Machine Learning-based Book Recommendation Systems: A Comparative Study of CFNN and KNN Algorithms

This study focuses on the implementation and optimization of a recommendation system using deep learning-based collaborative filtering algorithms. The system utilizes user-item interactions from provided datasets to predict user ratings for items in a test set. We introduce a hybrid model that incorporates both collaborative filtering and matrix factorization techniques to enhance prediction accuracy. The collaborative filtering approach exploits similarities between user preferences, while the matrix factorization method decomposes the user-item matrix to capture latent features. The effectiveness of the system is evaluated through various metrics, including precision and recall, with results indicating substantial improvements in recommendation accuracy and system robustness.

An innovative book recommendation tool called BookWise was created to help readers get individualized book suggestions based on their unique preferences. Effective recommendation systems are now more necessary than ever to guide readers through the overwhelming amount of options given the exponential expansion of the digital book market. Advanced machine learning algorithms and methodsfor natural language processing are used by BookWise to analyze a variety of features of books, including genre, author, narrative, and style, as well as user information like reading habits, ratings, and history. These elements are combined by BookWise to produce precise recommendationsthat are pertinent to the preferences of each user. BookWise's ability to learn from and adjust to the evolving preferences of its users over time is at the heart of the program. The system uses collaborative filtering and content-based filtering approaches to continuously gather user feedback on suggested books and take user interactions into account. By doing this, it is able to make increasingly accurate and interesting recommendations to users. By implementing a user review and rating system, BookWise also encourages social involvement and a sense of community among its members. Users can communicate with people who share their interests in literature, discussreviews, and offertheirthoughts. This community-driven strategy improves the quality of suggestions while also enhancing the reading experience by encouraging interactions and relationships amongst book lovers. Furthermore, BookWise is created to be very user- friendly, giving cross-platform compatibility, including online and mobile applications, to make sure users can get their tailored recommendations wherever they are and wheneverthey want. The system also offers a simple and user-friendly design that makes it simple to navigate and explore books of all genres, new releases, and popular works. Through thorough user testing and review, BookWise's efficacy has been confirmed, proving its capacity to provide precise and pleasing recommendations that are in line with consssumers' interests. BookWise represents a significant advancement in the field of book recommendation systems by providing a potent tool to improve reader experiences, promote literary research in the digital age, and ease discovery. Keywords—Personalize book recommendation; recommendation system; clustering; machine learning

Collaborative Filtering (CF) is one of the most successful algorithms in recommender systems. However, it suffers from data sparsity and scalability problems. Although many clustering techniques have been incorporated to alleviate these two problems, most of them fail to achieve further significant improvement in recommendation accuracy. First of all, most of them assume each user or item belongs to a single cluster. Since usually users can hold multiple interests and items may belong to multiple categories, it is more reasonable to assume that users and items can join multiple clusters (groups), where each cluster is a subset of like-minded users and items they prefer. Furthermore, most of the clustering-based CF models only utilize historical rating information in the clustering procedure but ignore other data resources in recommender systems such as the social connections of users and the correlations between items. In this article, we propose HMCoC, a Hybrid Multigroup CoClustering recommendation framework, which can cluster users and items into multiple groups simultaneously with different information resources. In our framework, we first integrate information of user--item rating records, user social networks, and item features extracted from the DBpedia knowledge base. We then use an optimization method to mine meaningful user--item groups with all the information. Finally, we apply the conventional CF method in each cluster to make predictions. By merging the predictions from each cluster, we generate the top-n recommendations to the target users for return. Extensive experimental results demonstrate the superior performance of our approach in top-n recommendation in terms of MAP, NDCG, and F1 compared with other clustering-based CF models.

