dlifeofjay/Customer-Review-DS-ML

This project shows an end to end data science and ML project finding the SWOT of 2 big hotels in lagos nigeria through finding insights from customer reviews # Customer Review DS-ML This project is a simple end-to-end customer review analysis workflow for hotel reviews, with a focus on Marriott and Radisson Blu properties. It brings together raw review data, sentiment analysis, exploratory notebooks, and an interactive dashboard so the feedback can be understood more clearly. The main goal is to turn review text into something useful: identify sentiment, highlight common themes, and present the results in a clean dashboard. ## What this project does The repository includes: - review datasets for Marriott and Radisson Blu hotels - preprocessing and analysis notebooks for exploring the data - sentiment classification artifacts for both hotel brands - a Streamlit dashboard for viewing positive and negative review insights ## Project structure - app.py: the Streamlit dashboard application - data/: raw and processed review datasets used by the project - Notebooks/: Jupyter notebooks for analysis and exploration - artifacts/: saved sentiment model files and tokenizers - images/: images related to the project - src/: supporting project files and utilities ## Setup This project is built with Python. A virtual environment is recommended. 1. Clone the repository 2. Create and activate a virtual environment 3. Install the required packages: ```bash pip install -r requirements.txt ``` If you run the dashboard for the first time, the app may download NLTK stopwords automatically. ## Running the dashboard To launch the dashboard locally: ```bash streamlit run app.py ``` The app reads the processed review data from the data folder and displays visual summaries of positive and negative review themes. ## Data and analysis flow The workflow is fairly straightforward: 1. Review data is loaded from the CSV files in the data folder. 2. The notebooks explore the reviews, clean the text, and prepare the data for analysis. 3. Sentiment labels are applied using the saved model artifacts. 4. The results are displayed in the Str …