Session Graph Research Articles

Sequence-Aware Graph Neural Network Incorporating Neighborhood Information for Session-Based Recommendation

Session-based recommendation is an important part of many e-commerce websites. Its purpose is to make recommendations based on the interaction behavior of anonymous users in a short period of time. Graph neural network can capture complex interactions in sessions, and they are a class of methods with better performance among existing session-based recommendation models. However, most existing models only learn item features for a single session. At the same time, GNNs are not good at capturing long-distance dependencies in a session, which leads to limited performance improvements for them. To address this deficiency, we propose sequence-aware graph neural network incorporating neighborhood information, named SAN-GNN. We construct a session graph and a neighborhood graph to learn item representations. For neighborhood graph, we propose a neighborhood Information extractor for the neighborhood graph to learn the neighbor information of nodes on the neighborhood graph. For the graph model of the session layer, we propose a session graph attention(SGA) module to learn the item representation of the target session. SGA uses Ta-LSTM to learn sequential dependencies in the target session and uses GCN with an integrated attention mechanism to learn node feature relevant to the target item. Exhaustive experiments on three public real-world datasets show that SAN-GNN outperforms the most advanced existing session-based method.

A Time-Sensitive Graph Neural Network for Session-Based New Item Recommendation

Session-based recommendation plays an important role in daily life and exists in many scenarios, such as online shopping websites and streaming media platforms. Recently, some works have focused on using graph neural networks (GNNs) to recommend new items in session-based scenarios. However, these methods have encountered several limitations. First, existing methods typically ignore the impact of items’ visited time in constructing session graphs, resulting in a departure from real-world recommendation dynamics. Second, sessions are often sparse, making it challenging for GNNs to learn valuable item embedding and user preferences. Third, the existing methods usually overemphasize the impact of the last item on user preferences, neglecting their interest in multiple items in a session. To address these issues, we introduce a time-sensitive graph neural network for new item recommendation in session-based scenarios, namely, TSGNN. Specifically, TSGNN provides a novel time-sensitive session graph constructing technique to solve the first problem. For the second problem, TSGNN introduces graph augmentation and contrastive learning into it. To solve the third problem, TSGNN designs a time-aware attention mechanism to accurately discern user preferences. By evaluating the compatibility between user preferences and candidate new item embeddings, our method recommends items with high relevance scores for users. Comparative experiments demonstrate the superiority of TSGNN over state-of-the-art (SOTA) methods.

A Session-Based Recommendation Model That Integrates the Temporal Sequence of Session Interactions and the Global Distance-Awareness of Items with Graph Neural Networks

In the session-based recommendation algorithm, a better class of methods is to model the complex interaction relationship within the session as a graph structure, and then use the graph neural network to capture the deep features of the item from it. However, most models do not deeply mine the effective information contained in the sequence temporal relationship, nor do they pay attention to the auxiliary contribution of the global distance in different sessions to the item representation. This paper proposes a session-based recommendation model that integrates the temporal sequence of session interactions and the global distance-awareness of items with graph neural networks (TSDA-GNN). First, according to all session sequences, a global graph, a session graph, and a feature graph are constructed. In the global graph, the average “spacing” between items is introduced to represent the association between two items in the global graph. In the session graph, the interaction timing of the item is combined as the weight information in the relationship matrix, and the connection matrix of the session graph is constructed, which can more accurately capture the association relationship between items. Finally, by fusing the feature information of the item, using a graph neural network can deeply mine user preferences hidden in the feature information of the item. The model can not only model the information transmission relationship between items in different sessions, but also further deeply mine the hidden information of global domain and item features. Multiple experiments on two benchmark datasets show that the model proposed in this paper achieves the best results in various indicators.

