Recommendation Algorithm Research Articles

Study on a User Preference Conversational Recommender Based on a Knowledge Graph

In the era of information explosion, as a form of personalized recommendation, dialogue recommendation systems provide users with personalized recommendation services through natural language interaction. However, in the face of complex user preferences, the traditional dialogue recommendation system has the problem of a poor recommendation effect. To solve these problems, this paper proposes a user preference dialogue recommendation algorithm (KGCR) based on a knowledge graph, which aims to enhance the understanding of user preferences through the semantic information of the knowledge graph and improve the relevance and accuracy of recommendations. This paper proposes a personalized conversation recommendation algorithm framework for user preference modeling. The framework uses a bilinear model attention mechanism and self-attention hierarchical coding structure to model user preferences to rank and recommend candidate items. By introducing rich user-related information, the recommendation results are not only more in line with users’ individual preferences but also have better diversity, effectively reducing the negative impact of information cocoons and other phenomena. At the same time, the experimental results on the open dataset prove the effectiveness and accuracy of the proposed model in the personalized conversation recommendation task.

The Impact of AI-Personalized Recommendations on Clicking Intentions: Evidence from Chinese E-Commerce

AI-personalized recommendation technology offers more accurate and diverse choices to consumers and increases click-through rates and sales on e-commerce platforms. Yet, data on consumers’ experiences of AI-personalized recommendations and their impact path on clicking intention are scarce. This article addressed these issues through three studies. In study 1, we adopted the Grounded Theory approach to conduct in-depth interviews with 30 Chinese consumers and constructed a scale to measure the impact of consumer experience on click intention. In study 2, we adopted the empirical research method to conduct reliability and validity tests on 347 valid questionnaires to finalize the scale officially. In study 3, based on the SOR theory, we constructed a model and formulated hypotheses and then conducted empirical analysis using 1097 valid questionnaires. We found that the relevance, inspiration, and insightful experiences of AI-personalized recommendations can significantly promote consumers’ clicking intention. Moreover, immersive experience mediates between the former three factors and clicking intention, and technology acceptance mediates between relevance, inspiration, and clicking intention. When consumers perceive a high degree of information privacy infringement, the immersive experience’s positive impact on clicking intention will be weakened. Meanwhile, the promoting effect of technology acceptance on clicking intention will also be inhibited. When information quality improves, the positive impact of technology acceptance on clicking intention will be enhanced. This research fills the gap in the literature on consumers’ experiences of AI-personalized recommendations and clarifies how these experiences affect the clicking intention. It offers valuable insights for e-commerce platforms to continuously optimize personalized recommendation algorithms and boost the click conversion rate of online shopping.

MP-LLAVRec: an agricultural product recommendation algorithm based on LLAVA and user modal preference

MP-LLAVRec: an agricultural product recommendation algorithm based on LLAVA and user modal preference

Building Serendipity into Recommender Algorithms on Online Platforms: Reviving the Chaos and Randomness of the Early Internet Aesthetic

Nostalgic comments about the early internet often praise its random, chaotic aesthetic. By contrast, the major platforms of today are typically viewed as corporate in aesthetic, with a one-size-fits-all profile and personalised recommendations of products. The curated life is the opposite of the serendipitous life. Instead of seeing strange or unusual items, online algorithms have been shown to trap users into ‘you loops’, surrounded by similar users in filter bubbles. Algorithms on major platforms give the illusion that users are getting what they want, yet users often complain of a form of aesthetic sameness, blandness and repetition regarding the products they see and are recommended. In this article, we consider forms of resistance against algorithmic ranking and efficiency, including nostalgia for the early internet. This includes a consideration of early websites, which were more chaotic compared with the algorithmic ranking-based sites of today. We argue that the movement towards a curated life has resulted in a loss of serendipity for the user and a narrowing of the user experience, limiting personal growth and the variety of content to which users are exposed. This ‘narrowing’ effect contributes to a flattening of culture, where culture starts to reflect what the algorithm rewards. We consider ways to shift back to a more random aesthetic, where less-popular items are sometimes shown to the user to increase the serendipity they experience on major platforms.

