A survey of recommender systems with multi-objective optimization

Recommender systems have been successfully applied to diverse areas, but their use in the healthcare domain is still rare. One challenge of applying recommender systems to this domain is related to legal concerns about the consequences of provided recommendations. In this work, we advance an expert-in-the-loop, explanation-first approach to tackle this challenge in a specific healthcare niche: gerontological care. A key aspect of the proposed approach is that both recommendations and explanations reflect the structured questionnaire employed by the practitioner to identify patient needs. Another key aspect is that a clinical dataset of patient assessments and respective assigned interventions is used to estimate effects of alternative interventions during the recommendation process. To evaluate the feasibility of this modelling approach, an explanation style was designed with help of practitioners, and a recommendation model was devised and evaluated against a clinical dataset, which was collected by a partner research group working on gerontological primary care. When compared to other traditional recommendation models, the attained precision was competitive across several evaluation conditions. The results suggest that the proposed approach is feasible and may point new ways of adapting recommender systems to play an assistive role in health care.

For multi-objective optimization under condition of complicated objective function, the data processing in the evaluation is sometime tediously long, special algorithm is needed to be adopted. Since the remarkable features of uniform distribution of test points within the test domain and the small number of tests, fully representative of each point, and easy to perform regression analysis, the uniform test design method is hybrid with the “intersection” method for multi-objective optimization to simplify the complicated data process in evaluation first. Furthermore, the "intersection" multi-objective optimization methodology is combined with sequential uniform design so as to get a more precise approximation for solving multi-objective optimization problem, the procedure for searching optimum of the “intersection“ multi-objective optimization methodology with sequential uniform design algorithm is put forward. A multi-objective optimization of linear programming problem with three variables is taken as our example, which involves a maximum for one objective and a minimum for another objective. The result for applying the novel approach to the example indicates the effectiveness of current hybrids.

Decision making for probabilistic fatigue inspection planning based on multi-objective optimization

How to automatically obtain cross-features with different weight values is a significant issue in the research of recommendation models. Traditional recommendation models cannot automatically learn the deep-level features of users and items to obtain cross-features. The mixed processing of dense numerical features and sparse categorical features will result in more information loss during dimensionality reduction. Cross features occupy the same weight in the recommendation process, which will lead to the non-prominence of critical features and reduce the accuracy of model recommendations. This paper proposes a personalized recommendation model (MSRN) for self-attention perceptron with automatic feature correlation. The model first processes the numerical features and category features in double towers to reduce the loss of feature information. Numerical cross-feature matrix and category cross-feature matrix use multilayer perceptrons to automatically mine the hidden knowledge and relationships between features. The model uses the Hadamard product to process it to obtain the cross feature matrix and uses the self-attention mechanism to assign different weights to the extracted cross-features. The experimental results on the public data set show that the recommended evaluation indicators of this model, MAE, and RMSE, are better than the current advanced recommendation models and have better accuracy and stability.

The conventional way of recommendation is exclusively focused on end-user personalization and accuracy improvement. Nowadays, business models are adapting a multi-sided fairness policy, and the multi-sided fairness could be achieved with a multi-stakeholder recommendation model. Recent studies suggested that the existing end-user focused recommendation model lacks in justifying other stakeholders’ needs and does not create any personalized recommendation environment. Thus, this research proposes a multi-objective optimization-based personalized recommendation framework suitable for preferences for every stakeholder in the system. Particularly this work considers the two stakeholders in the proposed recommendation model producer and system. We design two dedicated objective functions to address the preferences of producer and system stakeholders. The producer’s objective function is to enhance the producer’s exposure in the recommendation list, and the objective function of the system is intended to observe balance and fair recommendation. Next, to optimize the conflicting utilities of stakeholders, we propose a stakeholder optimization-based recommendation framework. The proposed model aims to improve utility gain and establish a balanced trade-off recommendation solution between stakeholders. The proposed model is evaluated based on cumulative utility gain against the baseline algorithms. The evaluation result describes the significant improvement in cumulative utility gain when the standard algorithms are altered with proposed objective functions.KeywordsMulti-stakeholder recommendationMulti-objective recommendationMulti-sided fairnessPersonalization