SMONE: A Session-based Recommendation Model Based on Neighbor Sessions with Similar Probabilistic Intentions

A session-based recommendation system (SRS) tries to predict the next possible choice of anonymous users. In recent years, graph neural network (GNN) models have been successfully applied to SRSs and have achieved great success. Using GNN models in SRSs, each session graph is processed successively to obtain the embedding of the node (i.e, each action on an item), which is then imported into the prediction module to generate recommendation results. However, solely depending on the session graph to obtain the node embeddings is not sufficient because each session only involves a few items. Therefore, neighbor sessions have been used to extend the session graph to learn more informative node representations. In this paper, we introduce a S ession-based recommendation MO del based on N eighbor sessions with similar probabilistic int E ntions(SMONE). SMONE models the intentions behind sessions in a probabilistic way and retrieves the neighbor sessions with similar intentions. After the neighbor sessions are found, the target session and its neighbor sessions are modeled as a hypyergraph to learn the contextualized embeddings, which are combined with item embeddings through GNN to produce the final item recommendations. Experiments on real-world datasets prove the effectiveness and superiority of SMONE.

Jointly modeling intra- and inter-session dependencies with graph neural networks for session-based recommendations

Recently, graph neural networks (GNNs) have achieved promising results in session-based recommendation. Existing methods typically construct a local session graph and a global session graph to explore complex item transition patterns. However, studies have seldom investigated the repeat consumption phenomenon in a local graph. In addition, it is challenging to retrieve relevant adjacent nodes from the whole training set owing to computational complexity and space constraints. In this study, we use a GNN to jointly model intra- and inter-session item dependencies for session-based recommendations. We construct a repeat-aware local session graph to encode the intra-item dependencies and generate the session representation with positional awareness. Then, we use sessions from the current mini-batch instead of the whole training set to construct a global graph, which we refer to as the session-level global graph. Next, we aggregate the K-nearest neighbors to generate the final session representation, which enables easy and efficient neighbor searching. Extensive experiments on three real-world recommendation datasets demonstrate that RN-GNN outperforms state-of-the-art methods.

Session Recommendation Based on Edge Information Clustering

Traditional session recommendation mainly uses the time sequence of users clicking items to construct a user session graph, which often ignores the similarity and differences between user groups. To improve the effect of recommendation, an E-SGNN (E-SGNN, Edge-Session Graph Neural Network) method combining edge information clustering and session recommendation is proposed. Firstly, similar users are clustered by edge information and divided into different session user groups. After extracting the data features of the user site relationship graph in the session, it is reset and updated through the gated graph neural network (GGNN); Secondly, a self-attention mechanism is introduced to adjust the proportion of users’ current preference and historical preference; Finally, the ranking score is obtained through linear transformation and softmax classifier. The higher the score, the more obvious the user’s preference for the item. Experiments show that compared with session-based graph neural network and cross-session information recommendation, the E-SGNN algorithm proposed in this paper has a significant improvement in recall rate and average reciprocal ranking. When the three edge parameters are combined, the recall rate reaches 98.97% and the average reciprocal ranking reaches 45.77%.

Interval-enhanced Graph Transformer solution for session-based recommendation

In many online recommendation services (e.g., multimedia streaming, e-commerce), predicting user’s next behavior based on anonymous sessions remains a challenging problem, mainly due to the lack of basic user information and limited behavioral information. The existing typical methods either model user behavior sequences based on RNN or capture potential relationships among items based on GNN. However, these pioneers ignore the importance of different time intervals in the behavior sequence, which implies the user preferences and makes the session sequence more distinguishable. Towards this end, we contribute an Interval-enhanced Graph Transformer (IGT) solution for the session-based recommendation, which takes both item relations and corresponding time intervals into consideration. Specifically, IGT consists of three modules: (i) Interval-enhanced session graph, which constructs all session sequences as session graphs with time intervals; (ii) Graph Transformer, which is embedded with time intervals is adopted to learn the complex interaction information among items. Among them, we design various time interval embedding functions, which can be flexibly injected into the framework; (iii) Preference representation and prediction, which uses an attention network to fuse the user’s long-term preferences and short-term preferences to predict the next click. By conducting extensive experiments on the DIGINETICA, YOOCHOOSE and Last.FM three real-world datasets, we validate that IGT outperforms state-of-the-art solutions.

