Due to the recent development of electronic commerce (EC) technology, sale amounts on EC sites have increased rapidly. Large amounts of consumer purchase history data can now easily be obtained, and many data-based approaches to identifying consumer purchase trends have been studied. On the other hand, since diversification of consumer consciousness such as their values and lifestyles are important for marketing research, many approaches to identifying the relationship between “consumers' values and consciousness” and “consumers' preferences underlying their purchase behavior” have been discussed in the existing literature. To investigate consumers' values and consciousness, a questionnaire survey is an efficient tool. It is, however, difficult to collect questionnaire responses from all consumers due to cost issues. Therefore, it is difficult to know the values and consciousness of consumers who have not responded to the questionnaire.
In this study, we propose a new model to identify both consumer purchase behavior and their consciousness based on a latent class model estimated by questionnaire responses and the purchase history data for all users. In addition, we apply the proposed model to actual data to analyze its effectiveness.
Automatic inspection machines for metal products can inspect products in large quantities and at a high speed while separating products without defects from defective products by using unified criteria. However, it is difficult to set accurate criteria to detect minute defects, and this results in misclassification. This study focuses on inspection techniques to improve the classification accuracy for an actual inspection process. There are differences in variables, such as raw materials, light color to capture images, and degree of degradation of light brightness, across different manufacturing plants. Previous studies utilized machine learning, image processing, and statistical methods. However, it is not clear how those different methods should be combined for improving the classification accuracy. Therefore, this study provides a comprehensive survey of an optimal combination of methods to increase robustness and improve classification accuracy. Furthermore, the results also suggest that a combination of methods can achieve classification accuracy and robustness exceeding those of previous methods.
It is very important to guarantee measurement results. For quantitative data, a methodology for obtaining accurate measurements in terms of both trueness and precision has been established according to ISO 5725. On the other hand, no methodology has been established for qualitative data. Moreover, in the previous study, the data are mostly supposed to be binary, rather than multinomial.
The purpose of this study is to validate a method of evaluating a measurement accuracy for ordinal-categorical (multinomial) data using the item response theory (IRT) developed in the fields of psychology and ability evaluation. The IRT model can estimate parameters indicating item difficulty and item discrimination and can evaluate whether the items can be classified appropriately according to the ability of the examinees. Two indicators of accuracy are proposed: correct-ordering probability and consistent-classification probability. We provide the characteristics of this method. The measurement data under different conditions were prepared, applied to the IRT model, and used to calculate the indicators. As a result, the two indicators could show appropriate values in most datasets, but some conditions did not work. In this study, we propose a new model to identify both consumer purchase behavior and their consciousness based on a latent class model estimated by questionnaire responses and the purchase history data for all users. In addition, we apply the proposed model to actual data to analyze its effectiveness.
In a two-level supersaturated design (L2SSD), there are more runs (columns) than factors (rows). This approach is commonly used in screening experiments, where the goal is low-cost identification of active factors (i.e., have major influence on the response). Several previous studies have considered methods for selecting the active factors of L2SSDs under the assumption of effect sparsity, with the Box–Meyer method (BMM) performing the best. However, to overcome various drawbacks of BMM, a modified BMM (MBMM) is proposed by Samset and Tyssedal(1998) to analyze the data of fractionated designs. However, MBMM is not used for L2SSDs so far. Therefore we propose here a modified BMM (MBMM) for analyzing L2SSDs. Although both methods can select active factors, MBMM selects fewer inactive factors compared to BMM. The present simulation results confirm that MBMM is an excellent analytical tool for L2SSDs.
Change point search methods are often used in quality engineering. These methods detect the points in time when important changes can occur, and are applied to various types of time series data. In this study, we search for change points in financial data from Japanese companies across various industries. By using these methods, we can easily and quantitatively identify companies turning points, helping financial institutions analyze financial indicators over a certain period of time.
To obtain sufficient data for the analysis, we use quarterly financial data over the period 2006 to 2015 to apply methods focusing on average values. For the analysis, we select financial indicators such as return on assets, capital adequacy ratio, and rate of sales growth. For each financial indicator, we use the t test to identify change points in data. Then, we run the analysis by grouping data for multiple financial indicators and employing the Hotelling Tsquare test. In this way, we are able to detect turning points that could not be found from a single financial indicator. Finally, we survey the major economic events that occurred around each turning point.