Wander

Inspiration

We were inspired by the trip planning journey, and our own personal experiences traveling to new cities and having trouble finding activities that interest us. We wanted to create a user-friendly way to explore a city tailored to you at the tip of your fingers.

What it does

Our application, Wander, is designed to revolutionize the way travelers explore new cities by leveraging machine learning. We have developed a machine learning model that curates a list of activities recommend to you to do in a new city based on past tourist attractions you have liked or disliked in other cities. By collecting data on tourist attractions across various cities and integrating user preferences, we provide a tailored experience that evolves with every trip.

We first select the city we are exploring. Now we are presented with a curated list of attractions from our database. We select which ones we would like to save for a future visit, and which ones we are not interested in based on a bunch of different factors. This feedback is used to train our recommender model that identifies patterns in user preferences. The model learns from each selection to refine recommendations for future trips. Finally, we can see all of our “liked” locations here!

How we built it

Since our application depends heavily on user-generated data, we used Generative AI to curate a list of tourist attractions in various cities including features such as rating, cost, address, suitability, and estimated time. Once of our team members, acted as a user of our app, deciding whether or not they would visit these stops. Based off of this, we were able to train multiple machine learning models (Logistic Regression, Decision Trees, Random Forest, Neural Networks, and CNN) and found that CNN resulted in 82 percent. However, we had trouble integrating CNN with our prediction dataset and used our Logistic Regression Model which had a high accuracy as well. We then used this to predict tourist attractions in Charlottesville the user would find interesting with our model and displayed in our Anvil App which integrated Google Collab (where we wrote the Machine Learning code with scikit-learn and TensorFlow.

Challenges we ran into

Integrating the CNN models, and getting a high accuracy rate since we had a small dataset to work with.

Accomplishments that we're proud of and What we learn

We learned so much through this process! We had never used Anvil before, and learned about various machine learning models, it would so exciting to see our app come together, and we look forward to learning more about Machine Learning and its applications.

What's next for Wander

We want to incorporate more features such as sharing tourist attractions with other people around you if you find an attraction interesting. Additionally we wanted to make our model more robust so the recommender algorithm is updated with new information with every city a user goes to and attractions they enjoy.