In this study, we propose a method to visualize the relationship between weather conditions and sales of soft drink categories in order to understand the characteristics of soft drink needs. We attempted to detect information on demand fluctuations by applying Non-negative Matrix Factorization, a time-series soft clustering method, to features created based on information on multiple demand quantities. By visualizing the relationship between the detected prescribed matrix and the discomfort index that indicates weather conditions, we have made it possible to detect relationships between weather conditions and demand fluctuations that cannot be grasped by time-series correlation. The analysis was performed using POS data related to the soft drink industry provided by Nikkei Shimbun at the Reiwa 5th Data Analysis Competition.

View full abstract

  In this study, based on purchase data from an e-commerce site, we use a control chart as a model to observe the sales amount (hereinafter referred to as “sales") transition of products. We propose a method for discovering products with distinctive sales trends. Specifically, an estimated range of sales is obtained with reference to past purchase data, and cases in which sales exceed the range are identified as “sales increase" and cases in which sales fall below the range are identified as “sales decrease". The number and timing of “sales increase" and “sales decrease" are used to discover products that have characteristics of sales increase or decrease. An application was created to put this method into practice. The application is used to discover products with characteristic sales increase/decrease and to visually grasp sales trends. We show that the proposed method can discover products with unique sales trends.

View full abstract

  Stated preference methods are a family of question-based social survey approaches to elicit people's preferences for goods and services and their characteristics. As the theoretical background of stated preference methods is consistent with microeconomics, the methods have been applied to a wide range of practices and empirical studies. However, when beginners apply these methods to their empirical tasks, they may have to use different software packages according to the methods. Herein, the authors have developed R packages for three main variants of the stated preference methods, namely contingent valuation, discrete choice experiments, and best-worst scaling, to make R a platform for learning and applying these methods. This paper outlines the three variants of stated preference methods, describes the roles and features of our R packages, and reveals the diffusion of such packages.

View full abstract

  Kriging is a widely used spatial statistical analysis method applied in various fields. In this study, Kriging techniques were employed to estimate chlorophyll-a (chl-a) concentrations in cloud-covered areas of satellite remote sensing data. Specifically, chl-a concentration data from the surface layer of the Gulf of Finland were interpolated using this method, and the spatial distribution of chl-a concentrations was modeled accordingly. The analysis was conducted using the “gstat” package in the R programming language, which enables variogram modeling and supports both Ordinary and Universal Kriging for spatial prediction. Preprocessing for the Kriging analysis involved data manipulation with QGIS, a free and open-source Geographic Information System (GIS). A comparison of the results showed that Universal Kriging, which incorporates distance from the coastline and bathymetry as covariates, outperformed Ordinary Kriging based solely on observational data, achieving lower mean square errors. Furthermore, Universal Kriging provided a more accurate representation of the spatial distribution of chl-a concentrations.

View full abstract