Hi, Product Hunters 👋 and thanks a lot @__tosh for the hunt and the last-minute product feedback! Alex is here along with my buddy and TravelRank co-founder @new_user_2805814e5b. In this totally odd time for launching travel-tech products, we’re asking for PH community feedback on TravelRank, the travel destination recommender powered by AI and neural network. Why we built this? We started working on this in a pre-COVID era, just about a year ago. I was looking and didn’t manage to find any tool, that had everything which I needed: - Combined travel history and stats (with the scratch-map look and feel) - Destinations (ideally pre-filtered, based on my travel history and preferences) - Single-best flight date, price and carrier suggestion (for the pre-selected destinations) - Social features (ranks, trending places, places visited by my friends, etc). So I decided to build one for myself and like-minded travel enthusiasts. We went on the mission to replace all the travel research heavy-lifting with just a few clicks via the AI and automation. So travelers like me can do less planning and enjoy more traveling. Our value proposition and key difference. We aim to: - Replace manual search and filters with accurate predictions - Offer unique and personalized travel recommendations - Save the time and nerves How we built this (and where the data are coming from). The product is built on MERN stack, AWS ML services (for matching and recommendations), AWS ElastiCache (for Ranks). We use Foursquare API to import the user checkins. The ML model was trained on an anonymized checkins dataset obtained from Foursquare. It contains 33M of checkins, created by 227K users in 415 most popular cities worldwide. On top of that, we calculate the matching score not only by interaction data from these 415 cites in our predictions (similar travel behavior), but also by meta-data similarity from a much larger city dataset (50K+, similar place), provided by 3rd party content API (based on Wikivoyage, Wikipedia, OSM city profile). Each of these 50K+ destinations is mapped to the closest airport in the 200km range dataset (0.5M of entries), so we could simultaneously offer the best flight price for each of our recommendations. Which is handy for the price comparison. Give us feedback While we didn’t manage to build everything that we wanted at once (social features, etc), done is better than perfect… We need some validation before we could continue, that’s why we can’t hold it until the pandemic ends. So give it a try, add your travel history, check if personalized travel recommendations hold value for you along with your travel stats and rank, and let us know, what was your impression. What features were useful and why? What features are missing and why you need it? We hope that you can put TravelRank recommendations into practice after travel safety improves. Meanwhile, we encourage you to be safe and avoid traveling. Stay healthy! Cheers, Alex & Ruslan! Animated icons by Icons8

Alex Shevchuk

TravelRank

Maker

5yr ago

Nataliya Vlasova

QRay app

Sounds interesting, but will it replace manual search? I, as an avid traveler, am unlikely to trust automatic selection, I need to compare everything manually. I wonder how accurate the predictions will be?

@nataliyavlasova I don't think any AI will replace manual ops or human sepervision completely any time soon. However it could greatly save the time on manual cumbersome operations, like category filtering or grouping the results, augmenting human in their research. At least we could aim for that, improving the quality and precision of recommendation with more data and user input.

@nataliyavlasova Another aspect, traveling is expensive. The technology should prove its capabilities first before the majority could shift the user habits and trust the AI to find perfectly matching destination. What do you think?