Fusion of diversified utility factors for high utility mining

Abstract

High utility mining is gaining prominence, and with the increasing set of business intelligence models, the scope of such significant practices is high. Rather than focusing only on profitability as one key utility metric, today’s organizations believe in having more robust levels of the multi-objective filtering process. In this manuscript, a contemporary model of the high utility mining process is proposed, wherein the multiple averages are used for grading the recommendation of the itemsets for merchandise. The model’s key advantage is its dynamic approach. The goods-related period of the average time interval can be flexible, alongside the fusion of multiple utility thresholds of diversified features chosen for itemsets recommendation. The performance analysis has been carried out by using a multi-fold cross-validation strategy. The results obtained for cross-validation show that the proposed model is outperforming the contemporary models with significant precision, specificity, sensitivity, and accuracy having values 97%, 95%, 98%, and 97% in respective order. Whereas, the contemporary models HUPM-MUO and MOEA-FHUI have obtained 93% and 90%, 88% and 82%, 89%, and 84%, and 89% and 83% in respective order of the corresponding metrics. The experimental study of the model denotes the effectiveness and ease with which the solution can generate results and produce significant output in the real-time environment for more dynamic and periodic decisions by different organizations.