Yoko Ishino

Knowledge Extraction of Consumers’ Attitude and Behavior: A Case Study of Private Medical Insurance Policy in Japan

Decision making in the marketing field requires the efficient acquisition and interpretation of consumers’ behavior from questionnaire data. However, the data are usually full of noise, and attributes in the questionnaire data may have a high association with one another. The problem of a domain expert (or analyst) lies in determining a method to acquire accurate and useful knowledge about consumers’ attitude and behavior from the noisy data. This study describes a novel method of extracting useful knowledge from questionnaire data by performing Bayesian network modeling, accompanied by feature selection, which incorporates Cramer’s coefficient of association as an indicator. This method is capable of treating multiple objective variables in one model, handling nonlinear covariation between variables, and solving a feature selection problem. The proposed method was verified by a case study of private medical insurance products in Japan, using real data on health consciousness and private medical insurance.

Method for Getting Parameters of Agent-Based Modeling Using Bayesian Network: A Case of Medical Insurance Market

To date, agent-based social simulation (ABSS) is a popular method to study the behavior of a social system and the interaction of the constituent members of the system. With the development of computer and information technologies, many ABSS approaches have been proposed with wide application. However, the definitive methodology for modeling of the agent’s behavior in ABSS has not been established yet. This study proposes a new methodology of modeling of the agent’s behavior in ABSS using Bayesian network based on the questionnaire survey. This method enables us to simultaneously perform the construction of the agent’s behavior model and the estimation of the internal parameters within the model. This study took a Japanese medical insurance market as an example, since this complicated market deserves detailed consideration. We verified the effectiveness of the proposed methodology by applying the scenario analysis to this case.

Analysis and Modeling of Customer-Perceived Value of Medical Insurance Products

The insurance industry in Japan has undergone drastic changes since the new insurance business law became effective in 1996. Understanding customers’ needs and values in purchasing insurance products has become more important for insurance institutions than ever before, because of not only environmental changes but also changes in customer behavior. This study aims to achieve two goals: (1) to clarify the structure of the customer-perceived value of insurance products related to the medical treatment; and (2) to clarify the dynamics of the belief formation of consumers in a small community as to medical insurance products. Specifically, a Bayesian network modeling based on the consumers’ survey data was used for the former purpose. A multi-agent simulation (MAS) was used for the latter purpose. Finally, the findings that can be utilized in marketing strategies were obtained.

Novel Computational Methodologies for Structural Modeling of Spacious Ligand Binding Sites of G-Protein-Coupled Receptors: Development and Application to Human Leukotriene B4 Receptor

This paper describes a novel method to predict the activated structures of G-protein-coupled receptors (GPCRs) with high accuracy, while aiming for the use of the predicted 3D structures in in silico virtual screening in the future. We propose a new method for modeling GPCR thermal fluctuations, where conformation changes of the proteins are modeled by combining fluctuations on multiple time scales. The core idea of the method is that a molecular dynamics simulation is used to calculate average 3D coordinates of all atoms of a GPCR protein against heat fluctuation on the picosecond or nanosecond time scale, and then evolutionary computation including receptor-ligand docking simulations functions to determine the rotation angle of each helix of a GPCR protein as a movement on a longer time scale. The method was validated using human leukotriene B4 receptor BLT1 as a sample GPCR. Our study demonstrated that the proposed method was able to derive the appropriate 3D structure of the active-state GPCR which docks with its agonists.

Japanese Health Food Market Trend Analysis

So-called health food has attracted increasing attention for more than two decades so that the health food market share has continued to increase. However, definitions and regulations regarding health food differ among countries; thus, comparing multiple countries simultaneously is difficult. This study focused on Japan because it has established health food rules and regulations, and both macro and micro data are available for analysis. In this study, government statistics were first used to determine the Japanese health food market’s macro trend. Then, a questionnaire was administered to identify consumer attitudes and behaviors, and important factors for consumers to purchase health food were determined by conducting chi-square tests using the survey data. The obtained factors corroborated the results of the regression analysis used to determine the macro trend. Then, the following hypothesis was posed: differences between the actual sales of health food and sales estimated by regression analysis are primarily due to advertising activities and government regulations. A straightforward agent-based social simulation (ABSS) was performed to investigate how advertising and government regulations influence health food sales. Several reasonable findings that can be utilized in a future ABSS were obtained.

Characterization of Consumers’ Behavior in Medical Insurance Market with Agent Parameters’ Estimation Process Using Bayesian Network

In medical insurance market as well as other markets, it is not straightforward for an institution to develop effective marketing strategies because consumers’ preferences and the environment surrounding consumers are constantly changing. This paper develops an agent-based model (ABM) of consumer’s behavior in purchasing medical insurance products and analyzes the characterization of consumers’ behavior to establish effective marketing strategies for the products. In general, the information propagation model of purchasing behavior has difficulty estimating the values of parameters only from ordinary marketing surveys, especially in the case of products that require a person to conduct advanced information processing, such as an insurance policy. To tackle this problem, this paper developed a method of estimating the probability parameters of agent’s behavior using Bayesian network based on questionnaire survey data, and then evaluated the effectiveness of the method by applying it to the actual insurance market. In the analysis using ABM constructed, we mainly focus on the power of influence of the sales activity using word-of-mouth communication between consumers. As the result we obtained several key findings regarding marketing strategies that can be utilized in the real marketing of insurance products.

Internet of Things for Health: Japanese Consumers’ Needs for Preventive Healthcare Products

The Internet of Things (IoT) plays a significant role in a broad range of healthcare applications. However, so far this effort is mainly restricted to actual patients actively undergoing treatment. From the perspective of IoT, these kinds of efforts should spread to the preventive activity of healthy people in the not-so-distant future. Healthy people, whose population is much larger than patients, presumably have various latent needs or wants for an IoT-driven healthcare system, compared with the actual patients. Discerning the real needs from the latent needs of healthy people for preventive care is the key to build up an innovative and favorable IoT-based system. This research tackled this problem from the product device side. We utilized multiple analyses including cluster analysis, graphical modeling, and text-mining, in order to extract from survey data real and useful consumer needs, which can be expected to propel the innovation of health-related home electronics. The effectiveness of the proposed method was verified by a case study using practical questionnaire data, while selecting a blood-pressure monitor and a sleep monitor as case study examples.