Behavior Based Group Recommendation from Social Media Dataset by Using Deep Learning and Topic Modeling

Md. Saddam Hossain Mukta, Jubaer Ahmed, Mohaimenul Azam Khan Raiaan, Nur Mohammad Fahad, Muhammad Nazrul Islam, Nafiz Imtiaz, Md. Adnanul Islam, Mohammed Eunus Ali, Sami Azam

Introduction

Traditional group recommendation systems often rely on explicit social ties—friendships, followers, or network structures. However, these connections do not always capture the underlying psychological motivations that drive collective behavior. This research explores "Basic Human Values" as a more robust foundation for identifying cohesive groups. By focusing on values like hedonism, the authors propose a system that predicts real-world interests based on digital footprints rather than just social graphs.

Methodology

The researchers developed a dual-track approach to identify groups based on psychological alignment, specifically targeting the value of Hedonism (the pursuit of pleasure and gratification).

Figure 1: Overall methodology flowchart showing the pipeline from data collection and preprocessing to model building and validation.

1. Graph-based Group Identification (GHV)

This method constructs a user graph where edges represent similarities in value scores derived from the IBM Personality Insight API. A Graph Neural Network (GNN) processes these connections to learn latent patterns, which are then partitioned using Spectral Clustering to find optimal groups.

2. Contextual-Psychological Identification (CPHV)

This pipeline utilizes deep learning and psycholinguistics. It employs BERT and Latent Semantic Analysis (LSA) to capture the contextual nuances of tweets, while Linguistic Inquiry and Word Count (LIWC) analyzes the psychological lexicon. Density-based clustering then identifies compact, psychologically aligned user groups.

Datasets and Validation

The study leveraged two distinct datasets scraped from Twitter:

  • Identification Dataset: 1,140 unverified users with 3.2 million tweets used for discovering value-based patterns.
  • Validation Dataset: 300 users categorized as "Movie Enthusiasts" or "Tech Enthusiasts" based on specific hashtags (e.g., #marvel, #gadgets). This served to verify if the identified "hedonist" groups actually shared these real-world hobbies.

Key Results and Insights

The CPHV model demonstrated superior performance, achieving a Silhouette Clustering Coefficient (SCC) of 76% and an Intra-cluster Correlation Coefficient (ICC) of 60%. These metrics confirm that the resulting groups are both distinct and internally cohesive compared to state-of-the-art models.

Figure 2: 2D Cluster Visualization showing the distribution of users into four distinct groups based on principal components.

The research highlights that psychological traits, particularly hedonism, are highly visible in digital behavior. Users valuing hedonism frequently discuss travel, fashion, and entertainment, making them ideal candidates for group-based recommendations even in the absence of direct social links. The "hedonist" groups identified by the model showed a high average hedonism score (approx. 0.70), correlating strongly with the interests of the validation groups.

Conclusion and Future Work

While the model shows promise, challenges remain regarding data sparsity for less active users and the temporal shifts in human values over time. Future research aims to extend this work to multimodal analysis—incorporating image data from platforms like Instagram and professional data from LinkedIn—to further refine these behavioral profiles.

Summary updated

Abstract

In this digital era, users frequently share their thoughts, preferences, and ideas through social media, which reflect their Basic Human Values. Basic Human Values (aka values) are the fundamental aspects of human behavior, which define what we consider important, and worth having and pursue them. Existing studies identify the values of individuals from different social network usages such as Facebook and Reddit. However, discovering the similarity (or diversity) of value priorities among the members in a group is important since we can reveal many interesting insights such as finding a set of target customers, identifying the chain of misdeed groups, searching for similar acquaintances in workplaces, etc. In this paper, a graph dataset is compiled using the strongest correlation among the features and then we apply a graph clustering technique to identify a suitable hedonist group (i.e., one dimension of values) for users’ recommendations. Then, we also propose a behavior based (i.e., value ) group recommendation technique by analyzing users’ contextual and psychological attributes. Finally, we validate those group members in real life by introducing two hypotheses. In particular, we analyze the tweets of a total of 1140 users collected from Twitter. We obtain a substantial intra-cluster correlation coefficient (ICC) and silhouette clustering coefficient (SCC) scores of 65% and 76%, respectively, among the members in our discovered group.
BibTeX
@article{mukta2024behavior,
  author = {Md. Saddam Hossain Mukta and Jubaer Ahmed and Mohaimenul Azam Khan Raiaan and Nur Mohammad Fahad and Muhammad Nazrul Islam and Nafiz Imtiaz and Md. Adnanul Islam and Mohammed Eunus Ali and Sami Azam},
  title = {{Behavior Based Group Recommendation from Social Media Dataset by Using Deep Learning and Topic Modeling}},
  journal = {Sn Comput. sci},
  year = {2024},
  doi = {10.1007/s42979-024-03055-1},
}
Figure