Challenges

Finding Authentic Dataset

Our initial plan was to incorporate various user personalized factors such as budget, user mood, time of year but there was no authentic datasets available which can cater to our needs. One way to do this was data Collection through user surveys but it would not have been scalable and time consuming. Hence, We will be using Yelp data set for helping us in building Businesses - User Ratings Matrix.

High Sparsity and Huge Size of Dataset

Yelp Dataset was huge in size and largely sparse and getting ratings matrix involved a lot of computation power. To address these challenges we incorporated AutoEncoder Model for sparsity. To handle huge size we split the data and used parallel processing to fasten computation.

Cold Start Problem

As dataset didn’t have any user profile data, it was difficult to address cold start problem. We tried web crawling on TripAdvisor for generating user profiles but scrapping is blocked on such sites. To address this, we developed new User interface and asked his perferences and provided recommendations accordingly.

Evaluation of System

As we know evaluating recommender systems is tricky, but we evaluated our recommneder systems of Matrix Factorization and Auto+Collabrative Filtering using RMSE as metric for changing epochs. Feedback through A/B testing would have really helped but is challenging. Usefulness of the tool demonstrated by conducting user testing with travel preferences and backgrounds.

Future Work

In conclusion, Travelix will be end tool for all travel needs. Now with Travelix, You will not need to go to Google flights for flight search, Trip Advisor for Hotels, Rental services website separately. Our tool will provide all things in one place hence simplifying the travel process for all users. It is a unique solution that fills the gap in the market by offering tailored travel recommendations, which are not currently provided by other platforms like TripAdvisor or Expedia. In the future, we plan to incorporate below mentioned things. Overall, Travelix has the potential to revolutionize the way people plan and experience their travel adventures. With continued development and improvements, Travelix can become a All in One solution for all travel-related needs, providing users with personalized and unforgettable travel experiences.

Building a complete Iternary by incorporating all aspects of travel such as preferred flights, places to eat, hotels to stay, rental services etc.

Incorporate various user personalized factors such as budget, user mood, time of year to provide a more personalized experience.

Expanding the dataset used for generating recommendations and incorporate additional features such as weather and seasonal trends to provide more relevant and accurate recommendations.