Michelin Journal: Data-Driven Insights for Travellers

Inspiration

We were inspired by the Michelin Guide, which provides rankings for restaurants and attractions across the world. Our goal was to expand on this concept using data-driven insights to help travelers make informed decisions about where to go, when to go, and what to expect, while also providing businesses with actionable insights to improve based on customer reviews.

What it does

The project analyzes restaurant, attraction, and point-of-interest (POI) data from multiple locations to offer travelers personalized recommendations based on preferences (e.g., sports, nightlife, cultural events). Using location clustering and sentiment analysis, we also help businesses understand the key drivers of positive reviews and improve their services.

How we built it

We collected and cleaned data from the given API, focusing on reviews and geographic data. Using K-means clustering, we identified optimal locations for tourists based on ratings and preferences. For sentiment analysis, we processed reviews to find key phrases that lead to positive reviews. The project also includes an interactive map where users can visualize the top-rated places. Notebooks for data processing, sentiment analysis, and map visualization were developed using Python, Jupyter, and other libraries.

Challenges we ran into

We faced challenges dealing with unbalanced and multilingual review data, as a significant portion of the reviews were not in English. This made sentiment analysis more complex. Additionally, some sentiment ratings provided by the dataset didn't correlate well with the polarity of the reviews, requiring further customization.

Accomplishments that we're proud of

We’re proud of successfully creating a system of an interactive map that not only helps travelers but also empowers businesses to make data-driven improvements. Our ability to integrate multiple components—such as location analysis, sentiment analysis, and interactive map visualization—into a cohesive project is a major accomplishment. We’re also proud of how we tackled challenges with data imbalance and translation.

What we learned

We learned how to handle real-world data challenges, such as dealing with imbalanced datasets and multilingual text processing. This project helped us improve our skills in data cleaning, machine learning algorithms (K-means clustering), and sentiment analysis. Additionally, we gained a deeper understanding of how data can drive both user experience improvements and business growth.

What's next for Michelin Journal: Data-Driven Insights for Travellers

Moving forward, we plan to expand the project to include more cities worldwide and improve our models for predicting business success based on customer reviews. We aim to develop a real-time recommendation system powered by large language models (LLMs) and offer even more personalized experiences for travelers. We also plan to refine the sentiment analysis to handle multilingual data more effectively.

Built With

• folium
• matplotlib
• natural-language-processing
• pandas
• python
• scikit-learn
• tableu