prince-dubeyy/customer-feedback-analytics-ai

This project is a platform for automatically analyzing large volumes of customer feedback, like reviews or support tickets, that would take a human team far too long to read one by one. You upload a spreadsheet of raw feedback text, and the system runs it through a machine learning pipeline that predicts whether each piece of feedback is positive, neutral, or negative, along with a confidence score for how sure the model is. A separate AI layer also tries to pull out specific topics and complaints, distinguishing genuine product defects from simple feature requests. The results feed into a dashboard showing summary statistics, a breakdown of sentiment by pie chart, and a bar chart of the most common complaints. It can also compare one batch of feedback to a previous one to flag whether a particular issue is trending up or down over time. The sentiment model was trained on a large, evenly balanced set of seventy five thousand Yelp reviews, and the README reports it correctly classifies feedback about seventy four percent of the time on data it had not seen during training. The project has a working frontend built with React and a backend built with Python's FastAPI framework, both already deployed and linked from the README so you can try the live demo without setting anything up locally. Underneath, the sentiment prediction uses a standard text classification technique combining word frequency features with logistic regression, while a large language model handles the more nuanced topic and complaint extraction. The README frames this as a portfolio style project demonstrating an end to end data pipeline, from raw text to a business facing dashboard, rather than a maintained open source library meant for others to adopt into their own products. No license is mentioned in the README.