This study presents a variable selection method for large-scale interaction models that include highly-correlated variables. SPRINTER (Sparse reluctant interaction modelling) is an important solution for pairwise interaction modelling. However, the SPRINTER algorithm initially applies Lasso for the variable selection of the main effect in the first and the last steps, which presents the problem of robust selection for highly correlated variables and leads to interaction selection failure. Therefore, this study proposes a combination method of CHANOL (Convex Hull Approximation of Nearly Optimal Lasso) and SPRINTER (referred to hereafter as, SPRINTER-CHANOL) for data with highly-correlated main effects and interactions. A series of experiments based on synthetic data demonstrated that the proposed method can effectively select the main effects and interactions while having a high true positive rate and a low false positive rate.

In supply chain management, the bullwhip effect leads to an increase in costs and opportunity loss risk, and imposes a significant burden on company management. Therefore, controlling the bullwhip effect in the supply chain is an important issue. The common strategy and information sharing among companies are effective in reducing the bullwhip effect. These effects have been shown in many studies. However, in recent years, in the increasingly complex supply chain, it is very difficult for each company to have an accurate and timely understanding of the information that affects their actions. This makes it difficult to share information and common strategies throughout the supply chain. This study considers a model of cooperation between some companies in the supply chain to share information and common strategies. Studies that aim to reduce the bullwhip effect and the total cost in the supply chains involve demand forecasting, inventory policy and safety stock optimization. The proposed model focuses on improving company strategies by combining multiple strategies and analyzes which policies and actions improve the performance of the entire supply chain. In this study, a multi-stage supply chain model is constructed using multi-agent simulation to identify the optimal strategy by analyzing the effect of cooperation between companies in terms of the bullwhip effect and the total cost. This study identifies a tendency for an the optimal strategy to be used by each company. In addition, the results will be validated by a beer game demonstration to examine effective strategies in a real supply chain.

In modern global supply chain system strategies, the importance of risk management to prevent disruption thereto is increasing. One efficient approach for overcoming this problem is a collaboration between manufacturers and retailers. This study extends a deterministic supply chain model based on the joint economic lot-sizing problem (JELP) to a stochastic programming model, which considers the uncertainty of price sensitive demand. A comparison of the results from numerical experiments based on individual and cooperative models suggests that a manufacturer-retailer collaboration could be beneficial in increasing the overall profit of the entire supply chain system. For instance, total profits were increased by 0.11-12.92% under joint optimization in the range of parameters used in this paper and the growth ratio was considerably higher for the cases in which the demand was more price sensitive. However, the above collaboration can decrease the retailer's profits. For this reason, this study proposes and formulates a collaborative model by adding profit constraints to the above model. As a result, the diminution of the retailer's profits is decreased to 0.0020-0.0164%, that is, the proposed model is able to maximize the total profit by minimizing the reduction of the profits for both the manufacturer and the retailer as much as possible.

In recent years, web-based questionnaire surveys have become widely used. Since the cost of web surveys is low, they are easily used as a basic research tool for planning various policies. However, web surveys often contain selection bias. The results of questionnaires that do not reflect the population to be surveyed may lead to the planning of wrong measures. Therefore, it is important to develop a method to correct the selection bias in order to use web surveys effectively. In previous studies, a correction method has been proposed by modeling the occurrence of selection bias using a selection model. In this study, we propose a new correction method for selection bias based on statistical decision theory. We present an optimal distribution estimation method that minimizes the loss function under the Bayes criterion. We show that the proposed method is not only theoretically optimal, but also has good performance in numerical experiments on artificial data.