Putting ‘filter bubble’ effects to the test: evidence on the polarizing impact of ideology-based news recommendation from two experiments in Germany and the U.S.

ABSTRACT Algorithmic news recommender systems (NRS) are present in many digital platforms. A decade after Eli Pariser introduced the infamous ‘filter bubble’ hypothesis, empirical evidence challenges the assumption that recommendation algorithms predominantly create homogeneous opinion environments. Studies indicate that algorithmic platform use may amplify users’ political polarization. Whether the link between platform use and polarization can be causally explained by ideological news filtering, however, is still an unanswered question as rigid causal designs to test the notion of ‘filter bubble’ effects are still largely lacking. To fill this gap, we conducted two experimental studies in Germany (n = 1,786) and the U.S. (n = 1,306) with running NRS selecting news items based on the political orientation and political interest of its users. For both national contexts, results indicate that an NRS with a bias towards users’ political preferences increases ideological polarization among politically moderate individuals, supporting the notion of ‘filter bubble’ effects for this group. No such pattern could be found for affective polarization. Yet, in the German data, affective polarization among moderate users was reduced by a politically balanced NRS (as compared to a randomized news diet), while the same NRS increased affective polarization of politically extreme participants. We discuss the democratic implications of these findings against the backdrop of increasing digital news consumption.

User Behavior Simulation with Large Language Model-based Agents

Simulating high quality user behavior data has always been a fundamental yet challenging problem in human-centered applications such as recommendation systems, social networks, among many others. The major difficulty of user behavior simulation originates from the intricate mechanism of human cognitive and decision processes. Recently, substantial evidence has suggested that by learning huge amounts of web knowledge, large language models (LLMs) can achieve human-like intelligence and generalization capabilities. Inspired by such capabilities, in this article, we take an initial step to study the potential of using LLMs for user behavior simulation in the recommendation domain. To make LLMs act like humans, we design profile, memory and action modules to equip them, building LLM-based agents to simulate real users. To enable interactions between different agents and observe their behavior patterns, we design a sandbox environment, where each agent can interact with the recommendation system, and different agents can converse with their friends via one-to-one chatting or one-to-many social broadcasting. In the experiments, we first demonstrate the believability of the agent-generated behaviors based on both subjective and objective evaluations. Then, to show the potential applications of our method, we simulate and study two social phenomena including (1) information cocoons and (2) user conformity behaviors. We find that controlling the personalization degree of recommendation algorithms and improving the heterogeneity of user social relations can be two effective strategies for alleviating the problem of information cocoon, and the conformity behaviors can be highly influenced by the amount of user social relations. To advance this direction, we have released our project at https://github.com/RUC-GSAI/YuLan-Rec .

From Connection to Isolation: The Role of TikTok Algorithmic Personalization in Computational Media and Cross-cultural Communication

With the popularity of social media platforms, recommendation algorithms play an important role in user content exposure. However, while the personalized recommendation mechanism of the algorithm may promote cross-cultural communication and understanding, it could also strengthen information barriers and affect users' cultural cognition and cross-cultural communication ability. This paper focuses on TikTok's recommendation algorithm and uses a related studies analysis method to explore its impact on multicultural content exposure and cross-cultural communication. According to the findings of numerous related studies, TikTok's algorithm enhances the probability of users being exposed to specific cultural content through strengthening interest - based matching. Moreover, TikTok's algorithm restricts the extensive dissemination of cross - cultural content on account of its "loneliness effect", thus diminishing the depth and quality of cross - cultural understanding. This paper has revealed the two-sided nature of recommendation algorithms in cross-cultural communication and provides inspiration for improving algorithm design and promoting cross-cultural communication.

The Influence of Information Cocoon Effect on the Spread of Feminism: Taking Douyin as an Example

The rapid expansion of social media platforms has provided new avenues for feminist discourse, enabling broader public engagement and visibility. However, algorithm-driven environments like Douyin (Chinese TikTok) introduce the information cocoon effect, which reinforces user biases by tailoring content to their preferences. This study investigates how these algorithmic constraints impact the dissemination of feminist content, shaping public perception and influencing the strategies of feminist creators. Using literature review and case study methods, the paper examines the effects of Douyin's recommendation algorithms on feminist narratives. It explores how feminist creators navigate the platform's algorithmic and audience-driven challenges, addressing the tension between generating visibility and delivering substantive, intersectional feminist discourse. The findings highlight the limitations imposed by the information cocoon effect, which prioritizes superficial, sensational content, thereby narrowing feminist perspectives and perpetuating misconceptions. The study underscores the importance of quality-driven, inclusive content strategies in breaking through these constraints to foster meaningful discussions and promote gender equality in digital spaces.

