Large-scale personalized marketing modeling for effective use of transaction data

2014 Fiscal Year Final Research Report

• Project/Area Number: 24683012
• Research Category: Grant-in-Aid for Young Scientists (A)
• Allocation Type: Partial Multi-year Fund
• Research Field: Commerce
• Research Institution: Tohoku University
• Principal Investigator: ISHIGAKI TSUKASA 東北大学, 経済学研究科(研究院), 准教授 (20469597)
• Project Period (FY): 2012-04-01 – 2015-03-31
• Keywords: マーケティング / ビッグデータ / 階層ベイズモデル / 次元圧縮 / パーソナライゼーション

Outline of Final Research Achievements

This research project aims to realize a large-scale personalized marketing model. Large-scale transaction data recorded in supermarkets or convenience stores essentially are sparse with respect to consumers, items, and purchase times. We combine a dimensional reduction model with the hierarchical Bayes binary probit model for overcoming the sparseness of data. For computational feasibility, we employ variational Bayes inference that has computational efficiency compared to the resource-intensive Markov chain Monte Carlo inference in large-scale problem. The result shows that the model is applicable to datasets involving tens of thousands of consumers and hundreds of product items.

Free Research Field

統計科学、サービス科学