Recommendation systems provide ways of directing users to items that may be relevant to them by guiding them to relevant items that will be suitable to the users according to their profiles. Collaborative filtering is one of the most successful and mature techniques of recommender system because of its domain independent ability. Bayesian Personalized Ranking Smart Linear Model (BPRSLIM) is model-based collaborative filtering (CF) recommendation algorithm that usually reconstructs a scanty user-item matrix directly; also, using only user-rating matrix usually prevents the algorithm from accessing relevant information that could enhance its recommendation accuracy. Therefore, this work reconstructs BPRSLIM user-item rating matrix via item feature information in order to improve its performance accuracy. Comprehensive experiments were carried out on a real-world dataset using different evaluation metrics. The performance of the model showed significant improvement in recommendation accuracy when compared with other top-N collaborative filtering-based recommendation algorithms, especially in precision and nDCG with 30.6% and 22.1% respectively.

Collaborative filtering is one of the most widely-used algorithms in recommendation systems. In user-based collaborative filtering algorithm, current users' nearest neighbors are used to recommend items because they have similar preference, but users' preference varies with time, which often affects the accuracy of the recommendation. As a result of the varying users' preference, many researches about recommendation systems are focusing on the time factor, to find a way to make up for the change in preferences of users. The existing time-related algorithms usually add time factor in the training phase and make this procedure more complicated. To catch the newest preference of the users and improve the accuracy of the recommendation without complicating the training phase, a timesensitive collaborative filtering model is proposed in this paper, which keeps the original training phase and make some changes in the prediction phase. During the recommendation process, the proposed model orders the items by time for each user as a sequence. The sequence is called time-behavior sequence. First it finds the last item from current user's time-behavior sequence which represents the newest preference of the current user. Secondly, it locates the item in nearest neighbors' timebehavior sequence and saves the timestamp of the item. Lastly, it recommends the items whose timestamps are greater than the saved timestamp from the nearest neighbors' time-behavior sequence. Experiments on the MovieLens dataset show that the proposed time-sensitive collaborative filtering model gives better recommendation quality than the traditional user-based collaborative filtering recommendation algorithm. It can catch the newest preferences of the users and increase the accuracy of recommendation, without changing the training phase.

This study investigated the integration of eye tracking technologies in recommender systems, focusing on their potential to enhance personalization, accuracy, and user engagement. Eye tracking metrics, including fixation duration and gaze patterns, provide a non-intrusive means of capturing real-time user preferences, which can lead to more effective recommendations. Through a comprehensive comparison of current studies, this paper synthesizes findings on the impact of eye tracking across application domains such as e-commerce and media. The results indicate notable improvements in recommendation accuracy with the use of gaze-based feedback. However, limitations persist, including reliance on controlled environments, limited sample diversity, and the high cost of specialized eye tracking equipment. To address these challenges, this paper proposes a structured framework that systematically integrates eye tracking data into real-time recommendation generation. The framework consists of an Eye Tracking Module, a Preferences Module, and a Recommender Module, creating an adaptive recommendation process that continuously refines user preferences based on implicit gaze-based interactions. This novel approach enhances the adaptability of recommender systems by minimizing reliance on static user profiles. Future research directions include the integration of additional behavioral indicators and the development of accessible eye tracking tools to broaden real-world impact. Eye tracking shows substantial promise in advancing recommender systems but requires further refinement to achieve practical, scalable applications across diverse contexts.

This research paper explores the application of collaborative filtering techniques to implicit feedback data within the Anime Recommendations Database. The study focuses on leveraging user behavior, such as viewing history and interactions, to provide personalized anime recommendations. We employ matrix factorization and nearest-neighbor approaches, comparing their effectiveness and efficiency in handling large datasets. Our results demonstrate significant improvements in recommendation accuracy and user satisfaction, highlighting the potential of collaborative filtering in the domain of anime recommendations. Recommender systems are super important for helping users find stuff they like, whether it's shows to watch, things to buy, or people to follow online. This study is all about using cool collaborative filtering techniques to make anime recommendations even better. We even tested these models and found that one called ALS works better with sparse data and gives more accurate recommendations than k-NN. Plus, we came up with a hybrid model that combines different approaches, and it's made a big difference in the quality of recommendations by solving the “cold-start” problem and offering more diverse suggestions. Our research shows that collaborative filtering is great for dealing with implicit feedback data, and we've got some practical ideas for making advanced recommendation systems for anime and other stuff too. This research paper explores the application of collaborative filtering techniques to implicit feedback data within the Anime Recommendations Database

