scribus01/amazon-customer-sentiment-analysis

This project is a data analysis portfolio piece that looks at Amazon customer reviews to figure out how customers feel about their purchases and what patterns show up in their feedback. It combines Python for the data work with Power BI for building an interactive dashboard, and the goal is to turn raw review text into insights a business could actually act on. The dataset includes details like the reviewer's name, country, rating, review title, review text, and review date. The workflow starts with cleaning the data: removing duplicate entries, filling in or handling missing values, and converting ratings and dates into a usable format. From there, the review text itself is cleaned up using standard natural language processing steps, including lowercasing everything, stripping out punctuation and special characters, removing common filler words, and reducing words to their base form. Each review is then sorted into a sentiment category based on its star rating: four stars or higher counts as positive, three stars as neutral, and two stars or below as negative. With reviews labeled this way, the project explores patterns such as how sentiment is distributed overall, how ratings vary by country, how sentiment trends change over time, and which words show up most often in positive versus negative reviews, including word cloud visualizations for each. All of this feeds into a Power BI dashboard that shows total review counts, average rating, and breakdowns by sentiment, country, and time, with filters so a viewer can slice the data themselves. The README also lists a set of business recommendations drawn from the findings, such as investigating recurring complaints and promoting frequently praised features, along with ideas for future work like adding proper machine learning based sentiment prediction and turning the project into a live web application. The author describes themselves as an aspiring data analyst building this as a skills demonstration.