Many‐objective multi‐criteria diet optimization problem

Abstract

Healthy eating continues to be a problem that affects a large part of the world's population. This research embraces the significance of dietary and nutritional habits, particularly given the devastating impacts of the Coronavirus Disease (Covid-19) pandemic [3]. With an escalating need for effective strategies, this study delves into the realm of user-oriented, practical, and enduring diet plans as a pivotal tool in encouraging healthier dietary choices. The transformation of the conventional diet problem into a multi-objective optimization one, inspired by the elaborate strategies of foraging animals [4,5], is the fundamental essence of this initiative. This dissertation significantly expands upon the traditional framework through enhanced modelling, enriched datasets, and innovative solution methodologies. Our quest mirrors the complexities of foraging animals, emphasizing the pursuit of numerous objectives in harmony with environmental constraints. By infusing the human diet problem with multiple objectives and constraints, we aim to achieve an optimal balance of various criteria simultaneously. In this pursuit, objectives span the spectrum from maximizing user preferences, meal plan diversity, ingredient availability, and public ratings to minimizing costs, preparation times, and carbon footprints. Concurrently, constraints are rooted in the dietary norms established by healthcare institutions [6], factoring in variables such as gender, age, body index, and activity level. Remarkably important is the unique ability to incorporate user-defined preferences, infusing a personalized touch to the recommended menus. This personalization relies on user-set preference rates, allowing the algorithm to accommodate individual tastes. Further, employing both content-based and collaborative filtering approaches, we efficiently rate unrated dishes using existing user-rated dishes, granting a wider scope for user-centric customization. The shift from individual food items to comprehensive, prepared recipe datasets marks a substantial enhancement, ensuring the generated meal plans are practical and viable. The multivariate nature of our challenge leads to a Pareto front characterized by diverse, optimized menu candidates. Yet this multiplicity necessitates informed decision-making. To address this, we employ a Fuzzy Inference System-based decision-making approach. This intelligent mechanism effectively guides users toward their preferences within the menu spectrum, ensuring a coherent, pragmatic meal schedule. The careful fine-tuning of membership functions using objective value distributions underlies the FIS's capability, granting control not only over user preferences but also over other optimization objectives. By contrasting two pivotal decision-making methodologies—the FIS-based Aggregation approach and the FIS-based Aposteriori approach—we have illuminated the intricate equilibrium between optimization efficiency and computational effectiveness. Consequently, our recommendations emerge not only as theoretically sound but as practically implementable dietary solutions, aligning with users' daily lives. The backbone of this study lies in the comprehensive evaluation of prominent Multi-Objective Evolutionary Algorithms (MOEAs) on the multi-objective diet problem (MODP). The rigorous comparison, based on metrics such as hypervolume (HV) and inverted generational distance (IGD), underscores the proficiency of our model in crafting tailored daily menus for various user archetypes spanning tastes, age, gender, and body mass index. While our primary focus is on individuals pursuing fitness, our methodology exhibits versatility and is readily adaptable to accommodate health conditions by introducing tailored constraints and objectives. Thus, our approach serves as both an individual's and a dietitian's tool, ushering in informed dietary decisions. As we delve into the conclusions drawn from these extensive undertakings, we unveil the successful transformation of a dietary challenge into an optimization journey, heavily informed by the principles of optimal foraging theory. The careful fusion of evolutionary algorithms, constraint handling, and innovative genetic operations has yielded not only a substantial array of solutions but also an optimized path within the Pareto front. The integration of real-world datasets, coupled with the strategic partitioning of the chromosome, has substantiated the practicality and effectiveness of our approach across diverse user profiles. In summation, this exploration has expanded horizons by skilfully tackling the complexities of multi-objective diet optimization from a refined, user-centric perspective. The outcome of these efforts is encapsulated in a decision-making framework empowered by Fuzzy Inference Systems, marking the dawn of a new era in personalized dietary recommendations. The convergence of careful experimentation, innovative algorithms, and insightful findings stand testament to the enduring impact of this research on promoting healthier eating habits and contributing to the scientific discourse surrounding nutritional optimization.