INF2179: Machine Learning Executive Project
This repository contains the machine learning pipeline and strategic executive report developed to detect fraudulent and deceptive reviews on Yelp. The core objective was to build a predictive model that protects platform integrity and ensures a trustworthy user experience.
Fake reviews erode user trust and degrade the core value proposition of review-based platforms. This project moved beyond simple accuracy metrics to focus on actionable Trust & Safety interventions.
- Strategic Threat Mitigation: Developed an NLP-driven classification strategy to identify deceptive language patterns, protecting the platform from astroturfing and coordinated review manipulation.
- Executive Translation: Synthesized complex NLP and model evaluation metrics into an Executive Report, providing clear product recommendations for integrating the model into a live moderation queue.
The project required a full-stack data science approach, focusing on text mining, natural language processing, and supervised learning.
- Natural Language Processing (NLP): Conducted extensive text pre-processing, tokenization, and sentiment analysis to engineer features from raw unstructured review text.
- Predictive Modeling: Trained, evaluated, and tuned classification models (including Logistic Regression and ensemble methods) to isolate the linguistic signatures of fake reviews.
- Model Evaluation: Prioritized precision/recall balance to ensure legitimate user reviews were not mistakenly flagged (minimizing false positives for a frictionless UX).
INF2179_Group 1 Executive Report.pdf: The strategic business report outlining the problem space, methodology, and final product recommendations.INF2179 - ML Yelp Fake Reviews.pdf: The technical presentation deck detailing the NLP pipeline and model performance.[Insert Your Notebook Name].ipynb: The core Python data analysis, NLP, and machine learning pipeline.