Cold Start Issues Research Articles

Addressing data sparsity and cold-start challenges in recommender systems using advanced deep learning and self-supervised learning techniques

ABSTRACT The efficacy of e-commerce conversion rates relies on precise and personalized product recommendations within recommendation systems (RS). While collaborative filtering-based RS has demonstrated success, challenges such as sparsity and cold-start issues in the user-item matrix can impede optimal functionality. To address these challenges, there is a need to integrate additional information sources, encompassing item/user profiles and textual reviews. This study introduces an innovative RS architecture that seamlessly combines self-supervised learning (SSL) and collaborative filtering techniques with BERT-DNN to surmount these obstacles. The distinctiveness of our approach lies in integrating self-supervised learning with collaborative filtering and contextualized data obtained from BERT-DNN, providing a profound understanding of item profiles to enhance comprehension of user preferences and item characteristics. This refined understanding, operational in conjunction with collaborative filtering models like ItemKNN and UserKNN, empowers the system to generate highly personalized recommendations. The proposed method entails several pivotal steps: developing the BERT language model for textual embeddings in item profiles, conducting dimensionality reduction, constructing a Deep Neural Network, implementing self-supervised learning with both UserKNN and ItemKNN CF methods, and employing an ensemble learning technique. Empirical results substantiate the efficacy of our approach, with a specific focus on the innovative fusion of BERT-DNN with self-supervised learning and KNN CF methodologies, showcasing substantial improvements across diverse performance metrics. This underscores the practical importance of leveraging contextualized BERT-DNN data, strengthening the recommendation mechanism, and ultimately enhancing the overall performance of the RS.

Co-clustering method for cold start issue in collaborative filtering movie recommender system

Co-clustering method for cold start issue in collaborative filtering movie recommender system

Analysis on recommendation systems based on ML and DL approaches

The explosive growth of the Internet has led to an unprecedented increase in available information, creating a major challenge in identifying content that aligns with user interests. Recommendation systems (RS) address this challenge by providing personalized content recommendations based on user behavior and preferences. This article systematically reviews the evolution of recommendation systems with a focus on the methodologies of machine learning (ML) and deep learning (DL). It highlights core technologies in ML-based RS, such as content-based recommendation, collaborative filtering (CF), and hybrid filtering (HF). The article evaluates the effectiveness of various DL models, including autoencoders (AE), convolutional neural networks (CNN), and recurrent neural networks (RNN). The review also addresses key challenges faced by RS, such as cold start issues and the need for improved model transparency and interpretability. DL methods are shown to significantly enhance recommendation accuracy by leveraging complex patterns in large-scale data, to improve the user experience. Future research directions include refining data preprocessing, enhancing feature engineering, compressing and accelerating DL models, and improving interpretability through advanced mechanisms. This study provides valuable insights and references for researchers aiming to advance recommendation system design and performance optimization.

Elevating recommender systems: Cutting-edge transfer learning and embedding solutions

In today’s information age and connected economy, Recommender Systems (RS) plays a vital role in managing information overload and delivering personalized suggestions to users. This paper introduces a multistage model that leverages multimodal data embedding and deep transfer learning to accurately capture user preferences and item characteristics, resulting in highly tailored recommendations. A key innovation in this model is the incorporation of an image dataset in the second phase, which addresses cold-start problems for new items by providing additional visual context. Our approach excels in overcoming challenges related to data sparsity and cold-start issues, thereby providing users with realistic and relevant product recommendations. To validate the effectiveness of the proposed model, we conducted extensive evaluations using three diverse datasets: data from Brazilian e-commerce platforms, the MovieLens 1M dataset, and the Amazon Product Review dataset. These evaluations involved comprehensive comparisons with standard RS methods to assess performance improvements. The results indicate that our proposed model significantly outperforms traditional RS techniques in terms of accuracy and reliability. Our model provides more accurate and meaningful recommendations by effectively addressing issues such as cold-start and data scarcity. Specifically, the model achieved Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) scores of 0.5883 and 0.4012, respectively, which demonstrate its superior performance metrics across all datasets tested.

