Academic literature on the topic 'Recommendation System; Cold Start Problem; Data Sparsity Problem'

AbstractThe concern has been raised regarding errors in drugs prescription and medical diagnostics that need to be carefully thought through. Both patient diagnosis and medication prescription are the responsibilities of healthcare providers. As the number of people with health issues rises, the healthcare professionals’ burden is increased. Medical errors may occur in the healthcare sector as a result of healthcare professionals prescribing drugs medicines based on inadequate information related to patient history and drug side effects. Therefore, this study aims to propose a drug recommender system to assist healthcare providers in decision making when prescribing drugs for patients depending on their diagnoses. Drug reviews sentiments are analyzed to find the drug effectiveness among the users. Furthermore, the most suitable recommender algorithm for recommending drugs based on the data from healthcare professionals are selected for this study. Opinion mining is applied on drug reviews, and a hybrid method is implemented to overcome the limitations of content-based and collaborative filtering methods, such as the cold start problem and increasing client preference. The system is developed and tested successfully. The proposed system can assist healthcare professionals in drug decision making and sustain the whole digital care pathway for various diseases.

Recommender systems have emerged as powerful tools for personalized infor- mation filtering and recommendation generation. However, these systems are not without their challenges and issues. This abstract explores the various issues faced by recommender systems and the implications they have on recommendation quality and user satisfaction. Data sparsity is a prevalent issue where recommender systems struggle to generate accurate recommendations due to limited or sparse user-item interaction data. This poses a significant challenge as it becomes difficult to capture user preferences and identify relevant items for recommendation. The cold start problem further compounds this issue by making it challenging to provide accurate recommendations for new users or items with limited or no historical data. Scalability is another issue, particularly for large-scale platforms with millions of users and items. As the volume of data grows, recommender systems encounter computational and efficiency challenges that hinder their ability to handle the increasing data size effectively. Diversity is an essential aspect of recommendations; ensuring users are exposed to a wide range of options. However, recommender systems often face a trade-off between accuracy and diversity. The overemphasis on popular or mainstream items can create filter bubbles and limit users’ exposure to novel or niche recommendations. Privacy and ethical concerns have gained significant attention in recommender systems. The collection and utilization of user data raise concerns about data privacy, algorithmic bias, and user manipulation. Ensuring transparency, fairness, and user control over their data and recommendations are critical aspects that need to be addressed. Evaluation of recommender systems poses its own set of challenges. Existing evaluation metrics primarily focus on accuracy and do not fully capture other important aspects such as diversity, novelty, and user satisfaction. Developing comprehensive evaluation frameworks that consider these factors is essential to assess the overall performance of recommender systems accurately. The research work highlights the issues faced by recommender systems, including data spar- sity, the cold start problem, scalability, diversity, privacy concerns, and evaluation challenges. Addressing these issues is crucial for enhancing the accuracy, diversity, and ethical consider- ations of recommender systems, ultimately improving the user experience and satisfaction. Future research should focus on developing innovative algorithms, evaluation methodolo- gies, and privacy-aware approaches to overcome these challenges and advance the field of recommender systems.

The traditional collaborative filtering recommendation systems have many deficiencies, which make them incompetent in the domain of clothing recommendation; we proposed a new ClothNet model based on CNN, RNN, collaborative filtering and the characteristics of the fashion industry. The accuracy and generalization performance of this model are improved compared with traditional systems. The visual information integrated into the ClothNet model enables the recommendation system to alleviate the cold start problem, and new clothes can be added to the recommendation list faster through the visual information. The addition of temporal information enables ClothNet sharply capturing the impact of seasonal and time changes on user preferences. However, because RNN and CNN have the disadvantage of requiring a large amount of data, combining RNN and CNN will make the model more difficult to converge, so we have adopted the LearningToRank training mode and obtained good results.

Widely used recommendation systems do not meet all industry requirements, so the search for more advanced methods for creating recommendations continues. The proposed new methods based on Generative Adversarial Networks (GAN) have a theoretical comparison with other recommendation algorithms; however, real-world comparisons are needed to introduce new methods in the industry. In our work, we compare recommendations from the Generative Adversarial Network with recommendation from the Deep Semantic Similarity Model (DSSM) on real-world case of airflight tickets. We found a way to train the GAN so that users receive appropriate recommendations, and during A/B testing, we noted that the GAN-based recommendation system can successfully compete with other neural networks in generating recommendations. One of the advantages of the proposed approach is that the GAN training process avoids a negative sampling, which causes a number of distortions in the final ratings of recommendations. Due to the ability of the GAN to generate new objects from the distribution of the training set, we assume that the Conditional GAN is able to solve the cold start problem.