Traditional recommendation models usually consider only one type of user behavior, some recommendation models focus on user behaviors that frequently occur (e.g. click, view), which have more interaction data but does not accurately reflect the users’ interests, and another types of recommendation models focus on user behaviors which reflect a stronger user preference (e.g. purchase), but has problem of data sparcity. Therefore, in order to improve the performance of the model, multi-behavior-based recommendation is a suitable solution. However, early efforts towards multi-behavior recommendation either do not utilize external knowledge to encode explicit relationships, or cannot fully explore the semantic information of user behavior. Hence, in this work, we solve these problems by constructing a heterogeneous interaction graph and proposing a new multi-behavior recommendation model named KMBFD(short for Knowledge-based Multi-Behavior Recommendation with Factor Disentanglement), which encodes knowledge-based information by multi-relational item graph and user graph and captures semantic information of multi-behavior interaction by a feature aggregation module with factor disentanglement method that is used to disentangle multi-dimensional latent factors hidden in the training data, which can enhance expressiveness and interpretability.

The information cocoon is a common phenomenon in recommendation systems, characterized by the continuous delivery of homogeneous content to users over prolonged interactions, resulting in reduced information diversity and a constrained cognitive scope. This paper begins by analyzing the multidimensional mechanisms behind its formation, systematically examining the roles of algorithmic, social, and human-computer interaction dimensions in the emergence of the cocoon effect. Building on this, it reviews the evolution of mainstream recommendation algorithms, including memory-based and model-based approaches, hybrid recommendation methods, and diversity-enhancing strategies. Their limitations in balancing accuracy and diversity are assessed. Furthermore, a dynamic systems modeling perspective is introduced, conceptualizing the recommendation system as a bidirectionally coupled dynamic process between the "user" and the "system," and exploring the periodic and synchronous characteristics of interest evolution. Building on the theories and applications of oscillator and synchronization models in physics, artificial systems, and neural networks, this study proposes integrating the concept of oscillation into machine learning and recommendation models. By introducing controlled phase perturbations and frequency modulation, the aim is to break information cocoons, enabling the periodic diffusion of content exposure and the dynamic expansion of the interest space. The oscillator mechanism is systematically incorporated into the research framework of recommendation systems, and, in light of recent developments in the application of oscillatory concepts in deep learning, feasible research directions for the future are proposed.

Recommender Systems (RS) influence everyday decisions, yet most remain optimized for short-term engagement or commercial gain. RS4SD aims to shift this focus by exploring how RS can contribute to sustainable development through behavioral change and nudging strategies. Aligned with the UN Sustainable Development Goals (SDG), RS4SD will highlight applications that promote responsible consumption, sustainable mobility, healthy eating, and digital well-being. In particular, we will focus on how AI and RS can be designed to foster sustainable behaviors through multi-objective optimization and ethically aligned interventions. These objectives are directly tied to the UN SDG, and we welcome all contributions showcasing RS in support of these goals. A central theme of the workshop is the integration of behavioral science and AI to design interventions that guide users toward more sustainable and healthier choices while preserving individual autonomy. Topics of interest include multi-objective recommendation, health-aware RS, eco-friendly product and tourism RS, as well as novel evaluation metrics that go beyond accuracy to capture societal impact. RS4SD will bring together researchers, stakeholders and practitioners from RS, AI, sustainability, and behavioral science to share models, datasets, frameworks, and real-world use cases. The workshop encourages interdisciplinary collaboration and aims to build a community dedicated to responsible, behavior-aware RS that benefit both individuals and society.