Energy analysis during cold start and warm up period of a methanol engine hybrid power system under several novel energy conversion strategies

The use of methanol in engine hybrid power system has a wide range of potential applications, with the primary challenge being cold start issues. This paper innovatively proposes four strategies for starting and warming up the methanol engine, including motor dragging, motor dragging coupled with exhaust gas recirculation (EGR), motor dragging coupled with exhaust control, and motor dragging coupled with both EGR and exhaust control. These strategies can be implemented in methanol hybrid power system with minimal hardware modification costs, and not generate any emissions compared to traditional gasoline assisted strategy. The warm-up mechanisms of these strategies were analyzed and their effectiveness was verified by experiments. The optimum strategy, motor dragging coupling both EGR and exhaust control, has been demonstrated to enhance power input and improve the energy conversion efficiency substantially. Compared to the traditional gasoline assisted strategy, the energy consumption, energy conversion efficiency and the warm-up time in this strategy are 17.9 % lower, 5.1 % higher, 4.2 s shorter, respectively. These strategies well solve the problem of methanol engines being unable to start at low temperatures, which is expected to eliminate the cold start of the engine and directly enter the hot start.

Transformer-based medication recommendation with a multiple graph augmentation strategy

Medication recommendation (MR) is a promising task that benefits both patients and medical practitioners. However, the individual differences in patients and drug–drug interactions (DDIs) in medication combinations are two key factors that severely limit the personalization and safety of the present MR models. To alleviate the challenges mentioned above, this paper proposes a novel MR model named Trans-GAHNet, which innovatively leverages transformer-based representation learning and constructs multiple graphs for drug information augmentation. More specifically, the transformer network architecture is deployed to fully encode the sequential EHR data for effective patient representations, which addresses the cold start issue and enhances the interpretability of representation learning. Next, multiple graphs are employed to capture the co-occurrence and exclusiveness in drug pairs and to mitigate potential DDIs in drug combinations, thereby enhancing the personalization and safety of the predicted medication combinations. Finally, extensive experiments on real-world datasets are conducted to evaluate the proposed model, and the experimental results demonstrate that Trans-GAHNet outperforms the state-of-the-art baseline models on multiple metrics, with Jaccard, F1 and PRAUC scores of 0.5238, 0.6786 and 0.7795, respectively.

Disentangled Cascaded Graph Convolution Networks for Multi-Behavior Recommendation

Multi-behavioral recommender systems have emerged as a solution to address data sparsity and cold-start issues by incorporating auxiliary behaviors alongside target behaviors. However, existing models struggle to accurately capture varying user preferences across different behaviors and fail to account for diverse item preferences within behaviors. Various user preference factors (such as price or quality) entangled in the behavior may lead to sub-optimization problems. Furthermore, these models overlook the personalized nature of user behavioral preferences by employing uniform transformation networks for all users and items. To tackle these challenges, we propose the Disentangled Cascaded Graph Convolutional Network (Disen-CGCN), a novel multi-behavior recommendation model. Disen-CGCN employs disentangled representation techniques to effectively separate factors within user and item representations, ensuring their independence. In addition, it incorporates a multi-behavioral meta-network, enabling personalized feature transformation across user and item behaviors. Furthermore, an attention mechanism captures user preferences for different item factors within each behavior. By leveraging attention weights, we aggregate user and item embeddings separately for each behavior, computing preference scores that predict overall user preferences for items. Our evaluation of benchmark datasets demonstrates the superiority of Disen-CGCN over state-of-the-art models, showcasing an average performance improvement of 7.07% and 9.00% on respective datasets. These results highlight Disen-CGCN’s ability to effectively leverage multi-behavioral data, leading to more accurate recommendations.

