Friendship Prediction in Social Networks Using Developed Extreme Learning Machine with Kernel Reduction and Probabilistic Calculation

Tharwat, Muhammed E. Abd Alkhalec; Fudzee, Mohd Farhan Md; Kasim, Shahreen; Ramli, Azizul Azhar; Madni, Syed Hamid Hussain

doi:10.1007/978-3-031-00828-3_6

Cited by 1 publication

(1 citation statement)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The other component of the proposed framework is the implicit link prediction-probabilistic reduced kernel extreme learning machine ILP-PRKELM (Tharwat et al, 2022) which is presented in the pseudocode of Algorithm 2. The input of the algorithm is the social network SN, the minimum considered a number of links for prediction NL and the size of the subset used for building the model size of the subset.…”

Section: Implicit Link Predictionmentioning

confidence: 99%

Community Aware Recommendation System with Explicit and Implicit Link Prediction

Tharwat,

Md Fudzee,

Kasim

et al. 2023

Journal of Computer Science

View full text Add to dashboard Cite

Recommendation systems are essential tools that help users discover content they may be interested in, amidst the vast amount of information available online. However, current methods, such as using historical user-item interactions and collaborative filtering, have limitations in accurately predicting user preferences. Our research aims to address these challenges and improve the performance of recommendation systems. In this article, we propose a new approach to recommendation systems using a method called Probabilistic Matrix Factorization (PMF). We transform the standard PMF method into a communitybased PMF that takes into account implicit relationships between users and items.To achieve this, we use a machine learning technique called Reduced Kernel Extreme Learning Machine (RKELM). Our proposed framework is designed to integrate these implicit relationships and identify communities of users with similar preferences based on PMF. We conducted a comparative analysis of our newly developed model against existing methods, using two well-known datasets. Various performance metrics, such as prediction errors, were employed to evaluate the effectiveness of our proposed community-based PMF approach with RKELM. Our model demonstrates improved performance, achieving a 7% improvement for the Douban dataset and a 4% improvement for the Last.fm dataset. Despite the improvements demonstrated by our model, potential limitations and challenges may still exist, such as scalability to larger datasets or adaptability to different domains. Future work could explore these aspects and investigate further enhancements to our approach.

show abstract

Section: Implicit Link Predictionmentioning

confidence: 99%