Iterative Optimization of Apriori Algorithm Based on Dynamic Transaction Filtering

With the deep integration of digital economy and information technology, the amount of global data has exploded. The traditional Apriori algorithm faces bottlenecks such as large I/O overhead, candidate set explosion and insufficient real-time performance when processing high-dimensional dynamic data. This paper proposes an iterative optimization model of Apriori algorithm based on dynamic transaction filtering (DTAO), which eliminates redundant transactions through a multi-dimensional dynamic filtering model, and optimizes the mining process by combining hierarchical pruning, adaptive threshold adjustment and incremental calculation mechanism. The experiment is based on a retail data set (5.28 million transactions) and a synthetic data set ($100,000-500,000$ transactions), and compares traditional Apriori, FP-Growth and other algorithms. The results show that when the minimum support is 0.02 and the confidence is 0.7, the DTAO model is 62.3% faster than the traditional Apriori and reduces the memory usage by 58.8%; when the data scale is expanded to 500,000, the running time is only 32.2% of the traditional algorithm, the memory peak is controlled within 1 GB, and the mining accuracy is maintained at more than 92%, which verifies its efficiency and stability in dynamic data scenarios.