Module title: Advanced Data Structures & Algorithms
Module code: COM713
Assessment Title: Data Structure & Algorithm with Python
Submitted by: Name Student ID
Introduction
In this assignment we have compared Content based recommender system,
collaborative filtering recommender system and graph based recommender systems from these
research papers. These systems aim to increase interaction and provide a richer experience for
users of e-commerce/online systems. The critical assessment of methodology used in each
paper is briefly discussed below.
1. A Graph-based Recommender System for Food Products
This paper presents a graph-based recommender system that utilizes reviews of the
food products rather than explicit item ratings to generate recommendations for users. These
reviews are provided by the users in textual data. Neighborhood-based and graph mining
techniques are used in [1]to create item profiles from review text and uses user review
feedback to find related products while balancing similarity and popularity. The proposed
algorithm divides the dataset into subgroups of similar products, making it scalable. The
recommender system follows a graph-based approach, incorporating k-nearest neighbors,
graph mining, and the PageRank algorithms.
Figure 1 The Workflow of Graph Based Recommender System
Page 2 of 5
2. An Improved Content Based Collaborative Filtering Algorithm
For Movie Recommendations
A hybrid approach of recommender systems, focusing on content-based and collaborative
filtering approaches is implemented in [2]. Content-based filtering considers the content of
items and user profiles, while collaborative filtering predicts user likes based on similar user
preferences.
The proposed algorithm utilizes tags and genres for content-based prediction and applies a set
matching comparator to find similarities between items. It reduces the sparsity of the user-
item rating matrix and uses collaborative filtering to generate final predictions. The
methodology is evaluated using Mean Absolute Error (MAE) and demonstrates better
performance compared to pure collaborative filtering and SVD on the Movielens dataset,
especially at higher sparsity levels.
Figure 2 Flowchart of Hybrid Collaborative Filtering algorithm for improved
3. Content Based Filtering Techniques in Recommendation System
using user preferences
The challenges faced by content-based filtering and collaborative filtering recommender
systems, such as the cold start problem, data sparsity, scalability, and accuracy are discussed
in [3]. The paper highlights the importance of recommendation systems in helping users find
valuable and interesting information in today's overwhelming data environment.
Recommender systems have become essential tools in e-commerce and online services,
helping users deal with information overload and find suitable products or services.
4. Evolution of Neural Collaborative Filtering for Recommender
Systems
The role of deep learning in enhancing recommender systems is explored in [4], with a focus
on its applications in modelling user-item interactions to provide more accurate
recommendations. The paper discusses influential deep learning architectures, such as the
Page 3 of 5
Wide and Deep model, Neural Collaborative Filtering, DeepFM, xDeepFM, and Neural
Factorization Machines, which have shown significant performance improvements over
traditional recommendation methods.
5. A Restaurant Recommender System Based on User Preference
and Location in Mobile Environment
[5] presents a novel approach to restaurant recommendation in a mobile environment. The
system is designed to address the limitations of traditional recommender systems in mobile
settings, which often struggle with dynamic and complex user contexts. The system builds a
user preference model based on the features of restaurants that users have visited. The
algorithm proposed in this paper combines user preference features with the distance between
the user and the restaurant. It employs cosine similarity to calculate the similarity between a
user's preferences and a restaurant's features.
Figure 3 The User Preference Model
Conclusion
Since the collaborative filtering is vulnerable to sparsity and the cold start problem due to its
dependency on user ratings. The proposed solution is content based filtering for
recommender systems where we have the customer feedback or review data. To enhance the
efficiency of the content based filtering recommender system we can combine it with deep
learning techniques as mentioned in [4].
Page 4 of 5
References:
[1] A. Mathur, S. K. Juguru, and M. Eirinaki, “A Graph-Based Recommender System for Food
Products,” in Proceedings - 2019 1st International Conference on Graph Computing, GC 2019,
Institute of Electrical and Electronics Engineers Inc., Sep. 2019, pp. 83–87. doi:
10.1109/GC46384.2019.00020.
[2] S. Aluru, Jaypee Institute of Information Technology University, University of Florida. College
of Engineering, IEEE Computer Society, IEEE Computer Society. Technical Committee on
Parallel Processing, and Institute of Electrical and Electronics Engineers, 2017 Tenth
International Conference on Contemporary Computing (IC3) : 10-12 August 2017, Jaypee
Institute of Information Technology, Noida, India.
[3] R. System, “Item Data Recommended Products User Item User Rating Data model Similarity
calculation Generate Recommendation,” 2016.
[4] A. I. Metsai, K. Karamitsios, K. Kotrotsios, P. Chatzimisios, G. Stalidis, and K. Goulianas,
“Evolution of Neural Collaborative Filtering for Recommender Systems,” in KST 2022 - 2022
14th International Conference on Knowledge and Smart Technology, Institute of Electrical and
Electronics Engineers Inc., 2022, pp. 86–90. doi: 10.1109/KST53302.2022.9729082.
[5] J. Zeng, F. Li, H. Liu, J. Wen, and S. Hirokawa, “A restaurant recommender system based on
user preference and location in mobile environment,” in Proceedings - 2016 5th IIAI
International Congress on Advanced Applied Informatics, IIAI-AAI 2016, Institute of Electrical
and Electronics Engineers Inc., Aug. 2016, pp. 55–60. doi: 10.1109/IIAI-AAI.2016.126.
Page 5 of 5