machine learning models, we predict the second-order optical susceptibility for these materials. Our framework for this work facilitates the rapid identification of vdW materials for advanced photonics, optoelectronics, and data storage applications.