Prerequisites-based course recommendation: recommending learning objects using concept prerequisites and metadata matching

The recommendation is an active area of scientific research; it is also a challenging and fundamental problem in online education. However, classical recommender systems usually suffer from item cold-start issues. Besides, unlike other fields like e-commerce or entertainment, e-learning recommendations must ensure that learners have the adequate background knowledge to cognitively receive the recommended learning objects. For that reason, when designing an efficient e-learning recommendation method, these challenges should be considered. To address those issues, in this paper, we first propose extracting pairs concept prerequisites using Linked Open Data (LOD). Then, we evaluate the proposed list of prerequisite relationships using machine learning predictive models. Then, we present the recommendation approach based on matching concept’s prerequisites relation and courses metadata through a similarity score. The experimental result of prerequisite identification was evaluated using four well-known machine learning algorithms while achieving an accuracy of 90%. Moreover, using three known evaluation metrics, the final prerequisite-based recommendation demonstrates very good results (NDCG@10 = 86%). This solution will enhance recommendations on online learning platforms. Additionally, it will overcome the cold-start issue and accomplish the needed prerequisites and background knowledge for learners to attain their learning objectives.

Application of recommendation system based on fuzzy logic in music culture communication and music teaching management

Music has been utilized throughout history as a medium for cultural communication and artistic expression, embodying various nations’ and societies’ ideologies and experiences. Music culture communication is crucial for encouraging cultural diversity and understanding and developing social cohesion and community building among people. Music teaching management is the process of setting up, arranging, and executing music education programs in a manner that successfully teaches students the essential skills and information necessary for becoming proficient musicians. Users’ exact preferences for various areas of attraction cannot be determined, nor are users’ choices for traditional music recommendations sufficiently accurate. A recommender system estimates or anticipates people’s preferences and offers appropriate recommendations. First, the sparsity problem emerges when insufficient data is accessible for the recommendation, and the coverage is one of the key drawbacks of social labeling. Cold start issues might be difficult since new music learners might not have given sufficient details about their musical tastes. Hence, the Hybridized Fuzzy logic-based Content and Collaborative Music Recommendation (HFC2MR) system is proposed to create personalized music teaching plans that are effective and engaging for each student based on their music preferences and learning outcomes. Enhanced Fuzzy C-Means clustering is used in collaborative recommendations to group users based on their shared musical tastes and to provide each user with more individualized, accurate music recommendations based on other users’ listening habits and preferences in the same cluster. Subsequently, an assessment of the recommender system using parameters like accuracy, precision, f1-score, and recall ratio is shown with optimal cluster selection. The coverage ratio is used to compare experimental data based on skill capacity covered through the assessment of music teaching. RMSE metric is used to evaluate the accuracy of students’ performance based on music attributes related to teaching goals.

Dual Homogeneity Hypergraph Motifs with Cross-view Contrastive Learning for Multiple Social Recommendations

Social relations are often used as auxiliary information to address data sparsity and cold-start issues in social recommendations. In the real world, social relations among users are complex and diverse. Widely used graph neural networks (GNNs) can only model pairwise node relationships and are not conducive to exploring higher-order connectivity, while hypergraph provides a natural way to model high-order relations between nodes. However, recent studies show that social recommendations still face the following challenges: 1) a majority of social recommendations ignore the impact of multifaceted social relationships on user preferences; 2) the item homogeneity is often neglected, mainly referring to items with similar static attributes have similar attractiveness when exposed to users that indicating hidden links between items; and 3) directly combining the representations learned from different independent views cannot fully exploit the potential connections between different views. To address these challenges, in this article, we propose a novel method DH-HGCN++ for multiple social recommendations. Specifically, dual homogeneity (i.e., social homogeneity and item homogeneity) is introduced to mine the impact of diverse social relations on user preferences and enrich item representations. Hypergraph convolution networks with motifs are further exploited to model the high-order relations between nodes. Finally, cross-view contrastive learning is proposed as an auxiliary task to jointly optimize the DH-HGCN++. Real-world datasets are used to validate the effectiveness of the proposed model, where we use sentiment analysis to extract comment relations and employ the k-means clustering algorithm to construct the item-item correlation graph. Experiment results demonstrate that our proposed method consistently outperforms the state-of-the-art baselines on Top-N recommendations.

