Twitter Account Type Prediction

Overview

This project focuses on predicting whether Twitter accounts are operated by bots or humans using machine learning techniques. Leveraging a dataset of 37,438 Twitter user accounts, I employed methods such as Logistic Regression, Random Forest, and Support Vector Machines (SVM) through PySpark for scalable data processing. My goal is to enhance understanding and classification accuracy of bot accounts, which are increasingly pertinent in social media analytics and cybersecurity.

Methodology

Data Preparation and Cleaning:

• I began by cleaning the data, removing redundant columns, and addressing missing values.
• Feature engineering was employed to create new indicators such as language presence, description completeness, and background image URL availability.
• Statistical validation and exploratory data analysis (EDA) helped me comprehend data distributions and correlations.

Machine Learning Model Implementation:

• PySpark was crucial for setting up my distributed data analysis environment.
• Using Spark SQL and data frame APIs, I efficiently manipulated and preprocessed the data.
• I implemented Logistic Regression, Random Forest, and SVM models, ensuring thorough preprocessing including categorical encoding and feature vector assembly.

Findings and Industry Use Cases

Findings:

• My experiments revealed that Random Forest achieved superior accuracy and precision compared to Logistic Regression and SVM.
• Logistic Regression, however, excelled in recall, a critical metric for accurately identifying bot accounts.
• Key features such as user activity metrics and profile characteristics emerged as significant in distinguishing bots from human-operated accounts.

Industry Use Cases:

• • Social Media Monitoring: Effective identification of bots can help mitigate misinformation and maintain platform integrity.
  • Cybersecurity: Enhanced bot detection aids in fraud prevention and strengthens security measures.
  • Marketing and Customer Service: Accurate identification of genuine customer interactions improves segmentation and enhances personalized engagement strategies.