The cold-start problem remains one of the most significant challenges in recommendation systems, particularly in music platforms where user preferences are diverse and personalized experiences are crucial. This research presents a novel approach to address the cold-start problem in music recommendation by integrating meta-learning and fuzzy logic techniques. Using the LFM-2b dataset which contains over two billion music listening events, we develop a hybrid recommendation framework that can rapidly adapt to new users and items with minimal interaction history. The proposed model employs a meta-learning strategy to transfer knowledge from existing users to new ones by learning generalizable patterns of music preferences. This is complemented by a fuzzy preference modeling component that captures the inherent uncertainty and gradation in user preferences for music genres, artists and acoustic features. Our framework introduces a novel prototype-based architecture that identifies representative user and item prototypes through a clustering mechanism enhancing both recommendation accuracy and explainability. Extensive experiments demonstrate that our approach outperforms state-of-the-art methods in cold-start scenarios, achieving a 15.2% improvement in recommendation accuracy for new users and a 12.7% improvement for new items compared to traditional collaborative filtering methods. The results show that the integration of fuzzy logic with meta-learning provides a robust solution for cold-start music recommendation by effectively modeling the uncertainty in user preferences while transferring knowledge across similar user groups.

In the realm of recommender systems, the prediction of diverse customer preferences has emerged as a compelling research challenge, particularly for multi-state business organizations operating across various geographical regions. Collaborative filtering, a widely utilized recommendation technique, has demonstrated its efficacy in sectors such as e-commerce, tourism, hotel management, and entertainment-based customer services. In the item-based collaborative filtering approach, users’ evaluations of purchased items are considered uniformly, without assigning weight to the participatory data sources and users’ ratings. This approach results in the ‘relevance problem’ when assessing the generated recommendations. In such scenarios, filtering collaborative patterns based on regional and local characteristics, while emphasizing the significance of branches and user ratings, could enhance the accuracy of recommendations. This paper introduces a turnover-based weighting model utilizing a big data processing framework to mine multi-level collaborative filtering patterns. The proposed weighting model assigns weights to participatory data sources based on the turnover cost of the branches, where turnover refers to the revenue generated through total business transactions conducted by the branch. Furthermore, the proposed big data framework eliminates the forced integration of branch data into a centralized repository and avoids the complexities associated with data movement. To validate the proposed work, experimental studies were conducted using a benchmarking dataset, namely the ‘Movie Lens Dataset’. The proposed approach uncovers multi-level collaborative pattern bases, including global, sub-global, and local levels, with improved predicted ratings compared with results generated by traditional recommender systems. The findings of the proposed approach would be highly beneficial to the strategic management of an interstate business organization, enabling them to leverage regional implications from user preferences.

Modeling user preferences using neural networks and tensor factorization model

Recommender systems play an important role in most modern e-commerce applications. They have allowed users to become aware of the myriad choices available to them. The ease of information and the abundance of options have helped users make educated decisions. A recommender system studies a user's preferences and continues learning the user's changing interests, so as to suggest items that incline with the user's interests. In cases where a user is new to the application, or the user prefers not to discourse preferences, the recommender system is unable to gather the user's preference on any item. This is called the cold start problem; wherein the system can make valid recommendations only once the user starts informing the system about his/her choices. In this paper, we discuss the challenges faced by the cold start problem and how this problem may be alleviated using social media. We suggest an approach where we collect public information from users' social media accounts and analyze this information to understand their preferences. In particular, we gather the new user's information using their Twitter profile; i.e., the user's interest and preferences are extracted from his/her Twitter profile by analyzing his/her tweets. These interests will help the system understand what kind of movies the user will be most interested in. We compare these preferences with the metadata about the individual items. Using this approach, we develop a movie recommendation system wherein we produce top-N movie recommendations for a user. We used the MovieTweetings dataset to model the application. Two sets of results have been produced. In the first, smaller set of 770 users, 72.67% of users have received 100% accurate movie recommendations while nearly 80% of users got more than 75% accuracy. For the second, larger set of more than 3,500 users, 53 % of users have received 100% accurate recommendations while 72% of users got more than 75% accuracy. These encouraging results have demonstrated that the approach is effectively in alleviating cold start problems in recommendation systems, and may be applicable to many other e-commerce applications.