Data-driven internal mobility: Similarity regularization gets the job done

This paper presents a novel approach to support career management by recommending job–employee matches within an organization through data-driven insights. We build a recommender system to propose matches. The presented approach extends upon a conventional collaborative filtering recommender system, which suggests matches based on the historic performance similarities of employees. To address the prevalent challenge of the cold start issue in internal placement, we incorporate personal employee data with a similarity-based regularization term. This regularization term finds latent representations that are closer to each other when employees share similar personal features. This approach is evaluated using three real-life datasets and demonstrates a highly competitive performance compared to state-of-the-art benchmark methods. Overall, we make three contributions to the field of HR analytics: (i) we present a comprehensive survey of job–employee recommender systems in the context of internal mobility, (ii) we implement a similarity regularization method, and (iii) we release a first-of-its-kind HR dataset on internal mobility. All code and public data used in this study are available on GitHub.

Netflix Movies Recommendation System

Netflix is a leading global entertainment company with over 247 million paying customers who can access TV series, films, and games in over 190+ countries. Many languages and genres are offered for the content. As much as they like, whenever and wherever they wish, participants are free to change their plans and play, pause, and resume viewing. With the exponential growth of content on streaming platforms like Netflix, users often face the challenge of finding relevant and enjoyable movies.The "Cold-Start" issue is a recommendation system issue wherein new users or items have little data, making it challenging to provide appropriate recommendations. In response, this paper proposes a data-driven movie recommendation system tailored for Netflix. Leveraging machine learning techniques, user preferences, and historical data, the model aims to enhance the user experience by providing personalized movierecommendations. This study explores algorithms such as collaborative filtering, content-based filtering, and hybrid models to achieve accurate and effective movie recommendations. While content-based filtering uses item features to suggest shows or movies a user has previously enjoyed, collaborative filtering examines user behavior and preferences to produce suggestions. Even with insufficient data, the system can deliver more accurate recommendations for new users or things by combining various approaches. By making recommendations based on both item features and user behavior, this hybrid approach helps to reduce the cold start issue.

Tackling the cold start issue in movie recommendations with a refined epsilon-greedy approach

With the rapid growth of the Internet and the consequent surge in data, the current era is characterized by information overload. As the domain of data processing and storage expands, recommendation systems have become pivotal tools in navigating this deluge, assisting users in filtering through vast information landscapes. A notable segment of this is movie recommendation systems. As living standards rise, so does the demand for cinematic experiences. Enhancing and refining the methodologies of these recommendation systems is, therefore, of significant value. However, a consistent challenge is the cold start problem encountered when new users join. Without prior viewing records or preferences, these users pose a dilemma for the system: how to offer relevant recommendations without historical data? Addressing this challenge, this paper proposes a unique method grounded in the N-armed bandit model, introducing an enhanced Epsilon-greedy algorithm specifically designed for movie recommendations for such users. By adjusting dynamically based on real-time user feedback, the algorithm aims to continuously hone its recommendation quality, ensuring a consistently better user experience.

Personalized LoRA for Human-Centered Text Understanding

Effectively and efficiently adapting a pre-trained language model (PLM) for human-centered text understanding (HCTU) is challenging since user tokens are million-level in most personalized applications and do not have concrete explicit semantics. A standard and parameter-efficient approach (e.g., LoRA) necessitates memorizing numerous suits of adapters for each user. In this work, we introduce a personalized LoRA (PLoRA) with a plug-and-play (PnP) framework for the HCTU task. PLoRA is effective, parameter-efficient, and dynamically deploying in PLMs. Moreover, a personalized dropout and a mutual information maximizing strategies are adopted and hence the proposed PLoRA can be well adapted to few/zero-shot learning scenarios for the cold-start issue. Experiments conducted on four benchmark datasets show that the proposed method outperforms existing methods in full/few/zero-shot learning scenarios for the HCTU task, even though it has fewer trainable parameters. For reproducibility, the code for this paper is available at: https://github.com/yoyo-yun/PLoRA.

