User Behavior Preference Research Articles

LSAB: User Behavioral Pattern Modeling in Sequential Recommendation by Learning Self-Attention Bias

Since the weight of a self-attention model is not affected by the sequence interval, it can more accurately and completely describe the user interests, so it is widely used in processing sequential recommendation. However, the mainstream self-attention model focuses on the similarity between items when calculating the attention weight of user behavioral patterns but fails to reflect the impact of user sudden drift decisions on the model in time. In this article, we introduce a bias strategy in the self-attention module, referred to as Learning Self-Attention Bias (LSAB) to more accurately learn the fast-changing user behavioral patterns. The introduction of LSAB allows for the adjustment of bias resulting from self-attention weights, leading to enhanced prediction performance in sequential recommendation. In addition, this article designs four attention-weight bias types catering to diverse user behavior preferences. After testing on the benchmark datasets, each bias strategy in LSAB is useful for state-of-the-art and can improve the performance of the models by nearly 5% on average. The source code listing is publicly available at https://gitee.com/kyle-liao/lsab .

Hierarchical Constrained Variational Autoencoder for interaction-sparse recommendations

Predicting potential user behaviors is of great importance in recommendation systems. Existing behavior prediction works mainly aim to construct the user behavior preference based on a large amount of user interaction data, but the prediction performance deteriorates when facing the users with a few interactions. A common and effective approach to alleviate the interaction-sparse problem is to incorporate the auxiliary information as much as possible, in which the hybrid (Variational Autoencoder) VAE method reports the optimal performance with the advantages of non-linear modeling and comprehensive integration. However, the hybrid VAE only focuses on the attribute data, which limits the richness and robustness of the user preference representation in interaction-sparse scenario. In this paper, we propose a novel hybrid VAE model, named Hierarchical Constrained Variational AutoEncoder Jointly Attributes and Neighbors (HiCVAE), to tackle the above limitations. Specifically, HiCVAE additionally introduces fine-grained neighbors’ behaviors as auxiliary information to enhance the richness and robustness of the user preference representation on the basis of the attribute data, by learning a hierarchical prior instead of a standard Gaussian prior. The prior on the attribute data is responsible for the community preference relevant to a group with the same attributes. Further, the conditional prior over the neighborhood data can capture the personalized information and provide further constraints for the new user preference representation. In addition, a Preference Consistency Loss is proposed to restrict the reconstructed interest to be consistent with the user’s real interest and eliminate the irrelevant information on the neighborhood data. Thorough and extensive experiments conducted on two real-world datasets demonstrate that the HiCVAE significantly outperforms the state-of-the-art methods in various interaction-sparse scenarios (sparse, highly sparse and extremely sparse).

Dynamic Multi-view Group Preference Learning for group behavior prediction in social networks

Group behavior modeling is an important research topic in the field of social network analysis. Existing methods regarding this topic can only learn the static group preference, ignoring the multiple characteristics of the group behavior, so they cannot model the group behavior in a complete way. In this paper, we propose a Dynamic Multi-view Group Preference Learning (DMGPL) model for group behavior prediction in social networks. Firstly, an area-aware user dynamic preference extracting module is developed concerning user representation learning, which can integrate the dynamic user behavior preference and the corresponding group structural information into group members’ representations. Then, to simultaneously obtain stability, propensity, potentiality, and heterogeneity of the group behavior, Multi-view Group Preference Aggregation (MGA) is designed for group behavior modeling. In MGA, various aggregation strategies are used to capture different kinds of group behavior properties, which can achieve the effect of highlighting different properties in different scenarios. Moreover, considering the influence of the member’s size on group behavior, we define the group scaling degree and perform group representation scaling to make MGA more flexible. Finally, extensive experiments are performed on two real-world datasets, and the results show that the accuracy of group behavior prediction by DMGPL improved by about 10% compared to the baseline models.

The impact of interaction on continuous use in online learning platforms: a metaverse perspective

PurposeAs online learning is the embryonic form of education in the metaverse, it is extremely important to explore the behavioral preferences of users. The aim is to explore the impact of interactive features on continuous use in online learning and to further explore what kind of interaction mode should be constructed for different types of students to obtain the best educational experience.Design/methodology/approachThe study developed an empirical model and used a real-world dataset to test hypotheses. Specifically, the interaction in online learning is analyzed from different dimensions, including the interaction intensity of multiple subjects, the immersion of interactive technology, the timeliness of interactive feedback, and the fun in interaction.FindingsThe authors found that the intensity of interaction, immersion, timeliness of feedback and fun in the interaction all had significant positive effects on continuous use. Among them, the most important is the interaction between teachers and students. With the growth of user grades, the role of parents in the interaction is getting smaller and smaller, and the fun in the interaction is gradually becoming unnecessary. For high school students, gamified interactions can even have a negative impact. In addition, from the perspective of gender, males prefer immersive interaction, while females pay more attention to themselves and have negative feedback on fees.Originality/valueThe authors deepened the interaction and summarized the impact of different interactive features on continuous use in online learning platforms. The authors focused on the impact of the immersive experience brought by the application of interactive technology, which can confirm the user behavior preferences of online learning in the context of the metaverse. The research also provides a reference for online learning institutions to set up course interaction modes and targeted marketing programs.