소셜 네트워크는 사용자들의 공통된 관심사, 경험 그리고 일상생활들을 함께 공유하기 위해 소셜 네트워크 상 사람들을 서로 연결시켜주는 거대한 커뮤니케이션 플랫폼이다. 소셜 네트워크상의 사용자들은 포스팅, 댓글, 인스턴스 메시지, 게임 소셜 이벤트 외에도 다양한 애플리케이션을 통해 다른 사용자들과 소통하고 개인 정보 관리하는데 많은 시간을 소비한다. 소셜 네트워크상의 풍부한 사용자 정보는 추천시스템이 추천 성능을 향상시키기 위해 필요한 큰 잠재력이 되었다. 대부분의 사용자들은 어떤 상품을 구매하기 전 가까운 관계이거나 같은 성향을 가진 사람들의 의견을 반영하여 의사 결정을 하게 된다. 그러므로 소셜 네트워크에서의 사용자 관계는 추천시스템을 위한 사용자 선호도 예측을 효율적으로 높이는데 중요한 요소라 할 수 있다. 일부 연구자들은 소셜 네트워크에서의 사용자와 다른 사용자들 사이의 상호작용 즉, 소셜 관계(social relationship)와 같은 소셜 데이터가 추천시스템에서 추천의 질에 어떠한 영향을 미치는가를 연구하고 있다. 추천시스템은 아마존, 이베이, Last.fm과 같은 큰 규모의 전자상거래 사이트 또한 채택하여 사용되는 시스템으로, 추천시스템을 위한 방법으로는 협업적 여과 방법과 내용 기반 여과 방법이 있다. 협업적 여과방법은 사용자들의 선호도 학습에 의해 사용자가 아직 평가하지 않은 아이템 중 선호할 수 있는 아이템을 정확하게 제안하기 위한 추천시스템 방법 중 하나이다. 협업적 여과는 사용자들의 데이터에 초점을 맞춘 방법으로 유사한 배경과 선호도를 가지는 사용자들로부터 정보를 수집하여 사용자들의 선호도 예측을 자동으로 발생시킨다. 특히 협업적 여과는 근접한 이웃 사용자들에 의해서 목적 사용자가 선호할 수 있는 아이템을 제시하는 것으로 유사한 이웃 사용자를 찾는 것이 중요하다. 좋은 이웃 사용자 발견은 사용자와 아이템을 고려하는 방법이 일반적이다. 각 사용자는 아이템 즉, 영화, 상품, 책 등에 자신의 선호도를 나타내기 위하여 평가 값을 입력하고, 시스템은 이를 바탕으로 사용자-평가 행렬을 구축한다. 이 사용자-평가 행렬은 목적 사용자와 유사하게 아이템을 평가한 사용자 그룹을 찾기 위한 것으로, 목적 사용자가 아직 평가하지 않은 아이템에 대하여 사용자-평가 매트릭스를 통해 그 평가 값을 예측한다. 현재 이 협업적 여과 방법은 전자상거래와 정보 검색에서 적용되어 개인화 시스템에 효율적으로 사용되고 있다. 하지만 초기 사용자 문제, 데이터 희박성 문제와 확장성 그리고 예측 정확도 향상 등 해결해야 할 과제가 여전히 남아 있다. 이러한 문제들을 해소하기 위해 많은 연구자들은 하이브리드, 신뢰기반, 소셜 네트워크 기반 협업적 여과와 같은 다양한 방법을 제안하였다. 본 논문에서는 전통적인 협업적 여과 방식의 예측 정확도와 추천 성능을 향상시키기 위해 소셜 네트워크에 존재하는 소셜 관계를 이용한 협업적 여과 시스템을 제안한다. 소셜 관계는 소셜 네트워크 서비스 중 하나인 페이스북 사용자들이 남긴 포스팅과 사용자의 소셜 네트워크 친구와 의견 교류 중 남긴 코멘트와 같은 사용자 행동을 기반으로 정의된다. 소셜 관계를 구축하기 위해 소셜 네트워크 사용자의 포스팅과 댓글 추출하고, 추출된 텍스트에 불용어 및 특수 기호 제거와 스테밍 등 전처리를 수행하였다. 특징 벡터는 TF-IDF를 이용하여 전처리된 텍스트에 나타난 각 단어에 대한 특징 점수를 계산함으로써 구축된다. 본 논문에서 이웃 사용자를 결정하기 위해 사용되는 사용자 간 유사도는 특징 벡터를 이용한 사용자 행동 유사도와 사용자의 영화 평가를 기반으로 한 전통적 방법의 유사도를 결합하여 계산된다. 