2024 Volume 75 Issue 3 Pages 89-101
In recent years, it has become possible to accumulate a large amount of browsing history data from users on websites and it is desirable to make use of such data for marketing purposes. In particular, customer browsing history collected on EC sites is an important resource for analyzing purchasing behavior because it shows the differences in user preferences and their motivations behind purchasing products. In a previous study, the Latent LSTM Allocation model was proposed for grouping customers' browsing history data. This is a model that combines the Long Short-Term Memory and the Latent Dirichlet Allocation models, but it does not consider the relationship between browsing behavior and purchase made. In addition, a customer's browsing time on each page is not taken into consideration, and a page is treated the same regardless of whether a user has spent a long time carefully browsing it, or if they have quickly moved on to the next page. In this study, a new analysis model is proposed that can analyze a user's willingness to purchase each product based on their relationship between their user's browsing behavior data including their browsing time and the purchasing behavior, based on the understanding that their browsing time for each page is also an important factor in the customer's browsing behavior. Specifically, the present paper proposes a new modified Latent LSTM Allocation model that is able to consider not only the information related to a user's browsing and purchasing behaviors but also the time intervals for their browsing behaviors on individual pages on the target EC site. The proposed model enables to the extraction of groups based on purchasing tendencies considering users' browsing times for each page. Finally, the proposed model is applied to a case study for an EC site that handles fresh flower products. The obtained findings are then analyzed by applying the proposed model to actual usage history data on the flesh flower EC site.
