In this study, we conduct an analysis of the network structure of users utilizing data provided by an operating company of a childcare application for mothers. First, we define a quantity that represents the depth of the relationship between users by using data on questions and answers within the app. Based on this, we construct a user network. Subsequently, we compute various network features and identify communities within the obtained network. Our findings suggest a high modularity value, indicating the existence of a community structure within the provided data.

View full abstract

  Now we suppose that a manufactory is interested in comparing performance between machines (treatments) of two types. In this study, we propose a sequential multiple comparison procedure that tests sequentially the null hypothesis of coincident of two population means based on data obtained when the p kinds of inspections are administered successively to investigate the difference of performance on two machines. When the null hypothesis is first rejected in the test, as we can detect a difference of performance between two machines, the test terminates. Thus, we can use it as the procedure that is feasible to draw an early conclusion with a little data. The purposes of the study are to develop the sequential test procedure and to investigate the features of the procedure. In the procedure, we derive an integral formula to determine the critical limits for satisfying a predefined type I familywise error rate. Furthermore, we show how to decide a required sample size at each inspection item with guaranteeing the power of test. In our simulation studies, we compare the superiority among the procedures based on three α spending functions under various arrangements of difference of two population means and consider the order of inspection items.

View full abstract

  In professional soccer games, ball touch-based play data (event data) and tracking data on player movements are obtained. It is difficult for coaches, players and fans to utilize the data in numerical form, so it is necessary to visualize the data in a format that is easy for non-specialists to handle. In this paper, we present examples of initiatives to visualize soccer data in Japan and abroad. In Japan, indicators have been developed to evaluate plays that lead to the creation of shooting chances and individual player characteristics using data from the J-League. On the other hand, in Europe and the United States, data scientists and soccer analysts from top clubs are visualizing the tactical value of spaces and conducting educational activities on data analysis in collaboration with universities. The efforts of data experts to visualize soccer data have significant value in ensuring that this data is used more effectively.

View full abstract

  In soccer, it is important that the coach appoints players who contribute to winning the game. It is also important to train such players and pursue prospective players from other teams by scouting. However, the current evaluation of players is often determined by the subjectivity of evaluators. To solve this problem, research is being conducted on indicators that quantify the value of every action. This has made it possible to evaluate players objectively, taking into account all types of attacking actions other than scoring and assisting. On the other hand, in team sports such as soccer, it is clear that there is ``compatibility'' among individual players. Therefore, to improve team performance, it is necessary to not only consider the performance of the individual players, but also the compatibility among the players when deciding on their lineup to participate in a match or when scouting new players. In addition to the performance of individual players, the compatibility among the players varies based on their playing styles. In this study, first the playing style is quantified using non-negative matrix factorization. Then, the relationship between the compatibility among players and their playing styles is analyzed using multiple regression analysis that considers interaction effects.

View full abstract