Enhanced graph neural network for session-based recommendation

A key problem of session-based recommendation is to predict items of interest based on anonymous user’s current interaction session. To improve the prediction accuracy, existing methods mainly adopted two paradigms, i.e., sequence pattern and co-occurrence pattern. Though decent results were achieved, these methods still could not better simulate the real intention of users, or consider the complex pattern transformation between interacted items. To address these limitations, we propose a novel framework called Enhanced Graph Neural Network (E-GNN) for session-based recommendation, where the interaction sequence of all anonymous users is modeled as a Weighted Global Item Graph (WGIG) and the current interaction session of a target user is modeled as a Local Session Graph (LSG). Specifically, the WGIG is first constructed to produce a weight matrix effectively revealing the connection patterns of all intereacted items. The LSG is then constructed to explicitly model the initial interaction patterns of items in the target user’s current session. Finally, the weighting matrix and LSG are input into our proposed fusion algorithm, and the E-GNN representing the target user’s real interaction pattern is the output. The main contributions of this paper are: (1) Multiple interaction patterns are explicitly modelled in one interaction session graph; and (2) A new fusion algorithm is proposed to reconstruct current session graph by integrating the extracted weights representing the directivity of interacted items from the WGIG. A large number of comparative experiments are conducted on two public datasets, and the results show that E-GNN has distinct advantages over other baseline models.

Session-based recommendation with time-aware neural attention network

In situations where user information and detailed knowledge of user behaviors are challenging to obtain, the session-based recommendation is essential. The session-based recommendation (SBR) relies on the current anonymous session sequence to predict the next action. Recently, sequence modeling methods such as recurrent neural networks have been widely used to model user preferences in session-based recommendations. However, the existing SBR methods usually ignore the click time of interacted items, resulting in inaccurate user preference models and poor recommendations. In this article, we propose a time-aware neural attention network to model dynamic user preferences in SBR tasks. First, a global session graph is constructed based on all sessions, and a graph neural network is used to learn the embedding of items. Then, a gated recurrent unit is used to refine the item embedding and capture the general interests of the user based on the current session. Next, a novel time-aware neural attention network is proposed and used to model the main purpose of the user in the session. Finally, the user’s general interests and main purposes are combined to obtain dynamic user preferences for generating recommendations. Extensive experiments show that our method outperforms state-of-the-art session-based recommendation methods significantly and consistently.

SGKT: Session graph-based knowledge tracing for student performance prediction

Knowledge tracing is a modeling method of students’ knowledge mastery. The deep knowledge tracing (DKT) model uses long short-term memory (LSTM) to process the sequence data of students exercises. However, the LSTM-based model pays more attention to the short-term response status of students while ignoring the long-term learning process. Moreover, existing graph-based knowledge tracing models focus on the static relationship between exercises and skills, ignoring the dynamic graphs formed by students exercises in a session. In this work, we propose a novel knowledge tracing model which is based on an exercise session graph, named session graph based knowledge tracing (SGKT). The session graph is used to model the students’ answering process. In addition, a relationship graph is used to model the relationship between exercises and skills. Then we use gated graph neural networks to obtain the students’ knowledge state from the session graph and use graph convolutional networks to obtain the embedding representations of exercises and skills in the relationship graph. Next, through the interaction mechanism, multiple interaction states composed of knowledge states and embedding representations are obtained. The attention mechanism is used to find the focus from these states and make predictions. Experiments are conducted on three publicly available datasets and the results show that our approach has advantages over some existing baseline methods.

