Histogram based rule verification in lifelong learning models

Lifelong learning models are popularly used with big data analysis as it learns better with the volume and variety of data. The model learns independently through an augmented learning mechanism that does not require manual support. Learning wrong and irrelevant rules are expected as it follows an unsupervised approach and therefore, the model is supported with a filtering mechanism. The rules that do not keep up the minimum par of utility to oldness ratio or strength from the following tasks, are filtered out. The existing approach uses flat thresholds to filter rules with a static approach, irrespective of the experience the model has gained. It slows the learning procedure in early tasks and creates congestion and bottleneck at high experience. In this research paper, histogram based rule verification mechanism is proposed that has rule verification thresholds adjusted according to the peaks of histogram of rules. The proposed approach improved topic coherence at earlier tasks while avoid congestion at high experience.