Frequent itemsets mining (FIM) is a fundamental task of data mining. The lack of consideration for the quantity and importance (price, profit, etc.) of items renders conventional FIM unsuitable for real-time approaches. To resolve this problem, high utility itemsets (HUIs) mining is presented that considers both the quantity and importance of items. HUIs mining generates more informative and actionable rules as compared to FIM because the utility of an item specifies business value. In the past decade, various approaches were proposed to mine the HUIs. Although the existing HUIs mining (HUIM) approaches discover the complete set of the HUIs which satisfy a predetermined minimum utility threshold \(min\_util\) , but, in general, it is a tough task for users to specify an appropriate minimum utility ( \(min\_util\) ) threshold. A very small \(min\_util\) may generate the excessive HUIs, while an overly large \(min\_util\) may generate quite fewer HUIs, which is of little use. In order to specify the appropriate \(min\_util\) threshold, top-K HUIs mining is presented, where K is the desired number of HUIs to be discovered. The traditional top-K HUIs mining approaches presume that items may consist of positive utility only. Although in real-world applications, items generally consist of both positive and negative utility values. Therefore, the discovery of HUIs with negative utility is crucial to discover more significant rules. In this paper, an efficient algorithm is proposed, namely TKHN (Top-K High utility itemsets mining with Negative utility items), to mine the top-K HUIs from the datasets considering negative item values. TKHN utilizes transaction merging techniques to compact the dataset. TKHN uses a real item utility (RIU) and presents a priority-queue-based strategy to gradually raise the internal minimum utility threshold. We also use sub-tree-based pruning strategies to accelerate the mining process. The experimental results show that the proposed algorithm outperforms benchmark methods.
