In today’s world, everything is transforming to digital forms. These yield large amount of data. A good analysis of these data can lead to new knowledge about the present situation as well as the future insight. While many advantages could be obtained from these large data, the issue on how to run the Machine Learning on a large dataset as effective and efficient as possible remains an open problem. In this paper, data processing simulation using machine learning algorithm of Linear Regression is conducted to learn from Bitcoin trading dataset. The simulation is carried out in Apache Spark cluster architecture and GPU. The running time and error of the algorithm implementation in both architectures are compared with each other. The simulation results show similar error performance between Apache Spark cluster and GPU. Yet, Apache Spark can run the simulation 2x faster than GPU.