Published Paper Details:

Abstract: A data stream represents a continuous, real-time, and ordered sequence of data items that arrive at high velocity from dynamic sources such as network traffic, sensor outputs, financial transactions, and call centre logs. Unlike traditional data mining, which deals with static and finite datasets, data stream mining involves extracting meaningful knowledge structures from unbounded and rapidly evolving data. The major challenge lies in handling the volume, velocity, and variability of continuous data streams while maintaining computational efficiency and limited memory usage. To address these challenges, several algorithms have been developed for mining frequent item sets in streaming environments, including Apriori, Partition, Pincer-Search, FP-Growth, Dynamic Itemset Counting, ECLAT (Equivalent Class Clustering and Bottom-Up Lattice Traversal), and RARM (Rapid Association Rule Mining). Among these, the ECLAT algorithm employs a vertical data format for efficient frequent pattern discovery, whereas RARM utilizes a tree-based structure to represent transactions for rapid rule generation and minimal database scans. This study presents a comparative experimental analysis of ECLAT and RARM algorithms for mining frequent item sets in data streams. The experimental results demonstrate that RARM outperforms ECLAT in terms of execution time, scalability, and memory efficiency, making it a more competent choice for high-speed streaming data environments. The findings highlight the potential of adaptive algorithms like RARM for real-time analytics, paving the way for future integration with Machine Learning-based stream mining models to enhance predictive accuracy and intelligent decision-making in Big Data applications.