Fusing frequent sub-sequences in the session-based recommender system

A session-based recommender system (RS) forecasts the item to be clicked next by taking advantage of the item sequence in the current session. The deep learning models, such as the recurrent neural network (RNN) and the graph neural network (GNN), are recently applied in the session-based recommendation. However, to the best of our knowledge, the existing methods ignore the fusion of the frequent sub-sequences which refer to those item sub-sequences that appear frequently in different sessions. Intuitively, the more frequently an item sequence appears, the more important it could be. In this paper, we propose the frequent sub-sequence modeling (FSM) method, by utilizing both the local and global information of item sequences. More specifically, we use the item sequence in the current session to build a local session graph and the frequent sub-sequences to construct a global session graph. In addition, we design a gated layer to control the contributions of these two session graphs. Our method is capable of learning both the session-level and global-level item embedding, and thus supports accurate predictions. It is tested on four benchmark datasets and the results show that our method is superior to state-of-the-art approaches consistently.

Time enhanced graph neural networks for session-based recommendation

Session-based recommendation (SBR) is a challenging task, aiming at recommending items according to the behavior of anonymous users. Previous research efforts mainly focus on capturing sequential transitions between consecutive items via recurrent neural networks (RNN) or modeling the complex transitions between non-adjacent items based on graph neural networks (GNN). Although these works have achieved encouraging performance on solving the session-based recommendation problem, few efforts have been dedicated to exploring the rich information related to the shifts of user interests within the transition relationships, which is the research gap we attempt to bridge in this work. In this paper, we propose a novel model, named Time Enhanced Graph Neural Networks (TE-GNN), which attempts to capture the complex user interest shift patterns within sessions. In TE-GNN, we construct a Time Enhanced Session Graph (TES-Graph) where transition relationships between items are treated adaptively with respect to the degree of user interest drift. In addition, a novel Temporal Graph Convolutional Network (T-GCN) is designed to learn item embeddings based on the TES-Graph. Moreover, we also introduce a Temporal Interest Attention Network (TIAN) to model the complex transition of items with a common user interest. Extensive experiments have been conducted on four widely used benchmark datasets, i.e., Diginetica, Tmall, Nowplaying, and Retailrocket, and the results show that our proposed approach TE-GNN significantly outperforms previous state-of-the-art baseline methods. The implementation of TE-GNN is available in https://github.com/GuTang1997/TE-GNN.

Session-Enhanced Graph Neural Network Recommendation Model (SE-GNNRM)

Session-based recommendation aims to predict anonymous user actions. Many existing session recommendation models do not fully consider the impact of similar sessions on recommendation performance. Graph neural networks can better capture the conversion relationship of items within a session, but some intra-session conversion relationships are not conducive to recommendation, which requires model learning more representative session embeddings. To solve these problems, an improved session-enhanced graph neural network recommendation model, namely SE-GNNRM, is proposed in this paper. In our model, the complex transitions relationship of items and more representative item features are captured through graph neural network and self-attention mechanism in the encoding stage. Then, the attention mechanism is employed to combine short-term and long-term preferences to construct a global session graph and capture similar session information by using a graph attention network fused with similarity. In order to prove the effectiveness of the constructed SE-GNNRM model, three public data sets are selected here. The experiment results show that the SE-GNNRM outperforms the existing baseline models and is an effective model for session-based recommendation.

A Spatiotemporal Graph Neural Network for session-based recommendation

Session-based recommendation is a challenging task dedicating to predict the user’s next item of interest based on a given sequence of anonymous behaviors. Most existing methods aim to model items’ temporal and spatial relationships, including recurrent neural network-based and graph neural network-based approaches. However, they have stagnated, expressed inadequacy, and rarely considered user behavior patterns hidden behind items that were common in the click process. In this paper, we propose a Spatiotemporal Graph Neural Network (SGNN) model for the session-based recommendation, which can effectively simulate the users’ behavior patterns from a spatiotemporal perspective to determine the next click. The proposed SGNN model mainly consists of two modules: a Spatiotemporal Session Graph (SSG) module and a Preference-aware AttentioN (PAN) module. In the SSG module, to simulate the possible behavior patterns of users, we model all session sequences as directed and weighted sequences. Then, these session sequences are integrated into a spatial graph, followed by an auto-regressive moving average convolution filter, thus availably assessing the time-decay of user preference. In the PAN module, considering that each item is not equally important to the user’s next click on the item, first of all, we divide the session into global behavioral interest and current interest, where the global behavioral interests fully capture the overall preferences in the session. Furthermore, we use the last clicked item in the session as the current interest to enhance the prediction of the next click behavior. Extensive experiments on two public datasets show that the proposed SGNN model consistently outperforms the comparative methods interns of accuracy and performance.