Deep Learning Aided SID in Near‐Field Power Internet of Things Networks With Hybrid Recommendation Algorithm

ABSTRACTIn the realm of power Internet of Things (IoT) networks, secure inspection detection (SID) is paramount for maintaining system integrity and security. This paper presents a novel framework that leverages deep learning‐based semi‐autoencoders in conjunction with a hybrid recommendation algorithm to enhance SID tasks. Our proposed method utilizes the deep learning‐based semi‐autoencoder to effectively capture and learn complex patterns from high‐dimensional power IoT data, facilitating the identification of anomalies indicative of potential security threats. The hybrid recommendation algorithm, which combines collaborative filtering and content‐based filtering, further refines the detection process by cross‐verifying the identified anomalies with historical data and contextual information, thereby improving the accuracy and reliability of the SID tasks. Through extensive simulations and practical data evaluations, our proposed framework demonstrates superior performance over conventional methods, achieving higher detection accuracy. In particular, the detection accuracy of the proposed scheme is more than 20% higher than that of the competing schemes.

Advancing Book Recommendation Systems: A Comparative Analysis of Collaborative Filtering and Matrix Factorization Algorithms

Introduction: Recommendation systems (RS) are pivotal in enhancing personalized user experiences, particularly in the educational domain, where tailored book recommendation systems (BRS) support individualized learning. Despite their importance, selecting an optimal recommendation algorithm remains challenging, especially in scenarios involving large and sparse datasets. Objectives: This study aims to evaluate the performance of four recommendation algorithms—Random Baseline, Popular, User-Based Collaborative Filtering (UBCF), and Singular Value Decomposition (SVD)—to identify their strengths and tradeoffs regarding accuracy and computational efficiency. The goal is to provide actionable insights for selecting appropriate algorithms for educational platforms and similar applications. Methods: The evaluation employed a dataset containing over 10,000 books, 53,424 users, and 981,756 ratings. The algorithms were assessed using Root Mean Square Error (RMSE) to measure predictive accuracy and training time to indicate computational efficiency. The analysis addressed challenges such as data sparsity and rating biases. Results: The results demonstrated that SVD achieved the highest accuracy (RMSE: 0.950), effectively uncovering latent relationships in sparse datasets. UBCF, with a slightly lower accuracy (RMSE: 1.020), offered a balance between accuracy and computational efficiency, making it suitable for real-time applications. Conversely, simpler algorithms like Random Baseline and Popular exhibited faster training times but significantly lower predictive accuracy, highlighting the limitations of non-personalized methods. Conclusions: This study underscores the tradeoffs between accuracy and efficiency among recommendation algorithms. While SVD is ideal for accuracy-driven applications, UBCF provides a practical alternative for scenarios requiring computational efficiency. The findings have substantial implications for educational platforms and e-commerce, where personalized recommendations enhance user satisfaction and engagement. Future research should focus on integrating deep learning models and expanding evaluation criteria, including user satisfaction and diversity, to further improve the performance of recommendation systems.

An analysis of artificial intelligence automation in digital music streaming platforms for improving consumer subscription responses: a review.

The rapid adoption and evolving nature of artificial intelligence (AI) is playing a significant role in shaping the music streaming industry. AI has become a key player in transforming the digital music streaming industry, particularly in enhancing user experiences and driving subscription growth. Through AI automation, platforms personalize music recommendations, optimize subscription offerings, and improve customer support services. This article reviews the role of AI in driving consumer subscription behaviors on digital music streaming platforms (DMSP), with a focus on recommendation algorithms, dynamic pricing models, marketing automation, and the future of AI in the music industry. Potential challenges related to privacy, ethics, and algorithmic biases are also discussed, showcasing how AI is revolutionizing the music streaming industry.