EV-Station-Grid Coordination Optimization Strategy Considering Psychological Preferences

This paper proposes the electric vehicle (EV)-station-grid coordination optimization strategy considering user preferences, which regulates the charging behaviors of EV users from the user side to ensure the stable and safe operation of the power grid. Firstly, the spatio-temporal prediction model of charging load based on speed-temperature is developed. The model of EV power consumption per unit mileage affected by temperature and EV speed is constructed, and the shortest path algorithm is applied to determine the driving paths of EVs so as to judge the charging demand in combination with the state of charge (SOC) of the battery and to determine the charging periods and locations of the EVs, thus obtaining the spatio-temporal information of the charging load. Secondly, a multi-attribute charging decision model considering user preferences is constructed. Fuzzy clustering and rough set theory are applied to mine user behavior preferences, combined with behavioral economics to describe users’ irrational charging decision-making psychology. Lastly, a real-time charging price model considering voltage fluctuation index and user charging cost is constructed to analyze the impact of price on guiding charging behaviors. The simulation results verify the effectiveness and performance of the collaborative optimization strategy.

Hybrid Graph Neural Network Recommendation Based on Multi-Behavior Interaction and Time Sequence Awareness

In recent years, mining user multi-behavior information for prediction has become a hot topic in recommendation systems. Usually, researchers only use graph networks to capture the relationship between multiple types of user-interaction information and target items, while ignoring the order of interactions. This makes multi-behavior information underutilized. In response to the above problem, we propose a new hybrid graph network recommendation model called the User Multi-Behavior Graph Network (UMBGN). The model uses a joint learning mechanism to integrate user–item multi-behavior interaction sequences. We designed a user multi-behavior information-aware layer to focus on the long-term multi-behavior features of users and learn temporally ordered user–item interaction information through BiGRU units and AUGRU units. Furthermore, we also defined the propagation weights between the user–item interaction graph and the item–item relationship graph according to user behavior preferences to capture more valuable dependencies. Extensive experiments on three public datasets, namely MovieLens, Yelp2018, and Online Mall, show that our model outperforms the best baselines by 2.04%, 3.82%, and 3.23%.

Temporal evolution of tagging subnetwork features and motif under different activity levels – take the Q&A community Zhihu as an example

Purpose This paper aims to study the characteristics and evolution rules of tagging knowledge network for users with different activity levels in question-and-answer (Q&A) community represented by Zhihu. Design/methodology/approach A random sample of issue tag data generated by topics in the Zhihu network environment is selected. By defining user quality and selecting the top 20% and bottom 20% of users to focus on, i.e. top users and bot users, the authors apply time slicing for both types of data to construct label knowledge networks, use Q-Q diagrams and ARIMA models to analyze network indicators and introduce the theory and methods of network motif. Findings This study shows that when the power index of degree distribution is less than or equal to 3.1, the ARIMA model with rank index of label network has a higher fitting degree. With the development of the community, the correlation between tags in the tagging knowledge network is very weak. Research limitations/implications It is not comprehensive and sufficient to classify users only according to their activity levels. And traditional statistical analysis is not applicable to large data sets. In the follow-up work, the authors will further explore the characteristics of the network at a larger scale and longer timescale and consider adding more node features, including some edge features. Then, users are statistically classified according to the attributes of nodes and edges to construct complex networks, and algorithms such as machine learning and deep learning are used to calculate large-scale data sets to deeply study the evolution of knowledge networks. Practical implications This paper uses the real data of the Zhihu community to divide users according to user activity and combines the theoretical methods of statistical testing, time series and network motifs to carry out the time series evolution of the knowledge network of the Q&A community. And these research methods provide other network problems with some new ideas. Research has found that user activity has a certain impact on the evolution of the tagging network. The tagging network followed by users with high activity level tends to be stable, and the tagging network followed by users with low activity level gradually fluctuates. Social implications Research has found that user activity has a certain impact on the evolution of the tagging network. The tagging network followed by users with high activity level tends to be stable, and the tagging network followed by users with low activity level gradually fluctuates. For the community, understanding the formation mechanism of its network structure and key nodes in the network is conducive to improving the knowledge system of the content, finding user behavior preferences and improving user experience. Future research work will focus on identifying outbreak points from a large number of topics, predicting topical trends and conducting timely public opinion guidance and control. Originality/value In terms of data selection, the user quality is defined; the Zhihu tags are divided into two categories for time slicing; and network indicators and network motifs are compared and analyzed. In addition, statistical tests, time series analysis and network modality theory are used to analyze the tags.