A cold-start recommendation method for We-map considering co-occurrence collaborative signals and contrastive learning

ABSTRACT We-maps, augmenting traditional maps, confront cold-start problems with prevailing recommendation techniques. The Contrastive Collaborative Filtering (CCF) model tackles We-map's cold-start issues by harnessing co-occurrence signals from warm data in a contrastive framework, thereby refining embeddings through the integration of content and co-occurrence insights. This methodology engenders unique embeddings for We-map entities, utilizing contrastive learning to bridge semantic gaps and enable indirect knowledge transfer, effectively mitigating cold-start challenges. Experimental validation on We-map and various public datasets confirms the method's practicality in addressing cold-start recommendations for We-maps. The study offers a new path for efficient geospatial data dissemination.

Heterogeneous Graph Neural Network with Personalized and Adaptive Diversity for News Recommendation

The emergence of online media has facilitated the dissemination of news, but has also introduced the problem of information overload. To address this issue, providing users with accurate and diverse news recommendations has become increasingly important. News possesses rich and heterogeneous content, and the factors that attract users to news reading are varied. Consequently, accurate news recommendation requires modeling of both the heterogeneous content of news and the heterogeneous user-news relationships. Furthermore, users’ news consumption is highly dynamic, which is reflected in the differences in topic concentration among different users and in the real-time changes in user interests. To this end, we propose a Heterogeneous Graph Neural Network with Personalized and Adaptive Diversity for News Recommendation (DivHGNN). DivHGNN first represents the heterogeneous content of news and the heterogeneous user-news relationships as an attributed heterogeneous graph. Then, through a heterogeneous node content adapter, it models the heterogeneous node attributes into aligned and fused node representations. With the proposed attributed heterogeneous graph neural network, DivHGNN integrates the heterogeneous relationships to enhance node representation for accurate news recommendations. We also discuss relation pruning, model deployment, and cold-start issues to further improve model efficiency. In terms of diversity, DivHGNN simultaneously models the variance of nodes through variational representation learning for providing personalized diversity. Additionally, a time-continuous exponentially decaying distribution cache is proposed to model the temporal dynamics of user real-time interests for providing adaptive diversity. Extensive experiments on real-world news datasets demonstrate the effectiveness of the proposed method.

Integration of Deep Reinforcement Learning with Collaborative Filtering for Movie Recommendation Systems

In the era of big data, effective recommendation systems are essential for providing users with personalized content and reducing search time on online platforms. Traditional collaborative filtering (CF) methods face challenges like data sparsity and the new-user or cold-start issue, primarily due to their reliance on limited user–item interactions. This paper proposes an innovative movie recommendation system that integrates deep reinforcement learning (DRL) with CF, employing the actor–critic method and the Deep Deterministic Policy Gradient (DDPG) algorithm. This integration enhances the system’s ability to navigate the recommendation space effectively, especially for new users with less interaction data. The system utilizes DRL for making initial recommendations to new users and to generate optimal recommendation as more data becomes available. Additionally, singular value decomposition (SVD) is used for matrix factorization in CF, improving the extraction of detailed embeddings that capture the latent features of users and movies. This approach significantly increases recommendation precision and personalization. Our model’s performance is evaluated using the MovieLens dataset with metrics like Precision, Recall, and F1 Score and demonstrates its effectiveness compared with existing recommendation benchmarks, particularly in addressing sparsity and new-user challenges. Several benchmarks of existing recommendation models are selected for the purpose of model comparison.

