Analysing Graph Data Model in Apriori Algorithm
Ade Hodijah, Urip Teguh Setijohatmo, Hilman Permana, Lolla Mariah, Devina Lusiana

Politeknik Negeri Bandung

Abstract

To find meaningful patterns in the big data is one of the active research in the data mining. Market basket analysis (MBA) is one of the most useful modeling technique in data mining. It involves the mining and analysis of Association Rules. The association rules problem carried by the Apriori algorithm is represented in various problems ranging from market basket analysis, co-occurrence in natural language processing, behavioral similarity, and so on which represent the recognition of co-occurrence tendencies. The problem of the speed of combinatorial itemset acquisition from the Apriori algorithm in previous studies using memory (RAM) is still very possible to be studied. In this study, graph-based technology was chosen based on the results of the study that there is the availability of graph algorithms (graph data science from Neo4j) to support the process of obtaining association rules from the Apriori algorithm. To mine transactional data and get detailed information about the products bought together in the market basket, the research performed network based approach. This gives insights into frequently bought products and products bought together de_ned as ^Communities^. Further, we analyzed these group of communities using different measures like Network density, Centrality and PageRank algorithms. From this study, we concluded that finding communities in comparison to the old traditional method of finding Association Rule is more informative and reliable in MBA.

Keywords: Market Basket Analysis- Graph Data Science, Apriori Algorithms

Topic: Artificial Intelligence (AI)