メインナビゲーションにスキップ 検索にスキップ メインコンテンツにスキップ

Identifying Purchasing Factors in Online Flea Markets Considering Thumbnail Images

研究成果: Article › 査読

抄録

In recent years, the market for online flea markets, which are consumer-to-consumer (C2C) services where goods are bought and sold among users, has been expanding. In such services, sellers (individuals that offer goods or services for sale) create product details, including price, condition, shipping method, and images, when listing their items for sale. We consider the product image to be the first thing users see when selecting a product as a thumbnail, significantly impacting their purchase decisions. In this study, we proposed a discriminant model of purchase decisions for online flea market data to clarify the factors influencing these decisions based on product details. Specifically, we used metadata such as price and delivery method, along with image labels for product thumbnails, as features. We created and compared models with three patterns: metadata only, image labels only, and a combination of metadata and image labels. We created four types of models in this study: logistic regression, decision tree, gradient boosting, and random forest. We selected the models based on their accuracy evaluations. Our analysis revealed that the model using both metadata and image labels as features, combined with the gradient boosting method, had the highest accuracy. The partial dependence plots of the selected models highlighted the features important for users’ purchase decisions.

本文言語English
ページ(範囲)340-348
ページ数9
ジャーナルJournal of Data Science and Intelligent Systems
巻4
号3
DOI
出版ステータスPublished - 2026 7月 28

UN SDG

この成果は、次の持続可能な開発目標に貢献しています

  1. SDG 12 - つくる責任つかう責任
    SDG 12 つくる責任つかう責任

ASJC Scopus subject areas

  • コンピュータ サイエンスの応用
  • 情報システム

フィンガープリント

「Identifying Purchasing Factors in Online Flea Markets Considering Thumbnail Images」の研究トピックを掘り下げます。これらがまとまってユニークなフィンガープリントを構成します。

引用スタイル