HUIL-TN & HUI-TN: Mining high utility itemsets based on pattern-growth.

Area of Science:

• Data Mining
• Database Systems
• Artificial Intelligence

Background:

• High utility itemsets (HUIs) mining is crucial for extracting valuable patterns from datasets.
• Existing HUI mining algorithms often suffer from computational inefficiency and multiple data scans.
• There is a need for more efficient and scalable HUI mining techniques.

Purpose of the Study:

• To propose two novel pattern-growth based algorithms for efficient high utility itemsets mining.
• To introduce a new data structure, the tail-node tree (TN-tree), for effective utility information management.
• To evaluate the performance and scalability of the proposed algorithms against state-of-the-art methods.

Main Methods:

• Development of two HUI mining algorithms: HUIL-TN and HUI-TN, utilizing a tail-node tree (TN-tree).
• The TN-tree efficiently stores and accesses utility information of itemsets, avoiding redundant computations.
• Algorithms are designed to bypass the candidate generation phase and minimize dataset scans.

Main Results:

• The proposed HUIL-TN and HUI-TN algorithms demonstrate superior or comparable performance in running time across various datasets.
• These algorithms significantly improve efficiency by avoiding the candidate generation stage and multiple data scans.
• Scalability tests reveal that HUIL-TN and HUI-TN exhibit the most stable performance curves among all tested methods.

Conclusions:

• HUIL-TN and HUI-TN offer an efficient and scalable approach to high utility itemsets mining.
• The novel TN-tree structure is a key innovation enabling improved performance.
• The proposed algorithms represent a significant advancement in the field of data mining for HUI discovery.