Sentilyze

📝 Notes

Sentilyze is a machine learning-based sentiment analysis platform developed from scratch to classify public opinion on the topic “No-Code vs. White Coding.” The project involves building a custom dataset by scraping and manually annotating over 1,000 Reddit comments. Using natural language processing (NLP) techniques and various classification algorithms—including Naive Bayes, SVM, Logistic Regression, Random Forest, and XGBoost—Sentilyze analyzes user sentiments as positive, negative, or neutral. Achieving the best precision of 67% with Naive Bayes, this project highlights our ability to collect real-world data, engineer effective NLP pipelines, and evaluate machine learning models for opinion mining and social media analysis.

📄 Project Description

Sentilyze is a machine learning-based sentiment analysis platform developed to understand public opinion on the topic of No-Code vs. White Coding. Unlike traditional approaches, this project involved building a custom dataset from scratch using real-world Reddit data.

• 📥 Custom Dataset Creation – Scraped and manually labeled over 1,000 Reddit comments for sentiment (positive, negative, neutral).
• 🧹 NLP Preprocessing – Applied techniques like tokenization, stop-word removal, lemmatization, and TF-IDF vectorization to prepare text data for analysis.
• 🤖 Multi-Model Sentiment Classification – Trained and compared five machine learning models (Naive Bayes, SVM, Logistic Regression, Random Forest, XGBoost), with Naive Bayes achieving 67% precision.

Sentilyze showcases our ability to build end-to-end NLP pipelines, perform real-world sentiment classification, and extract meaningful insights from social media platforms using machine learning.