Details

Autor(en) / Beteiligte

• Kanwal, Moona
• Khan, Najeed A.
• Ismat, Najma
• Khan, Aftab A.
• Khan, Muzammil Ahmad

Titel

Machine Learning Approach to Classification of Online Users by Exploiting Information Seeking Behavior

Ist Teil von

• IEEE access, 2024-01, Vol.12, p.1-1

Ort / Verlag

Piscataway: IEEE

Erscheinungsjahr

2024

Quelle

EZB Electronic Journals Library

Beschreibungen/Notizen

• In today's world, technology has engulfed the internet with an excessive amount of unfiltered, spontaneous, and incessant data from multiple sources. Complex algorithms are designed to present information effectively based on user intent. The online experience of users is a combination of various behaviors exhibited to seek information, including searching, sharing, and verifying information. However, this multifaceted user behavior is yet to be explored comprehensively. This research contributes towards proposing a user intent-machine learning model for classifying users based on their online search, share, and verification behavior, identifying different types of users based on their online engagement, and demonstrating that dynamic online interactions can be classified based on their searching, sharing, and verifying behavior. User feedback on online behavior and practices is gathered through a questionnaire, encompassing participants from diverse gender, occupational, and age demographics. Following the extensive feature engineering, the significant features are presented to K-Mean clustering to identify user intent classes or profiles and their characteristics. A supervised learning Linear Discriminant Analysis Classifier (LDAC) is then trained on data to classify these classes. The proposed framework successfully predicted the user intent class with 80% accuracy. The model is further tested on users' dynamic interaction data gathered through a second user study. The information search, share, and verify activity data is transformed to fit the model and labeled by human raters using the user profiles resulting from clustering. The research achieves an Inter-rater reliability (IRR) of 60%, whereas the model predicted the user with 67% accuracy. This research indicates that a user's purpose in seeking information, their willingness to share information on social media, and their inclination to view information as credible can all contribute to understanding their intentions, identifying behavioral similarities, and can be used to recognize intent through dynamic interactions that can be used in targeted marketing, and search engine optimization.