Abstract
With the increasing amount of data produced daily, it becomes very difficult for users to find resources suitable to their needs. Recommendation systems, which are capable of providing individualized suggestions or guiding the user to interesting or relevant resources within a large data space, are proposed for this purpose. In this article, the authors do a comprehensive assessment of recommendation models, propose a categorization scheme, analyze challenges, and explore unresolved issues. In addition, they highlight new trends and future visions in this field of study, emphasizing the need of merging ontologies and machine-learning algorithms to increase the accuracy and efficiency of the recommender systems.
TopIntroduction
As e-commerce expands and Big Data becomes more widespread, a massive quantity of data becomes available, and the number of Internet users increases. On the other hand, users are finding it difficult to acquire the products they desire. In a great knowledge area, the challenge is to help users in discovering and selecting resources. Recommendation systems have recently been a popular topic for researchers. Several big companies, such as Amazon and Netflix (Paul et al., 2017) have adopted these systems. Recommender Systems explore users' preferences in order to supply them with items that best meet their needs.
According to Klašnja-Milićević et al. (2015), Recommender systems are software tools and algorithms that provide suggestions for items that a user could find useful. These systems leverage the dependence principle between user-based and item-based tasks to select the most relevant item (Aggarwal, 2016).
Recommender systems remain to be a significant business tool for both Internet users and service providers; on the one hand, they improve company's sales, profits, and revenues, while also reducing the price of discovery and adoption in an online shopping.
However, Recommender systems are not limited to marketing products but emerged to support the healthcare community for decision-making and predict healthiness. In order to make user recommendations, the Recommender systems collect efficiently simple and standard data from different data sources, such as user evaluations and suggestions. Data belongs to different types and are mainly related to the elements proposed and the users receive the appropriate recommendations. Moreover, Data can be more informational, for example, users or items descriptions or constraints, social relationships, and user’s activities (Portugal et al., 2018).
In addition, with the explosion of the Web Services on the internet, such as YouTube, Amazon, eBay, and many others, Recommender systems are becoming increasingly important in our life. Recommender systems are now inevitable in our daily online trips, from e-commerce (suggest articles to buyers that may be of interest) to online advertising (suggest the proper contents to consumers based on their preferences).
Overall, recommender systems are becoming increasingly important in a variety of fields, most particularly healthcare. Here some examples:
Recently, in order to provide users with better recommendations, these systems have introduced Machine-Learning algorithms. However, given the large number of methods presented in the literature and the effectiveness of each approach, selecting an appropriate Machine-Learning algorithm for a Recommender system is challenging (Portugal et al., 2018).
Recommender systems are usually used to manage massive amounts of data and knowledge. Ontologies play a crucial role in knowledge representation, exchange, and reuse in these systems. Ontology-based recommenders are knowledge-based systems that employ ontologies to describe information about items and users in the recommendation process. Indeed, including ontological information in the recommendation process enriches the data with semantics and can overcome the limitations of conventional recommender systems.
According to recent studies (Chicaiza and Valdiviezo-Diaz, 2021), combining ontology domain information about users and items increases the accuracy and quality of suggestions while reducing the downsides of traditional recommender approaches like cold start and score dispersion. The ontology is useful for constructing user profiles with several dimensions, such as user comments, reviews, and ratings. Furthermore, the ontological model makes it easier to comprehend user preferences by representing them from several viewpoints.
Thus, the main contributions of this work are:
Key Terms in this Chapter
Ontology: The word ontology is of a Greek origin. In computer science, an ontology is a structured set of concepts that makes sense of information.
Items: Entities that a recommendation system recommends. For example, books are the items that a bookstore recommends.
Big Data: Big data also known as mega-data refers to all the digital data produced by the use of new technologies for personal or professional purposes.
Recommender system: Software tool that suggests recommendation, which users might apply to achieve their goals.
Collaborative Filtering: Recommends items to a specific user based on the interests of many other users.
Dataset: Datasets are commonly used in machine learning. They regroup a set of linked data that can be in various formats (texts, numbers, images, videos, etc.).
Cold Start: An issue that a recommendation system encounters when there is very little data available about the user or the item.
Machine Learning: Machine learning is an artificial intelligence technology that allows computers to learn without being explicitly programmed for this purpose.