Stochastic variational method as a machine learning approach to large-scale few-body system calculations is performed with Gaussian expansion method. As a basic research to far beyond the three-body systems, the methodology is confirmed in three-body system called \(^4\) He trimer. When the number of particles is large, it becomes impossible to perform the calculations at around seven-body system even with a massively parallel computer such as Fugaku supercomputer, mainly due to memory capacity limitations, but this limitation can be significantly alleviated by reducing the number of necessary basis functions. Therefore, a machine learning approach that searches for basis functions with as few basis functions as possible is expected. In the present calculation with a machine learning approach based on SVM, the calculated minimum energ with 625 basis functions, which is a standard number in GEM, is reproduced with learned significantly less 54 basis functions.