Recommendation technique of goods by fusion to subjective evaluation of sellers and data-mining

Recommendation technique of goods by fusion to subjective evaluation of sellers and data-mining

0.00 Avg rating0 Votes
Article ID: iaor20081380
Country: Japan
Volume: 14
Issue: 3
Start Page Number: 63
End Page Number: 79
Publication Date: Oct 2005
Journal: Journal of the Japan Society for Management Information
Authors: , , ,
Keywords: datamining, analytic hierarchy process, programming: mathematical, statistics: inference
Abstract:

We perform a new proposal in this paper about the ‘recommendation technique of goods’ which a seller recommends next purchase goods appropriately for every specific customer. The new recommendation technique of goods was able to conquer the problem of the conventional collaborative filtering method based on data-mining, or the contents analyzing method by the past purchase track record data, and take in subjective evaluation of a seller's sales promotion person group. (Hereafter, it is called the spot). Namely, it aims at the research on the new recommendation technique of goods and the construction of a practical model by fusion to IT (Information Technology) technique and humane sensitivity. The collaborative filtering and the contents analyzing method have the problem that a lot of track record data is needed, and it cannot apply about recommendation of the new goods before sale in the planning stage since recommendation is performed based on the similarity of the purchase track record data for every customer. Moreover, big deviation with experience of the spot and the conference was seen, and the fusion of mathematical analysis and on-site feeling of the goods evaluation only depending on data analysis was often a subject. The feature of the new technique to propose is to derive the ‘attribute relation matrix’ which evaluates the strength of the relation of a ‘customer attribute’ and a ‘goods attribute’ by the mathematical programming model. According to this attribute relation matrix, the recommended candidate goods can be extracted for customers with a certain characteristic, and the recommended goods were narrowed down on the spot by making the recommended candidate goods of alternatives, and adding the subjective evaluation by AHP (Analytic Hierarchy Process). In the application experiment result in the real shop of the new recommendation technique, the recommended candidate goods was able to mention goods with unexpected nature or freshness while the goods of the high evaluation for the spot were contained.

Reviews

Required fields are marked *. Your email address will not be published.