Supervised Machine Learning Based Fake Profile Detection Using User Ratings and Reviews in Recommender Systems

Ümmügülsüm Mengutaycı; Selma Ayşe Özel

Makale detayları

Supervised Machine Learning Based Fake Profile Detection Using User Ratings and Reviews in Recommender Systems

Yazarlar:

Ümmügülsüm Mengutaycı | Selma Ayşe Özel

Yayıncı:

Çanakkale Onsekiz Mart University

Ana Başlıklar:

Supervised Learning

Dergi:

Journal of Advanced Research in Natural and Applied Sciences Vol.11 Issue 2, pp.144-155

Belge Tipi:

Makale

Yayın Tarihi:

6/2025

Belge Dili:

İngilizce

Document ID (Org No.):

2001353228

pISSN:

2757-5195

Anahtar kelimeler:

Özet:

 Recommendation systems produce content based on users interests and aim to increase user satisfaction. In this way, the system keeps the user constantly active. Therefore, the reliability and robustness of these systems are essential. However, in recent years, with the influence of popular culture, recommendation systems have been struggling with fake users to highlight a particular product more or, conversely, to reduce the popularity of the product. Fake accounts mimic real user data and provide misleading information to the systems. This affects the accuracy of recommendation algorithms. This paper proposes a novel approach to detect fake user profiles by combining two different data sources: rating data and product reviews by using machine learning techniques, such as Decision Trees, Logistic Regression, Support Vector Machines, k-Nearest Neighbors and Naive Bayes algorithms. We also test the impact of integrating ensemble learning techniques on classification success. The research results show that the ensemble learning method Stack Classifier model has the highest detection success with an F1-score of 81.11. This highlights that the innovative approach using multiple data sources together provides a more robust and reliable solution for detecting fake profiles, thus improving the accuracy and efficiency of recommender systems.

 Recommendation systems produce content based on users interests and aim to increase user satisfaction. In this way, the system keeps the user constantly active. Therefore, the reliability and robustness of these systems are essential. However, in recent years, with the influence of popular culture, recommendation systems have been struggling with fake users to highlight a particular product more or, conversely, to reduce the popularity of the product. Fake accounts mimic real user data and provide misleading information to the systems. This affects the accuracy of recommendation algorithms. This paper proposes a novel approach to detect fake user profiles by combining two different data sources: rating data and product reviews by using machine learning techniques, such as Decision Trees, Logistic Regression, Support Vector Machines, k-Nearest Neighbors and Naive Bayes algorithms. We also test the impact of integrating ensemble learning techniques on classification success. The research results show that the ensemble learning method Stack Classifier model has the highest detection success with an F1-score of 81.11. This highlights that the innovative approach using multiple data sources together provides a more robust and reliable solution for detecting fake profiles, thus improving the accuracy and efficiency of recommender systems.

 Recommendation systems produce content based on users interests and aim to increase user satisfaction. In this way, the system keeps the user constantly active. Therefore, the reliability and robustness of these systems are essential. However, in recent years, with the influence of popular culture, recommendation systems have been struggling with fake users to highlight a particular product more or, conversely, to reduce the popularity of the product. Fake accounts mimic real user data and provide misleading information to the systems. This affects the accuracy of recommendation algorithms. This paper proposes a novel approach to detect fake user profiles by combining two different data sources: rating data and product reviews by using machine learning techniques, such as Decision Trees, Logistic Regression, Support Vector Machines, k-Nearest Neighbors and Naive Bayes algorithms. We also test the impact of integrating ensemble learning techniques on classification success. The research results show that the ensemble learning method Stack Classifier model has the highest detection success with an F1-score of 81.11. This highlights that the innovative approach using multiple data sources together provides a more robust and reliable solution for detecting fake profiles, thus improving the accuracy and efficiency of recommender systems.

 Recommendation systems produce content based on users interests and aim to increase user satisfaction. In this way, the system keeps the user constantly active. Therefore, the reliability and robustness of these systems are essential. However, in recent years, with the influence of popular culture, recommendation systems have been struggling with fake users to highlight a particular product more or, conversely, to reduce the popularity of the product. Fake accounts mimic real user data and provide misleading information to the systems. This affects the accuracy of recommendation algorithms. This paper proposes a novel approach to detect fake user profiles by combining two different data sources: rating data and product reviews by using machine learning techniques, such as Decision Trees, Logistic Regression, Support Vector Machines, k-Nearest Neighbors and Naive Bayes algorithms. We also test the impact of integrating ensemble learning techniques on classification success. The research results show that the ensemble learning method Stack Classifier model has the highest detection success with an F1-score of 81.11. This highlights that the innovative approach using multiple data sources together provides a more robust and reliable solution for detecting fake profiles, thus improving the accuracy and efficiency of recommender systems.

Makale detayları

Supervised Machine Learning Based Fake Profile Detection Using User Ratings and Reviews in Recommender Systems

Yazarlar:

Ümmügülsüm Mengutaycı | Selma Ayşe Özel

Yayıncı:

Çanakkale Onsekiz Mart University

Ana Başlıklar:

Supervised Learning

Dergi:

Journal of Advanced Research in Natural and Applied Sciences Vol.11 Issue 2, pp.144-155

Belge Tipi:

Makale

Yayın Tarihi:

6/2025

Belge Dili:

İngilizce

Document ID (Org No.):

2001353228

pISSN:

2757-5195

Recommender system

fake profile detection

machine learning

robustness

Sayı

Makale detayları

Supervised Machine Learning Based Fake Profile Detection Using User Ratings and Reviews in Recommender Systems

Yazarlar: Ümmügülsüm Mengutaycı | Selma Ayşe Özel

Yayıncı: Çanakkale Onsekiz Mart University

Ana Başlıklar: Supervised Learning

Dergi: Journal of Advanced Research in Natural and Applied Sciences Vol.11 Issue 2, pp.144-155

Belge Tipi: Makale

Yayın Tarihi: 6/2025

Belge Dili: İngilizce

Document ID (Org No.): 2001353228

pISSN: 2757-5195

Recommender system

fake profile detection

machine learning

robustness

Sayı

Yazarlar:

Ümmügülsüm Mengutaycı | Selma Ayşe Özel

Yayıncı:

Çanakkale Onsekiz Mart University

Ana Başlıklar:

Supervised Learning

Dergi:

Journal of Advanced Research in Natural and Applied Sciences Vol.11 Issue 2, pp.144-155

Belge Tipi:

Makale

Yayın Tarihi:

6/2025

Belge Dili:

İngilizce

Document ID (Org No.):

2001353228

pISSN:

2757-5195