A multiobjective optimization methodology under uncertainty is developed. Model-based reliability analysis methods are used to compute the probabilistic constraints. Both component-level and system-level reliability analyses are included in the optimization formulation. Several approaches in multiobjective optimization are used to generate the Pareto frontier for decisionmaking. The proposed method is successfully applied to a car door subsystem, in which two important quality issues, door closing effort and wind noise, are of concern and a balanced design needs to be achieved.

In chemical engineering design, process plants are often capable of turndown and flexible operation to maximise profitability depending on demand and price signals, and the optimisation of these processes needs to take into account the variability in the plant operation. By performing the optimisation with flexible operating conditions the overall plant performance can be improved. The optimisation process makes use of two computational modelling techniques: multi-objective optimisation (MOO) and surrogate modelling. Multi-objective optimisation allows for the optimisation of two or more competing variables simultaneously. In the cases studied here, the long individual simulation time coupled with the large number of individual simulations required meant that the total simulation time was prohibitive for dynamic MOO. In order to reduce computation time surrogate models of the processes are developed using artificial neural networks. A statistical dynamic modelling framework within a multi-objective optimisation methodology is developed to optimise a process plant’s performance. Here the dynamic modelling is used to optimise the economic and environmental performance of natural gas combined cycle power plant systems fitted with carbon capture and storage (CCS) and solar thermal energy (STE) in a time-integrated MOO framework. This framework allows for variable process inputs and flexible operation based on external signals, such as product price. By introducing a probabilistic approach to the modelling of flexible operation allows for more data to be included in the analysis. The methodology for dynamic modelling of flexible operation could be applied to other industrial processes that face similar dynamic economic and/or environmental conditions. From a CCS standpoint dynamic MOO was used to show that variable capture can be used to make the CCS plant more profitable over steady state operation. At a capture rate of 80 % the NPV of a gas turbine combined cycle with CCS improved by 38 m$AU by utilising flexible operation of the capture process. This also allows for various economic scenarios to incentivise the reduction of CO₂ emissions from natural gas based power plants, with the use of CCS and STE used to reduce the greenhouse emissions. The economic strategies of a carbon price and governmental grants are investigated to determine the levels of policy help to incentivise implementing CO₂ reduction technologies. In order to achieve a financial incentive for limiting CO₂ emissions from a combined cycle power plants equipped with CCS a government co-investment grant of 50 % of the capital costs and a carbon price of 83 $AU/tCO₂ is required.

This article presents a comprehensive and objective systematic review of existing research on recommendation systems with regards to core theory, latest studies, various applications, current attitudes, and potential future applications. The research is mainly based on exploring professional peer-reviewed studies and articles and using their abstracts to create a comprehensive and unbiased review of existing research. The following search terms were used to identify articles and studies for the research: recommendation systems; recommender systems; core theory of recommender systems; current attitudes towards recommendation systems; latest studies on recommendation systems; applications of recommendation systems; potential studies on recommendation systems; and future potential applications of recommendation systems. The research also used the advanced search filter to locate recent studies for comparison by limiting the search by year to find studies published from 2021 onwards. Most literature on this area highlights the importance of recommendation systems in almost all aspects of modern life. Specifically, recommendation systems have become critical components in business, health care, education, marketing, and social networking domains. Additionally, most studies identified reinforcement of learning and deep learning techniques as significant developments in the field. These techniques form the backbone of most modern recommendation systems. The primary concern that could hinder further evolution systems is their consequent filter bubble effects which many studies showed to be problematic. Healthcare is a central area that shows tremendous potential for these systems. Although recommender systems have been implemented in this domain, there remains a lot of untapped potential that, if unleashed, could revolutionize medicine and healthcare. But the problems facing these systems have to be tackled first to establish trust. Keywords: Recommendation systems, Recommender systems, Deep learning, Reinforcement learning DOI: 10.7176/CEIS/13-4-04 Publication date: August 31 st 2022

