In 2001, as part of a research group project called “Thursday Meeting” that Dr. Masakatsu Murakami has been organizing since the early 1990s, we conducted three public interviews with Dr. Chikio Hayashi about his academic experiences and philosophy of statistics and science.

The first and third interviews were about the history of Dr. Hayashi’s research and the invention of the Hayashi method of quantification, which have already been published in the Japanese Journal of Bahaviormetrika (Takahashi, 2004) and The Society and Survey (Takahashi, 2012). This is the second open interview with Dr. Chikio Hayashi about his experiences and thoughts on collaborative research with researchers in various fields such as criminology, linguistics, public opinion research, psychology, economics, marketing research, forestry, medicine, and economics. Since the late 1980s, Dr. Hayashi has been putting forward the idea of “Science of Data” and the importance of looking at the total process from the design of data collection, pre-testing, data collection, data cleaning, data aggregation, data analysis, and summary for policy making, and the cyclical improvement is especially emphasized. This article should make readers aware of the major difference between his “Science of Data” and the superficial techniques of “data science” that are currently in vogue.

So far, from the standpoint that the vulnerability of cognitive function causes the damageof special fraud in the elderly, we have derived the logic that can measure and judge the vulnerabilityof cognitive function, and formulate it as a fraud resistance judgment formula. Incontrast to the logistic regression which is used in our judgment formula, this study proposesan improvement to the judgment formula using a fast-and-frugal decision tree. By using adecision tree that applies fast-and-frugal heuristics, the accuracy of fraud resistance judgmentis improved, and the comprehension of advice based on the judgment is also improved.Furthermore, we discuss the problem of imbalanced data that occurs when analyzing specialfraud victim data mixed with general data and the problem of small samples that inevitablyaccompanies special fraud victim data.

Numerous studies have been conducted on bankrupt companies, wherein researchers have often employed financial numerical data for analyzing the continuity of such companies. Additionally, several studies have been conducted in which the signs of bankruptcy have been quantitatively identified through textual data from, for example, economic reports, bulletin boards, and securities reports for corporate evaluation and stock price forecasting.

In this study, discriminant analysis was applied for identifying corporate bankruptcy by utilizing both financial numeric data and textual data. The data used in this study from a part of the Japanese annual securities report. Furthermore, a data set was created for the textual data through text mining.

The results indicate that, the macro-average F-score can reach the value of 0.941 uponincorporating both numerical and textual data, which is significantly higher than the macroaverageF-scores obtained while solely utilizing financial numerical data (0.880) or financialtextual data (0.895).