Machine Learning to Classify Religious Communities and Detect Extremism on Social Networks: ML to CRCs and DE Through Text Tweets on SNs

Introduction

Social networks are the best area for chatting, expressing opinions, and sharing beliefs and opinions (Allcott et al., 2020). This means that any expression can freed from the boundaries to be communicated to a potentially unlimited audience due to the ease and the speed of use. This work aims to discover religious orientation and extremism in different communities by studying textual data from social networks, more precisely Facebook and Twitter.

Artificial intelligence is important to understand the social context of extremism based on atheist ideology and religious texts within a big data context. Machine learning and deep learning algorithms create an excellent approach to analyzing and understanding texts within social networks (Liu et al., 2018).

Text mining (or “knowledge extraction”) is a specialized form of data mining (Xie et al., 2020)and is part of the field of Artificial Intelligence. Text mining is a form of computer processing whereby knowledge extracted according to criteria (features) in human-production texts. In practice this means modeling linguistic theories in computer systems for learning and statistics, coupled with technology for understanding the natural language itself.

The challenge is how to use these technologies to effectively classify religious orientation and detect extremism. In this work, we focus on investigating four research problems:

• Q1: Can religious expressions extracted from textual data reveal religious affiliation?

• Q2: Can extremism sentiment features extracted from textual data reveal extremism trends?

• Q3: What are the appropriate smart approaches and techniques for classifying religious communities from the texts in social networks?

• Q4: What are the smart approaches and techniques for detecting extremism within the texts in social networks?

R: We noticed that textual data contains religious elements such as references, books, names of scholars, special supplications, and specific religious phrases that distinguish one religion from another. This paper provides an overview of how religious elements of tweets and comments extracted and used in Machine Learning to detect religious affiliation and extremism.

Contributions

We processed textual data (In English language) related to Christianity, Judaism, Islam, and Atheism using four machine-learning algorithms, namely support vector machines (SVM), decision tree (DT), random forest (RF), and Naive Bayes (NB).

• 1.

  Collection and pre-processing of a dataset containing (3373) comments and tweets for religious community classification.

• 2.

  Collection and pre-processing of a dataset containing (29423) comments and tweets for extremism detection.

• 3.

  Creation of a set of dictionaries for the features of each religion (religious phrases, divine books, supplications, religious terms and religious figures ... etc.).

• 4.

  Development of an intelligent processing model for religious community classification using data collection and feature dictionaries. We found that Random Forest (RF) classification achieved the greatest accuracy (93.77%).

• 5.

  An intelligent prediction classifier for extremism detection has been developed using data collection as well as feature dictionaries. We found that Linear SVC achieves the greatest accuracy (97%).

The rest of the paper is structured as follows: Section 2 discusses related works. We will conduct a comparative study to illustrate the characteristics of each of them. Section 3 proposes a two-axis approach; religious classification and extremism detection. The empirical results and evaluation of the proposed classifier are discussed in section 4; and in the final section, recommendations and future research scopes are suggested.

Related Work

In this section, we review methods related to our field by examining works that propose similar or different approaches. In particular, we examine studies that rely on databases and dictionaries of features to solve the same problem. Where there some works that dealt in their contributions on machine learning techniques, while others deal on databases and dictionaries.