Optimizing Learning Path Design in Mobile Learning Platforms for Online Courses

With the rapid advancement of information technology, online education, particularly vocational education, has become a vital avenue for enhancing skills and knowledge. Vocational education necessitates flexible and personalized learning path design to accommodate the diverse needs and behaviors of learners. Mobile learning platforms, as a pivotal form of modern online education, provide learners with the convenience of accessing educational resources anytime and anywhere. However, existing methods for optimizing learning paths exhibit notable limitations, primarily in accurately capturing learners’ dynamic behaviors and in providing personalized and intelligent path design. Therefore, how to optimize learning paths based on learners’ dynamic behavior data has become a research hotspot in academia and educational practice. At present, many studies focus on matching analysis based on students’ static characteristics with course content but overlook learners’ behavioral changes and dynamic needs during the learning process. Traditional recommendation algorithms and rule-based path design methods are difficult to cope with complex learning behaviors and diverse learner needs. This study addresses these limitations by proposing an optimization model for mobile learning path design based on multi-view prediction of dynamic student behaviors. The model extracts mobile interaction features from student groups, constructs a multiple mobile behavior collaborative encoder, and employs multiple task label prediction techniques to achieve personalized and intelligent optimization of learning paths. The results demonstrate that this approach significantly enhances the learning efficiency and experience of learners, offering novel insights and technological support for the path design of mobile learning platforms.

Research on Personalized Design and Recommendation Systems for Cultural and Creative Products Based on User Behavior Data

Personalized recommendation technology involves the process of soliciting user data and prescribing a user interest model and active recommendations for some users. Such kinds of creative products involve subjective judgments and intricate patterns in the aesthetics, and it is always difficult to encode algorithms that are capable of understanding and recommending these subjective aspects. In this paper, the beetle swarm optimization algorithm is incorporated into a refined deep neural network model called the beetle swarm-drive refined deep neural network (BS-RDNN) for the analysis of personalized design and recommendation systems for user behaviors. Information about user behavior and feedback was collected as part of this study. The data were preprocessed using Min-Max normalization. t-distributed stochastic neighbor embedding (t-SNE) is employed to reduce the dataset dimensions. The proposed method is discussed with other types of recommendation algorithms. The proposed method is implemented with the aid of Python software. This result proves that the BS-RDNN method has better performance in terms of precision (91.34%), accuracy (93.24%), F1-score (92.23%), recall (92.44%), AUC (91.42%), and overall satisfaction of the users. Therefore, the use of the suggested system to coordinate with the design ideas of different individuals can benefit the field of cultural and creative products.

The Advantages and Disadvantages of Internet Platform Algorithms Promoting Personalized Recommendations From the Echo Chamber to the Deep Impact of Ideological Polarization

The convenience of society has increased significantly with the development of the Internet and social media. Nonetheless, it has also contributed to the rapid spread of information and the intensification of opinion polarization. This paper investigates how personalized recommendation algorithms can contribute to the further polarization of users' views through information filtering, the echo chamber effect, and the spiral of silence mechanism. The findings presented in the research indicate that when participants encounter information that contradicts their existing beliefs, rather than adopting a more neutral stance, they tend to strengthen their initial viewpoints. At the same time, ordinary users chose to remain silent for fear of online harassment, allowing extreme voices to dominate the network. This paper concludes that personalized recommendation algorithms enhance user experience, limit exposure to multiple viewpoints, and exacerbate social fragmentation. To tackle this issue, it is advisable for individuals to enhance their media literacy and proactively interact with a variety of viewpoints. Additionally, platforms and policymakers ought to modify their algorithms to facilitate the spread of diverse information, thereby fostering a more transparent communication landscape

A Neural Collaborative Filtering Recommendation Algorithm Based on Attention Mechanism and Contrastive Learning

The neural collaborative filtering recommendation algorithm is widely used in recommendation systems, which further applies deep learning to recommendation systems. It is a universal framework in the neural collaborative filtering recommendation algorithm, however, it does not consider the impact of different features on recommendation results, nor does it consider the issues of data sparsity and long tail distribution of items. To solve the above problems, this paper proposes a recommendation algorithm based on attention mechanism and contrastive learning, which focuses on more important features through attention mechanism and increases the number of samples to achieve data augmentation through contrastive learning, therefore it improves model performance. The experimental results on two benchmark datasets show that the algorithm proposed in this paper has further improved recommendation performance compared to other benchmark algorithms.

