It has recently become easier for retail stores to obtain mass customer purchase history data. Analyzing these data, it is possible to understand the preferences of each customer and to use the results for marketing strategies. At the same time, it is important to take into account item sea- sonality in supermarkets planing marketing policies. It is, therefore, necessary to understand whether each customer purchases items based on seasonality throughout the year. In this study, we propose a new latent class model for analyzing customers' purchasing behavior focusing on the seasonality of items, and demonstrate an analysis using our model. Moreover, we show that analysis of customers' purchase behavior using both conventional latent class models and our latent class model provides more useful results than using only one model.

This paper develops an analysis procedure in which an eye tracking-based reverse inference approach is implemented in a restaurants' environment. The analysis procedure enables us to realize the reverse inference approach concept in eye-tracking data analysis. The analytical procedure developed is composed of steps for defining base/reference characteristics of cognitive processes considering key contexts, developing gaze taxonomy and interpreting eye gaze data. We performed a case study where the eye-tracking data of 42 participants at two Japanese Udon restaurants, as well as questionnaire responses, were collected. The potential uses of our analysis procedure are discussed.

Recently, the shared use of mobile tools has been expanding, such as bicycles being pro- vided in areas designated for rental bicycles. However, concentrating mobile tools at a node is a critical issue. To resolve this problem, we use a closed queuing network to calculate the average standby num- ber and average waiting time for the mobile tools at each node, and then select an optimal node from among the candidate nodes. As the model itself assumes stationarity for optimal node placement using the closed queuing network, characteristic quantities are obtained after sufficient time has passed. User trends are fluctuating, and in many cases, the trends cannot be considered using only basic statistics such as averaging. Hence, we simulate this model for optimal node placement in a closed queuing network and obtain detailed information along the time series as information that can correspond to the actual model. To demonstrate applicability to the actual model, we chose Hamanako Garden Park. We installed some mobile tools in the park and evaluated the distribution of the tools when used by visitors. The information obtained from this simulation includes the percentage of time when there is no mobile tool at the node, the percentage of time of node overcapacity, and the degree of influence on the increase and decrease in the number of mobile tools at each node. We propose a new optimal node placement platform by linking the simulation to optimal node placement using the closed queuing network proposed, and expect that it will be applicable to a wide range of models.