Show HN: Flight Search on steroids: find better flights through Machine Learning | Hacker News Reader

Show HN: Flight Search on steroids: find better flights through Machine Learning

tripdelta.com

16 pointsThePhysicist18 comments

I know naming startups and getting domains is hard, but Delta is the 2nd largest airline in the world [1]. Naming your air travel site TripDelta seems likely to raise some significant trademark issues.

[1] https://en.m.wikipedia.org/wiki/World's_largest_airlines

We talked to lawyers and think that tripdelta, as a combination of the (for us) important keyword trip and the math term delta to indicate savings, is a word creation in its own. Hope they’re right. Thank you for sharing your concerns

Cool UI, tried it on my recent flight, but for my case I could find better deal on Orbitz.

Example: WAW -> SFO (July 18, Aug 2)

IMO best: WAW -> AMS -> SFO (14h25m), SFO -> CDG -> WAW (14h5m) 1256$

Your flights are a bit cheaper, but the shortest available outbound is 18h40m.

Maybe you can tweak your algorithm to optimize for shortest flight for people willing to pay 5-10% more if that saves several hours.

@jakozaur thanks for your feedback. We are always working on new features, trip optimization is high on the list! ;).

Hm, so what's the difference between you and the other guys? Don't really get it. Any examples for additional savings I can get with you?

@cneumann81 thanks for trying tripdelta. A few examples that might explain what we do differently.

-We find hidden flights by combining rival airlines (https://tripdelta.com/results/SFO/LON/28-10-15)

-Our smart area search looks at every airport near your start and destination and checks every single combination to find better alternatives (https://tripdelta.com/results/FKB/BOH/01-10-15/ )

-We split your return flights into open jaw tickets (https://tripdelta.com/results/SFO/YVR/11-09-15/YVR/SFO/13-09... )

-Wanna do a short world trip? Better check this one out (https://tripdelta.com/results/SFO/BER/18-09-15/BER/DXB/25-09...)

I hope this explains what we do ;).. If you have any more questions, let me know

Interesting, but how does this translate into concrete savings for me? Any concrete examples?

Of course. Look at the San Francisco to Vancouver (return) example:

Regular: SFO - YVR: $385 Tripdelta: Oakland - Bellingham (two tickets): $164

You save $221 by picking two airports that are, by locals, considered good airports (I live in Vancouver, we all use Bellingham!)

The great thing is: Even when you're not a local, you just enter Vancouver, and we know about it!

Thanks for pulling this out. Even though I assume this is a best case example, savings in that are interesting for sure. Will check you guys out for my next booking!

Thanks for that. Actually, the example above does not showcase a unique flight combination but something like this happens rather often. I hope next time you check it out you'll face something similar.

Let us know about it ;)

Just tried it and it did return some pretty good prices. Care to elaborate how it does that? Machine Learning could mean anything or nothing.

@harryhunt711 Thanks for your feedback. To answer your questions. We use Apache Spark with a mixture of different algorithms to analyze over 1.5b flight prices and come up with millions of new routing possibilities. Meanwhile on every request our algorithm learns from past mistakes and corrects its decision based on new data

Hi I am Max.

I am one of the co-founders of tripdelta. If you have any questions, feel free to get in touch.

We are always happy about your feedback.

It's a bit frustrating to search for flights to a specific airport and then to only receive results for a different but somewhat nearby airport. For example: SFO -> ONT or SFO -> SNA will only return results for SFO -> LAX

@ktsmith thanks for you feedback. Maybe we should clarify that . We have filter for that on the top to show only your "original results". Those show you only what you where looking for and no further suggestions

Sorry for the delayed reply. I tried the filter before I commented here and my results were exactly the same.

Nice job! UI is very good!

Thank you! Glad you like it!

Reply on news.ycombinator.com