Prebaking runtime environments to improve the FaaS cold start latency

Function-as-service (FaaS) platforms promise a simpler programming model for cloud computing, given that providers take care of the overall resource management while the developers can concentrate only on writing their applications in the scope of a function, not having to care about installing or scaling resources. As FaaS users are billed based on the execution of the functions, platform providers have a natural incentive not to keep idle resources running at the platform’s expense. However, this strategy may lead to the cold start issue, in which the execution of a function is delayed because there are no ready resources to host the execution. Despite the time to provision the computational resources, starting the runtime environment that manages the function execution can take hundreds of milliseconds to seconds, a prohibitive non-deterministic overhead to latency-critical applications. This work describes a practical technique, Prebaking, to reduce the function start-up time based on restoring snapshots of previously executed processes. The foundation of the Prebaking technique is that deciding when to create a snapshot of a function is essential. For example, based on a prototype we developed using the CRIU checkpoint/restore Linux tool, the performance improvement on the start-up of a properly snapshotted function is up to 25 times. Fortunately, generating a good function snapshot is easy: One only needs to checkpoint a warm function, that is, a function that has processed a request. To show the feasibility of using the Prebaking technique, we discussed integrating it into OpenFaaS, an open-source FaaS platform, including how to generate and restore function snapshots. We also evaluated our prototype by running a comprehensive set of experiments that compare the function start-up duration and cold start latency against the standard fork-exec procedure. We analyze the JVM, CPython, and Node.js runtimes. The results indicate that the technique can improve the function replica start-up time even for no-warmed functions: we improved the function replica start-up time from 3.5 times for a “do-nothing” function running in Node.js up to 12 times when considering an Image Resizer function running in CPython. Finally, we reinforce the practicability of the proposed technique by comparing our proposal to another snapshot-based technique called SEUSS. That evaluation shows that Prebaking provides better cold start latency even for slightly complex functions like the Markdown Renderer.

Real-Time Movie Recommendation: Integrating Persona-Based User Modeling with NMF and Deep Neural Networks

The proliferation of uncategorized information on the Internet has intensified the need for effective recommender systems. Recommender systems have evolved from content-based filtering to collaborative filtering and, most recently, to deep learning-based and hybrid models. However, they often face challenges such as high computational costs, reduced reliability, and the Cold Start problem. We introduce a persona-based user modeling approach for real-time movie recommendations. Our system employs Non-negative Matrix Factorization (NMF) and Deep Learning algorithms to manage complex and sparse data types and to mitigate the Cold Start issue. Experimental results, based on criteria involving 50 topics and 35 personas, indicate a significant performance gain. Specifically, with 500 users, the precision@K for NMF was 86.01%, and for the Deep Neural Network (DNN), it was 92.67%. Tested with 900 users, the precision@K for NMF increased to 97.04%, and for DNN, it was 95.55%. These results represent an approximate 10% and 5% improvement in performance, respectively. The system not only delivers fast and accurate recommendations but also reduces computational overhead by updating the model only when user personas change. The generated user personas can be adapted for other recommendation services or large-scale data mining.

Deep learning architecture towards consumer buying behaviour prediction using multitask learning paradigm

A recommendation system serves as a distributed information filter, predicting customer preferences in reviews, ratings, and comments. Analysing customer behaviour aids in understanding needs and predicting intentions. E-commerce tracks product usage and sentiment to provide a personalized network based on consumer preference modelling. The challenge lies in optimizing item selection for suitable consumers to enhance performance. To address this, an imperative is the item recommendation approach for modelling future consumer behaviour. However, traditional machine learning methods often overlook dynamic product recommendations due to evolving user interests and changes in preferences reflected in customer ratings, causing cold-start issues. To overcome these challenges, a comprehensive deep learning approach is introduced. This approach incorporates a deep neural network for consumer preference prediction, utilizing a multi-task learning paradigm to accommodate variations in consumer ratings. The research contribution lies in applying this network to predict consumer preference scores based on latent multimodal information and item characteristics. Initially, the architecture manages changing consumer aspects and preferences by extracting features and latent factors from customer review rating data. These latent factors include customer demographic information and other concealed features that signify preferences based on experiences and behaviours. Extracted latent features are processed using a sentiment analysis model to generate embedding latent features. A finely-tuned deep neural network with hyper-parameter adjustments serves as a prediction network, forming a customer performance-oriented recommendation system. It processes embedded latent features along with associated sentiments to achieve high prediction accuracy, reliability, and latency. The deep learning architecture, enriched with consumer-specific discriminative information, generates an objective function for item recommendations with minimal error, significantly enhancing predictive performance. Empirical experiments on Amazon review datasets validate the proposed model’s performance, showcasing its enhanced effectiveness and scalability in handling substantial data volumes.