Parallel K-Nearest Neighbor Implementation on Multicore Processors

As the industry moves from single chip processors to multi-core processors in the general purpose community, it is becoming increasingly important to develop techniques to find and expose enough parallelism in the application programs. Parallel programming is classified in to two major groups as code parallelism and data parallelism. In order to exploit the power of multi core processors it is essential to change programming of conventional application to parallel programming paradigms. Some compiler tools have been developed to help the programmer to develop parallel applications. However, it is still a challenging problem to programmer to extract full parallelism in general applications. Here we propose a case study of classification of huge database like electoral data of Kolhapur constituency in to age wise groups using popular technique of classification using K-Nearest Neighbor on multi core CPUs. Such a classification of data will predict the age group of constituency which will help the contestant to arrange their campaign accordingly. Also trend of voting can be associated to age groups for analysis. This application demonstrates how parallel programs can be developed using multi core processors to take full advantage of parallel programming on desktop.