Cyber-physical systems (CPS) are often characterized as smart systems, which intelligently interact with other systems across information and physical interfaces. Cyber-physical systems (CPS) are fitting into modern society. CPS integrate multiple techniques, including distributed computing, communication and automatic control, to support variety of intelligent services and applications in many fields, such as transportation, healthcare, entertainment and city infrastructure. Recommender systems in CPS, which always provide information recommendations for users based on historical ratings collected from a single domain only, suffer from the data sparsity problem. Recently, several recommendation models have been proposed to transfer knowledge across multiple domains to alleviate the sparsity problem, which typically assumes that multiple domains share a latent common rating pattern. Recommender systems always results into faster and efficient options/choices as per user’s demands. This technology helps us in retrieval and access of different services like in Healthcare sector, Education, E-Commerce, etc. During the last few decades, with the rise of Youtube, Amazon, Netflix and many other such web services, recommender systems have taken more and more place in our lives. From e-commerce (suggest to buyers articles that could interest them) to online advertisement (suggest to users the right contents, matching their preferences), recommender systems are today unavoidable in our daily online journeys. In this paper, a proposed system architecture focusses on IoT applications utilizing recommender systems for offering number of services to users. Also, various Cyber Physical Recommender Systems are proposed to offer standard quality of service (QoS) for various IoT applications.

Recommender systems are typically biased toward a small group of users, leading to severe unfairness in recommendation performance, i.e., User-Oriented Fairness (UOF) issue. Existing research on UOF exhibits notable limitations in two phases of recommendation models. In the training phase, current methods fail to tackle the root cause of the UOF issue, which lies in the unfair training process between advantaged and disadvantaged users. In the evaluation phase, the current UOF metric lacks the ability to comprehensively evaluate varying cases of unfairness. In this paper, we aim to address the aforementioned limitations and ensure recommendation models treat user groups of varying activity levels equally. In the training phase, we propose a novel Intra- and Inter-GrOup Optimal Transport framework (II-GOOT) to alleviate the data sparsity problem for disadvantaged users and narrow the training gap between advantaged and disadvantaged users. In the evaluation phase, we introduce a novel metric called ?-UOF, which enables the identification and assessment of various cases of UOF. This helps prevent recommendation models from leading to unfavorable fairness outcomes, where both advantaged and disadvantaged users experience subpar recommendation performance. We conduct extensive experiments on three real-world datasets based on four backbone recommendation models to prove the effectiveness of ?-UOF and the efficiency of our proposed II-GOOT.

Many simple decision processes are based on a single objective such as minimizing cost, maximizing profit, minimizing runtime and so forth. However decisions must be made in an environment where more than one objective, constrains the problem, and the relative value of each of these objective is different. Such problems where multiple objectives are to be optimized are known as Multi-objective optimization (MOO) problems. The problem becomes challenging in case of mutually conflicting objectives i.e. the optimal solution widely varies with the shifting of focus from one objective to the other while all of them are quite relevant for the problem. MOO is used in various fields viz. Production Planning, Structural Design etc. There are various genetic approaches to solve multi-objective optimization problems like Non-dominated sorting GA (NSGA-II), Strength Pareto Evolutionary Approach (SPEA) etc. Each of these methods has their pros and cons and none of these is found to be perfect. In this paper we present a comparative study of the various available methods based on evolutionary genetic algorithms for Multi-Objective Optimization on different performance metrics.

Recommendation system plays a significant role in helping people to get effective information from mass data. Traditional recommendation systems focus on the recommendation accuracy, which is not sufficient. In this paper, we also consider the various needs of users to achieve more diverse recommendation. However, accuracy and diversity are two conflicting goals for recommendation system. Hence, we model the recommendation system as a multi-objective optimization problem, and aim to find tradeoff solutions between the two goals. Because the rating matrix is rather sparser in recommendation, we first use singular value decomposition to get the recommendation list, then multi-objective immune algorithm is used to optimize it. The experimental results illustrate that the proposed algorithm can get more diverse and accurate recommendation results.