GC–HGNN: A global-context supported hypergraph neural network for enhancing session-based recommendation

• In order to fully consider the global- and local-level context information, all session sequences are modeled as a hypergraph and the current session sequence is modeled as local session graph. In particular, each node in the local session graph constructed in this paper is connected to an implicit dynamic mediation node. • Our proposed model uses hypergraph convolutional neural network to capture complex higher-order relationships between items and graph attention network to learn the pairwise item transiting relationships. • In our work, sum-pooling operation is adopted to fuse the global and local-level context information, and reversed position information is incorporated. Specifically, the proposed model uses the attention mechanism to get the final session representation. • The results of extensive experiments demonstrate that our proposed model consistently outperforms the state-of-the-art methods. Session-based recommendation (SBR) aims to predict the next most likely item to be interacted based on the current session. Given a short individual session, it is difficult to make accurate recommendations by using only the information of the sequence itself. To improve the recommendation effect, it has become a trend to comprehensively consider the interactive information of other session sequences. In this paper, GC–HGNN, a hypergraph neural network, is proposed to enhance the SBR effects. The model first learns the preferences of current users by transiting information among items in all session sequences. On this basis, the GC–HGNN model fully considers the global context information and local context information of items, and constructs the global session hypergraph and local session graph. The global context information is obtained by the hypergraph convolutional neural network, and the graph attention network calculates the local context information by learning the pairwise relationship between items. The sum-pooling method is utilized to fuse the two kinds of information and the fused information is applied to accomplish the prediction. Extensive experiments were conducted on three public datasets, Diginetica, Tmall and Nowplaying, and the experimental results demonstrate that our proposed GC-HGNN model outperforms the baseline models in SBR.

Collaborative Co-Attention Network for Session-Based Recommendation

Session-based recommendation aims to model a user’s intent and predict an item that the user may interact with in the next step based on an ongoing session. Existing session-based recommender systems mainly aim to model the sequential signals based on Recurrent Neural Network (RNN) structures or the item transition relations between items with Graph Neural Network (GNN) based frameworks to identify a user’s intent for recommendation. However, in real scenarios, there may be strong sequential signals existing in users’ adjacent behaviors or multi-step transition relations among different items. Thus, either RNN- or GNN-based methods can only capture limited information for modeling complex user behavior patterns. RNNs pay attention to the sequential relations among consecutive items, while GNNs focus on structural information, i.e., how to enrich the item embedding with its adjacent items. In this paper, we propose a Collaborative Co-attention Network for Session-based Recommendation (CCN-SR) to incorporate both sequential and structural information, as well as capture the co-relations between them for obtaining an accurate session representation. To be specific, we first model the ongoing session with an RNN structure to capture the sequential information among items. Meanwhile, we also construct a session graph to learn the item representations with a GNN structure. Then, we design a co-attention network upon these two structures to capture the mutual information between them. The designed co-attention network can enrich the representation of each node in the session with both sequential and structural information, and thus generate a more comprehensive representation for each session. Extensive experiments are conducted on two public e-commerce datasets, and the results demonstrate that our proposed model outperforms state-of-the-art baseline model for session based recommendation in terms of both Recall and MRR. We also investigate different combination strategies and the experimental results verify the effectiveness of our proposed co-attention mechanism. Besides, our CCN-SR model achieves better performance than baseline models with different session lengths.

Session Recommendation Model Based on Context-Aware and Gated Graph Neural Networks.