제안하는 시스템은 목표 사용자와 제안한 방법을 통해 결정된 이웃 사용자 집단을 기반으로 목표 사용자가 평가하지 않은 아이템에 대한 선호도를 예측하고 Top-N 아이템을 선별하여 사용자에게 아이템을 추천하게 된다. 본 논문에서 제안하는 방법을 확인하고 평가하기 위하여 IMDB에서 제공하는 영화 정보 기반으로 영화 평가 시스템을 구축하였다. 예측 정확도를 평가하기 위해 MAE 값을 이용하여 제안하는 알고리즘이 얼마나 정확한 추천을 수행하는지에 대한 예측 정확도를 측정하였다. 그리고 정확도 재현율 및 F1값 등을 활용하여 시스템의 성능을 평가하였으며, 시스템의 추천 품질은 커버리지를 이용하여 평가되었다. 실험 결과로부터 본 논문에서 제안한 시스템이 보다 더 정확하고 좋은 성능으로 사용자에게 아이템을 추천하는 것을 볼 수 있었다. 특히 소셜 네트워크에서 사용자 행동을 기반으로 한 소셜 관계를 이용함으로써 추천 정확도를 6% 향상시킴을 보였다. 또한 벤치마크 알고리즘과의 성능비교 실험을 통해 7% 향상된 추천 성능의 결과를 보여준다. 그러므로 사용자의 행동으로부터 관찰된 소셜 관계를 CF방법과 결합한 제안한 방법이 정확한 추천시스템을 위해 유용하며, 추천시스템의 성능과 품질을 향상시킬 수 있음을 알 수 있다.

Nowadays, recommendation systems are widely used to recommend items to the users that are specific to their individual preferences and most appropriate. For this reason, many academic libraries try to establish an effectiveness and efficiency book recommendation system which could enhance students' performance. This research presents the process of book recommendation by using the collaborative filtering (CF), one of the most popular techniques widely used in recommendation systems, for university students. Since data sparseness is the one key issue limiting the success of collaborative filtering, we also adopt bias matrix factorization technique to handle this problem. Book recommendation of each student has been generated by using existing borrowing records with a time stamp. In our experiments, different techniques of similarity calculation are compared. The performance evaluations are conducted using both accuracy measure and student satisfaction evaluation with the book recommended by the system.

Integrating contrastive learning and adversarial learning on graph denoising encoder for recommendation

Improvement of reading platforms assisted by the spring framework: A recommendation technique integrating the KGMRA algorithm and BERT model.