多目標最佳化互動式趨近法於補充飲食攝取選擇之應用

本研究藉由建立多目標最佳化 (Multi-objective Optimization, MOO) 之互動式趨近法 (Interactive Approach) 模型，探討國人對於人體必需營養素其補充飲食攝取偏好之權衡選擇。　　全球迅速發展的同時，國人生活飲食習慣改變，慢性疾病發生機率隨之增長，欲透過改善國人飲食習慣，攝取補充飲食以滿足一天營養所需，進而提升國人健康狀況。本研究以台灣20至40歲女性為研究對象，將額外的價格成本及熱量攝取列為消費者(即決策者)主要考量因素，並利用多目標最佳化互動式趨近法中之加權法 (Weighting Approach) 及達成尺度化函數法 (Achievement Scalarizing Functions, ASF) 與受測者進行互動式之研究。藉由反覆互動的過程，瞭解並修正其對於額外價格成本及熱量攝取之權衡選擇，直至獲得妥協滿意解。經由兩方法之建構模型分析，並以視覺化圖形提供受測者多個介於極值範圍之間之柏雷多解，由受測者選擇最終滿意解，進而探討加權法及達成尺度化函數法與最終選擇之差距。　　案例研究中，結果發現相較於達成尺度化函數法，受測者在加權法中對於兩目標之權重意涵表達直接感受不如達成尺度化函數直接訂定理想目標明確，故與受測者以視覺化圖形所選擇之最終理想解有較大差距，且數據較為分散。判斷受測者對於權重之理念與實際確切數值之概念相比較為模糊，較難掌握其權衡之選擇。而本研究以兩種方法探討改善國人健康品質之飲食決策問題，具有實用性且有效降低研究之成本，並藉由方法之比較，提供決策者一種較可信之方法。然而，本研究案例為線性多目標最佳化之應用，若有非線性現象則需擴充，進行更深入之探討。未來研究建議可依照案例之不同階段選用不同特性的方法，以增進決策方法於實際生活所應用。

關鍵字

多目標規劃；線性多目標最佳化；互動式趨近法；加權法；達成尺度化函數法；補充飲食攝取

並列摘要

In this study, we constructed a Multi-objective Optimization (MOO) interactive approach model to explore the trade-off choices of Taiwanese people for their dietary supplements. 　　With the rapid development of the world, people’s daily life and eating habits have changed, and the probability of the occurrence of chronic diseases has increased. In order to improve the dietary habits of Taiwanese people, through the intake of supplementary diets and thus promote the health of people. This study is based on women aged 20 to 40 in Taiwan, the additional cost and calorie intake are the main considerations for consumers (i.e. decision makers). Using the Weighting Approach and the Achievement Scalarizing Functions (ASF) in the interactive approach of Multi-objective Optimization to interact with the subjects through a process of repeated interactions. To understand and correct the trade-offs between additional price costs and calorie intake, reciprocate the process of interaction until a satisfactory solution is obtained. This study provides information about the model before being tested, and then conducts tests in two ways. The subject chooses the final satisfied solution through the construction model analysis of the two methods, and then discusses the difference between the Weighting Approach and the Achievement Scalarizing Functions. The results of this study, the weighted expression of the Weighting Approach for two goals is not as distinct as the Achievement Scalarizing Functions Approach. So there is vary widely between the final ideal solution selected by the subject, and the data is scattered. Judging the concept of weight is vaguer than the concept of actual exact value because it is difficult to grasp the choice of its trade-off for the subjects. This study explores dietary decision-making problem that improve the health of Taiwanese people in two ways. It is practical and effective in reducing the cost of research. By comparing methods, a more plausible method for decision makers is provided. However, this research case is the application of linear multi-objective optimization. If there is a nonlinear programming, it needs to be expanded. It is suggested that future research can select different characteristics according to different stages of the case to enhance the application of decision-making approach in life.