Generic Itemset Mining Based on Reinforcement Learning

One of the biggest problems in itemset mining is requirement developing a data structure or algorithm, every time user wants to extract different type itemsets. To overcome this, we propose method, called Generic Itemset Mining based on Reinforcement Learning (GIM-RL), that offers unified framework train an agent for extracting any In GIM-RL, environment formulates iterative steps target itemsets from dataset. At each step, performs action add remove item current itemset, and then obtains reward represents how relevant resulting type. Through numerous trial-and-error where various rewards are obtained by diverse actions, trained maximise cumulative so it acquires optimal policy forming as many possible. this framework, can be long suitable defined. The extensive experiments high utility itemsets, frequent association rules show general effectiveness one remarkable potential (agent transfer) GIM-RL. We hope GIM-RL opens new research direction towards learning-based mining.