Machine Learning for Book Recommendations
Machine Learning for Book Recommendations
- Introduction
- Proposed System
- References
- Conclusion
- Experiment Results
Volume 8, Issue 5, May – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Book Recommendation System Using Machine
Learning
Shivam Goswami Shivam Mishra
Department of Electronics and Communication Engineering Department of Electronics and Communication Engineering
KIET Group of Institutions KIET Group of Institutions
Ghaziabad, India Ghaziabad, India
Shobhit Chaudhary Ajeet Pratap Singh
Department of Electronics and Communication Engineering, Department of Electronics and Communication Engineering
KIET Group of Institutions KIET Group of Institutions
Ghaziabad, India Ghaziabad, India
Abstract:- In today’s world, people focus on reviews and An overview of the literature on associated studies is
ratings available online. Recommendation system works given in Section 2. Section 3 outlines the specific ways in
like a filtering technique that provide information on the which the suggested system. The evaluation process and its
basis of previous user experience. These technologies findings are given in Section 4. In Section 5, we made some
used in various websites and apps to provide better conclusion.
experience to user and also used to gain profit. This
study suggests fast and simple book recommendations II. RELATED WORK
method which help user to select right books based on
their choice. Its thorough description is offered together During the last few years, the study of recommender
with the general architecture. In this system system has expanded significantly. Recommendation
collaborative filtering was employed. lastly, additional systems employ a variety of strategies to make appropriate
details are depicted with test results based on the online recommendations. Conventionally Both collaborative and
survey. content- based filtering are employed. In content-based
filtering system recommends an item to user on the basis of
Keywords:- Book, Collaborative Filtering, Recommendation user preference for an item [2]. Contrarily, it is not
System, User Ratings. compulsory that user preferred content in past will also
prefer same content in the future. Information about their
I. INTRODUCTION. preferences can be gathered based on how they rate items.
The effective application of recommender system is the
The magnitude of knowledge accessible nowadays is Amazon corporation recommends a variety of their items in a
exploding, especially online. It gets difficult to find the data very effective and efficient manner. Furthermore, Hybrid
you required. Such issues are what recommendation systems recommender system is used that combines the two
are meant to address. Without having to actively search the previously mentioned methods which can be used by most of
web, one can swiftly acquire pertinent information with their the online sites to enhance their efficiency [3].
assistance. As a result, recommendations systems are used
by numerous websites today to market and sell their items Today, there are numerous book recommendation
[1]. A long row of things such as books, music, movies, systems. The developers of these technologies leverage data
services and many more can be suggested to the customer from social media to understand user preferences in order to
on the basis of their online profiles or even social media make the best recommendations.
accounts, online history, including links they have visited,
search activity, including the time and duration of their In this work, we provide a recommendation system
activity, and other online behaviors. These technologies are that assesses the buyer’s interests in order to offer the
helping online stores boost up their revenue. customer the best choice. Using collaborative filtering,
association rule mining and content-based filtering, the
In this study, we show that recommendation algorithm quality and content are taken into account from [4]. Our
works like a filtering system that shows user a book on the aim is to provide user recommendations by considering
basis of rating, reviews and past activity of an user. We experience of search in a coherent and well organized
developed a system that gain an understanding of user manner.
preferences by analyzing user rating on books and generate
a record or table that user is most likely want to read based
on their choice.
IJISRT23MAY253 [Link] 816
Volume 8, Issue 5, May – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
III. PROPOSED SYSTEM
The current services require a better user profile
information [5]. customers sign up for such platforms,
peruse reading material, and give them ratings, write
reviews, pass along suggestion. A system formulates its
recommendations using this data. Our recommendation
engine emphasizes efficiency and quickness. After applying,
the user must choose their top books from at least three
different categories. The suggestion offered by the system
based on this data. The user can keep rating books, buying
them, and adding them to their reading list, all of which will
help the recommender system to improve accuracy. A user
can update a database by rating the books using a Fig 2 web page
straightforward search and filtering interface, and they will
then receive the right book recommendations. These C. Providing Recommendations Following user preference
suggestions based on a collaborative filtering process. learning, the system offers suggestions (see Fig. 3).
Recommended list can be bisected intodifferent group or
class.
Fig 3 suggestion row
We employed collaborative filtering to make the
suggestions for our system. It gathers an immense quantity
Fig 1 collaborative filtering of data regardinga customer tastes collaborative filtering is a
method for providing recommendations on various products.
A. Accumulate Details. As it was previously said [7], the idea behind this method is
The data used in a recommendation system is crucial. that if two users (users A and B in Fig. 4) have similar views
In recommender system data should be gathered to make for one item, it is more likely that they will have similar
system more accurate and reliable. We selected information preferences for other items rather than with a different user.
for that aim from already-available resources such as a Realizations from a collaborative filtering can take many
website that offers thousands of books that contains various diverse forms. One of them is constructed using
genres and categories. Data accumulated and stored in neighborhood-based methods that compute similarity
MYSQL. There are roughly 20,000 popular books total. between two items.
B. Evaluation of User liking
There are two distinct kinds of customer in the system:
unregistered and registered customers [6]. Visitors can
search a book according to their preference and read basic
information about that book. Registered users have the
advantage to rate the books and get system recommendation.