Research on the Analysis of students’ English Learning Behavior and Personalized Recommendation Algorithm based on Machine learning

The goal of this study is to create a personalized recommendation system for English learning resources by analysing students’ English learning behaviors using a Generalized Regression Neural Network (GRNN). For efficient language acquisition, tailored educational support is essential due to the diversity of students’ linguistic backgrounds and learning demands. Performance ratings, personal preferences, and the amount of time students spent on various content categories were among the data gathered for this study on how students interacted with English language learning materials. Our initial analysis of the patterns and discrepancies in the learning behaviors of the pupils involved the use of the GRNN model. Strong insights into the correlations between various aspects and learning results were obtained by the neural network, which was especially well-suited for this investigation due to its affinity for handling non-linear interactions and its low need for preprocessing data. These observations led us to create an individual recommendation engine that recommends educational resources and activities based on each user’s learning preferences and skill level. Using a varied set of students, a controlled study was conducted to assess the efficacy of the individual suggestions. Comparing the preliminary findings to conventional, non-personalized methods of learning, efficiency in learning and student engagement have significantly improved. In addition to showcasing GRNN’s potential for educational applications, this work offers an adaptable framework for adaptive learning systems across a range of academic fields.

A Critical Analysis of AI-Driven Recommendation Algorithms in E-Commerce: Impacts, Opportunities and Challenges

This paper conducted a critical analysis and review of literature, spanning from 2019 to 2023, on how recommendation algorithms affected consumer behavior in e-commerce. The study explored the role of AI-driven recommendation algorithms in influencing consumer behavior and examined the impacts of various recommendations on e-commerce dynamics. By screening relevant articles from the Web of Science Core Collection database, with predefined keywords, the study applied visualization techniques, using the Bibliometrix and Biblioshiny tools within R Studio. The research investigated the different recommendation algorithms, documented in prior studies and their effects on consumer behavior in e-commerce. Our analysis uncovered significant insights into how these AI-driven recommendations shape consumer behavior, highlighting its influence on purchase decisions and product engagement. This study provides actionable insights for e-commerce businesses, to leverage recommendation algorithms, to boost consumer engagement and enhance sales effectiveness. By understanding the impacts of diverse recommendations, businesses can tailor their recommendation strategies to better align with consumer preferences. This paper could contribute to broader understanding of the role of artificial intelligence in consumer decision-making processes and offer valuable perspectives for future research in this field.

Research on Personalized Recommendation Algorithm Integrating Cross-Grained Sentiment and Rating Interaction Features

Research on Personalized Recommendation Algorithm Integrating Cross-Grained Sentiment and Rating Interaction Features

Personalized recommendation algorithm based on electricity users’ usage behavior portrait

Personalized recommendation algorithm based on electricity users’ usage behavior portrait

Algorithm-Driven Identity Politics: Role of Social Media in Polarisation and Mobilisation

This is an essay that explores the role of recommendation algorithms in the communication of identity politics and their sociological and political implications, focusing on how the phenomena of filter bubbles and information cocoons reshape users' information exposure and cognitive structures. By combining Baudrillard's theory of mimetic environments, Bourdieu's theory of cultural capital, and Mead's theory of symbolic interaction, the article reveals how recommender algorithms reinforce group identities and exacerbate social divisions through selective pushing of information. In addition, the proliferation of disinformation exhibits higher dissemination efficiency in algorithm-driven environments, posing challenges to the openness and inclusiveness of the public sphere. The article calls for multi-dimensional interventions in technological design, policy regulation and public education to mitigate the masking effect of algorithmic communication and the negative impact of disinformation, and to provide solution paths for building a pluralistic and open digital public sphere. This article aims to provide theoretical support for understanding the double-edged sword effect of recommendation algorithms in identity politics and practical suggestions for the governance and optimisation of digital society.