Fraud Detection in Social Media
NLP with text mining approach to detect cyberbullying, fake reviews and misinformation in social media platforms.
1. Project Overview
NLP with text mining approach to detect cyberbullying, fake reviews and misinformation in social media platforms.
2. Problem / Challenge
Social media platforms face significant challenges with cyberbullying, fake reviews and misinformation. This project aimed to detect such harmful content automatically.
3. Approach / Methodology
Applied Natural Language Processing and text mining techniques including TF-IDF vectorisation and Naive Bayes classification to identify fraudulent or harmful social media content.
4. My Contribution
Designed the NLP pipeline, implemented TF-IDF feature extraction, trained and evaluated the Naive Bayes classifier, and documented the findings as part of my MSc dissertation.
5. Results / Outcome
The project demonstrates how natural language processing and text mining can be applied to social media content to flag cyberbullying, fake reviews and misinformation. A TF-IDF representation combined with a Naive Bayes classifier was used to establish a working text-classification baseline, implemented and documented in Google Colab.
6. What I Learned
Working on this project built practical experience with text preprocessing, feature extraction using TF-IDF, and the limitations of bag-of-words representations. It also highlighted that evaluating a classifier on imbalanced data requires precision and recall rather than accuracy alone, and that the cost of a false positive is a product decision as much as a technical one.