After registering, firstly a customer was requested to
choose two or more of his preferred type (see Fig. 2) and
then rate at least 10 books to know their preferences. An
inbuild search box given in a system for a customer to find
book on the basis of their preferred choice. Fig. 4. Preference table
IJISRT23MAY253 [Link] 817
Volume 8, Issue 5, May – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
IV. EXPERIMENT RESULTS [3]. Pijitra Jomsri, “Book recommendation System for
Digital Library based on User profile by using
Three types of recommendation systems are Association rule”, Thailand, IEEE, 2014, pp 130- 134.
distinguished: A user study analyses quantitative and [4]. Raymond J Mooney and Loriene Roy. “Content-based
qualitative metrics and watches the behavior of a group of book recommending using learning for text
participants who will interact with the system, while an categorization.” In: Proceedings of the fifth ACM
offline experiment simulates user behavior. Actual users of conference on Digital libraries. ACM. 2000, pp. 195–
the system perform tasks while it is being evaluated. We 204
chose to conduct our research via an online survey since, as [5]. Anand Shanker Tewari, Kumari Priyanka ,”Book Re-
far as we are aware, there is no database that is currently commendation system based on Collaborative filtering
available that is appropriate for our work and because online and association rule mining for college
evaluation has not yet been implemented. As a result, we students”,IEEE,2014,pp 135-138
made the decision to develop an online tool to evaluate the [6]. Xiaoyuan Su and Taghi M Khoshgoftaar. “A survey of
performance of our system. collaborative filtering techniques.” In: Advances in
artificial intelligence 2009 (2009), p. 4
We chose a group of independent readers and asked [7]. Jinny Cho, Ryan Gorey, Sofia Serrano, Shatian Wang,
them to rate the system on factors like the relevance of the JordiKai Watanabe-Inouye, “Book Recommendation
recommendations and the practicalities and the system's System” Winter 2016.
functionality (0–10). A higher score suggests that the advice
is more pertinent.
Fig 6 Experiment Result
Despite the excellent outcomes of the internet survey,
users identified several system-related issues. One of them
was the issue that suggestions are less helpful when the
same book appears in multiple categories. This is due to some
genres being closely related, making it possible for things
from those genres to appear in both genres.
V. CONCLUSION
In this research, book recommendation system based
on collaborative filtering described. The primary objective
was to make recommending system efficient. The main
aim to develop a system that can provide users with
appropriate suggestion without requiring extensive profile
information, search history, etc. The output of the
experiments shows that the suggested method offers
appropriate suggestion to users.
REFERENCES
[1]. Zhao kai, Lu Peng-Yu ,“ Improved Collaborative
filtering approach based on User similarity
Combination”,IEEE International Conference on
Management Science & Engineering ,Helsinki, Finland
,2014,pp 238-241.
[2]. Pasquale Lops, Marco De Gemmis, and Giovanni
Semeraro. “Content-based recommender systems:
State of the art and trends.” In: Recommender systems
handbook. Springer, 2011, pp. 73–105.
IJISRT23MAY253 [Link] 818
Common questions
Powered by AICollaborative filtering provides personalized recommendations by identifying patterns in user behavior and aligning them with similar users' preferences, offering significant advantage in user engagement and satisfaction . However, limitations include its dependence on extensive user data, potential challenges in new user scenarios (cold start problem), and reduced effectiveness when book suggestions overlap categories, leading to less diverse recommendations .
The system's architecture contributes to speed and efficiency by structuring data collection and processing in a streamlined manner. By requiring minimal user interaction data initially, focusing on SQL databases for rapid data retrieval, and employing an efficient collaborative filtering algorithm, the system offers quick, relevant recommendations. This architecture minimizes processing time while maximizing the relevance and personalization of the user experience .
Data is critical for the effectiveness of the proposed recommendation system as it forms the basis for understanding user preferences. The system accumulates data from user ratings, preferences, and social media activity. This comprehensive collection allows the system to make accurate recommendations using collaborative filtering techniques laid in databases such as MYSQL with large volumes of book titles and categories .
Online interactions and social media data are integral to the recommendation algorithms as they provide insights into user preferences and behaviors beyond direct user ratings. The algorithms perceive patterns in online activity, such as links visited and content interacted with, allowing for a more nuanced understanding of user interests, which can then be translated into better-targeted book recommendations .
The book recommendation system enhances user experience by employing collaborative filtering techniques. It analyzes user ratings on books and generates recommendations based on these preferences. The system learns from the past behavior of similar users to suggest books a user is likely to enjoy, thereby providing a personalized experience without requiring extensive personal data. This method allows for efficient and relevant suggestions, which are crucial for user satisfaction .
Hybrid recommender systems hold significance as they combine both collaborative and content-based filtering techniques to enhance recommendation accuracy and reliability. In the study, while collaborative filtering was primarily employed, the concept of hybrid systems is acknowledged as a means to overcome limitations of individual methods by leveraging strengths from both approaches, thus providing more comprehensive user recommendations .
The recommendation system uses user feedback to refine its suggestions dynamically. By allowing registered users to rate books and provide feedback, the system adjusts its algorithm to better align with current user tastes and trends. This continuous loop of feedback and adjustment helps in increasing the accuracy of recommendations and enhancing user satisfaction by tailoring to evolving preferences .
The system mitigates the cold start problem by initially asking new users to select their favorite categories and rate a minimum of ten books. This provides an initial dataset from which preferences can be inferred, allowing the system to make plausible recommendations even when there's limited historical data. These initial actions create a minimal profile that guides the system in offering meaningful suggestions despite the lack of extensive user interaction history .
The evaluation of the recommendation system employed both online surveys and user studies because no appropriate existing database was available. This approach focused on gathering qualitative and quantitative user feedback on recommendation relevance and system functionality. The findings indicated that while the recommendations were generally relevant, challenges such as the overlap of book suggestions across genres were identified .
User behavior analysis is crucial as it involves studying patterns in user interactions, such as book ratings, purchase history, and search activity. By analyzing these behaviors, the system identifies trends and preferences that help in constructing accurate user profiles. This information directly informs the collaborative filtering process, improving recommendation precision by closely aligning suggestions to user-established preferences .