The graph neural network (GNN) based approach has been successfully applied to session-based recommendation tasks. However, in the face of complex and changing real-world situations, the existing session recommendation algorithms do not fully consider the context information in user decision-making; furthermore, the importance of context information for the behavior model has been widely recognized. Based on this, this paper presents a session recommendation model based on context-aware and gated graph neural networks (CA-GGNNs). First, this paper presents the session sequence as data of graph structure. Second, the embedding vector representation of each item in the session graph is obtained by using the gated graph neural network (GGNN). In this paper, the GRU in GGNN is expanded to replace the input matrix and the state matrix in the conventional GRU with input context captured in the session (e.g., time, location, and holiday) and interval context (representing the proportion of the total session time of each item in the session). Finally, a soft attention mechanism is used to capture users' interests and preferences, and a recommendation list is given. The CA-GGNN model combines session sequence information with context information at each time. The results on the open Yoochoose and Diginetica datasets show that the model has significantly improved compared with the latest session recommendation methods.

Attention-Enhanced Graph Neural Networks for Session-Based Recommendation

Session-based recommendation, which aims to match user needs with rich resources based on anonymous sessions, nowadays plays a critical role in various online platforms (e.g., media streaming sites, search and e-commerce). Existing recommendation algorithms usually model a session as a sequence or a session graph to model transitions between items. Despite their effectiveness, we would argue that the performance of these methods is still flawed: (1) Using only fixed session item embedding without considering the diversity of users’ interests and target items. (2) For user’s long-term interest, the difficulty of capturing the different priorities for different items accurately. To tackle these defects, we propose a novel model which leverages both the target attentive network and self-attention network to improve the graph-neural-network (GNN)-based recommender. In our model, we first model user’s interaction sequences as session graphs which serves as the input of the GNN, and each node vector involved in session graph can be obtained via the GNN. Next, target attentive network can activates different user interests corresponding to varied target items (i.e., the session embedding learned varies with different target items), which can reveal the relevance between users’ interests and target items. At last, after applying the self-attention mechanism, the different priorities for different items can be captured to improve the precision of the long-term session representation. By using a hybrid of long-term and short-term session representation, we can capture users’ comprehensive interests at multiple levels. Extensive experiments demonstrate the effectiveness of our algorithm on two real-world datasets for session-based recommendation.

WGIN: A Session-Based Recommendation Model Considering the Repeated Link Effect

Session-based recommendation systems have high application value. Determining how to make better use of anonymous user sessions to recommend items of interest is a considerable challenge for current recommendation systems. Existing research has mainly focused on sequential session patterns; however, due to the complexity and diversity of user interests, such interests cannot be effectively modeled in this way. Therefore, in this paper, we investigate the transition patterns between items by constructing a session graph and propose a novel model called Weighted Graph Interest Networks (WGIN) that collaboratively considers hidden user preference information and the potential order of items in the session graph for a session-based recommendation system. Specifically, we propose a repetitive weighted graph neural network (RWGNN), which pays attention to the transitions between frequent items in a session to deeply explore the preferences of users. In addition, we establish a new Transformer structure to model long-term and short-term user preferences and obtain rich session embeddings. Extensive experiments on two real datasets illustrate that the proposed model outperforms other state-of-the-art session-based recommendation methods.

Session-Based Recommendation with Graph Neural Networks

The problem of session-based recommendation aims to predict user actions based on anonymous sessions. Previous methods model a session as a sequence and estimate user representations besides item representations to make recommendations. Though achieved promising results, they are insufficient to obtain accurate user vectors in sessions and neglect complex transitions of items. To obtain accurate item embedding and take complex transitions of items into account, we propose a novel method, i.e. Session-based Recommendation with Graph Neural Networks, SR-GNN for brevity. In the proposed method, session sequences are modeled as graphstructured data. Based on the session graph, GNN can capture complex transitions of items, which are difficult to be revealed by previous conventional sequential methods. Each session is then represented as the composition of the global preference and the current interest of that session using an attention network. Extensive experiments conducted on two real datasets show that SR-GNN evidently outperforms the state-of-the-art session-based recommendation methods consistently.