Intelligent Link Prediction Management Based on Community Discovery and User Behavior Preference in Online Social Networks

Link prediction in online social networks intends to predict users who are yet to establish their network of friends, with the motivation of offering friend recommendation based on the current network structure and the attributes of nodes. However, many existing link prediction methods do not consider important information such as community characteristics, text information, and growth mechanism. In this paper, we propose an intelligent data management mechanism based on relationship strength according to the characteristics of social networks for achieving a reliable prediction in online social networks. Secondly, by considering the network structure attributes and interest preference of users as important factors affecting the link prediction process in online social networks, we propose further improvements in the prediction process by designing a friend recommendation model with a novel incorporation of the relationship information and interest preference characteristics of users into the community detection algorithm. Finally, extensive experiments conducted on a Twitter dataset demonstrate the effectiveness of our proposed models in both dynamic community detection and link prediction.

Exploring Geographic Information Effects for POI Recommendation in LBSNs

With the rapid development of global positioning technology, location-based social networks (LBSNs) have emerged. Massive check-in data provides development possibilities for mining user behavior preferences, and also brings a variety of location-based social services. For example, Point of Interest (POI) recommendation service can help users find their desired places. In this paper, we propose a novel POI recommendation model by integrating users’ preference and locations geographic information such that an efficient and accurate POI recommendation can be achieved. The model explores the spatial distribution of check-in points visited, the relevance of the category of check-in points visited, and the difference in the frequency of check-in points visited to generate a recommendation list for a particular user. Experimental results on the New York and Tokyo datasets from Foursquare show that this model has higher recommendation performance than other comparison models.

Cross-Border E-Commerce Personalized Recommendation Based on Fuzzy Association Specifications Combined with Complex Preference Model

Since cross-border e-commerce involves the export and import of commodities, it is affected by many policies and regulations, resulting in some special requirements for the recommendation system, which makes the traditional collaborative filtering recommendation algorithm less effective for the cross-border e-commerce recommendation system. To address this issue, a simple yet effective cross-border e-commerce personalized recommendation is proposed in this paper, which integrates fuzzy association rule and complex preference into a recommendation model. Under the constraint of fuzzy association rules, a hybrid recommendation model based on user complex preference features is constructed to mine user preference features, and personalized commodities recommendation is realized according to user behavior preference. Compared with the traditional recommendation algorithm, the improved algorithm reduces the impact of data sparsity. The experiment also verifies that the improved fuzzy association rule algorithm has a better recommendation effect than the existing state-of-the-art recommendation models. The recommendation system proposed in this paper has better generalization and has the performance to be applied to real-life scenarios.

Smart caching based on user behavior for mobile edge computing

With the explosive growth in wireless data traffic, it is difficult to support the massive consumption of backhaul link bandwidth. Mobile edge computing (MEC) uses edge caching to store content closer to the end user in case to reduce the computational overhead and latency. Caching decisions depend on user demand, which is associated with user behavior. Popular content shows the preferences of user behavior. Combining user context behavior to benefit caching, our optimization objective is to maximize the hit rate of contents within limited time. This optimization problem can be modeled as a knapsack problem. Even if the context and content popularity are known for all circumstances, it is still NP-hard. We thus propose a heuristic Smart caching algorithm (Smart-caching) within the MEC paradigm. Smart-caching constructs the user behavior preferences to reflect different user demands for content. Based on this construct, Smart-caching employs contextual zoom caching to find the context-aware content popularity with diversity via partitioning and exploring the context space. Our experiments are implemented with a real large-volume dataset, i.e., China Mobile User Detail Record (UDR). The results show that Smart-caching achieves a better cache hit rate and stability with low overhead compared with other related caching algorithms.

Research of Social Network User Behavior Preference Prediction Based on Social Influence

The development of internet and mobile communication technology makes the analysis of user behavior become a significant research area. And the popularity of all kinds of social networks provides massive and real data for analysis and prediction of user behavior. In this paper, considering the users’ social influence in online social network, we proposed a framework of user behavior preference prediction based on social influence. Then given the existing assessment methods, we also analyzed and summarized the ways of evaluating the prediction model.

사용자 행동 패턴 선호도 학습을 위한 퍼지 귀납 학습 시스템

본 논문은 스마트 홈과 같이 다양한 센서 및 제어 네트워크가 밀집되어 있는 유비쿼터스 환경 하에서 복잡한 인터페이스의 사용에 대한 사용자의 인지 부담(cognitive load)을 줄이고, 개인화된(personalized) 서비스를 자율적으로 제공하기 위한 새로운 사용자 행동 패턴 선호도 학습기법을 제안하였다. 이를 위해 지식 발견(knowledge discovery)을 위한 평생 학습(life-long learning)의 관점에서 퍼지귀납(fuzzy inductive) 학습 방법론을 제안하며, 이것은 수치 데이터로부터 입력 공간에 대한 효율적인 퍼지 분할(fuzzy partition)을 얻어내고 일관성 있는(consistent) 퍼지 상관 롤(fuzzy association rule)을 얻어내도록 한다. Smart home is one of the ubiquitous environment platforms with various complex sensor-and-control network. In this paper, a now learning methodology for learning user's behavior preference pattern is proposed in the sense of reductive user's cognitive load to access complex interfaces and providing personalized services. We propose a fuzzy inductive learning methodology based on life-long learning paradigm for knowledge discovery, which tries to construct efficient fuzzy partition for each input space and to extract fuzzy association rules from the numerical data pattern.