Accelerating Data Mining Application in R Using CUDA C

This paper focuses on an innovative approach of implementing parallel processing using NVIDIA’s Graphics Processing Unit (GPU) to accelerate a data mining application in R. In order to accomplish this, one of the most apposite and efficient solution is to use CUDA (Compute Unified Device Architecture). We have used the k-means clustering algorithm to demonstrate the effectiveness of CUDA C in terms of speed-up and reduced latency. It is a widely used unsupervised learning technique in data science applications. The currently existing sequential R programming technique using C for k-means algorithm was converted to a more optimized and efficient code that uses concepts of parallel computing using GPU. The efficiency of C and CUDA C codes has been compared on the basis of execution time.