An Improved Apriori Algorithm Based on Data Mining

For multi-core architectures, making full use of GPU’s high-floating-point computing resources to achieve load balancing among multiple cores. This paper presents the way to use an efficient GPU graphics processing unit to speed up frequent pattern mining algorithms (GPFPM). This paper uses a parallel GPU process allocation mechanism to determine the number of GPU threads required based on the number of sorted items. Using GPU to calculate the threshold can reduce the time spent in support calculations, decrease the times of checks and comparisons, improve the candidate set confirmation time, and return the results to the CPU for the next-order operation. The experimental results show that in the multi-core platform and the hybrid platform of multi-core and multi-GPUs, the algorithm performs well, which excavates frequent set association rules with high efficiency and accuracy, satisfies the demand for efficient data mining of multi-core and heterogeneous platforms, and validates the effectiveness and feasibility of data mining algorithms based on multi-core and multi-graphics processing units.