That's pretty interesting -- Be great if there was some kind of API where a site owner could pump in all the destinations (say in our case we have about 600 destinations across SEA) and it could be used as a localised tool for site readers ... obviously pointing the user to a Travelfish result rather than a Tripadvisor one
Yes, clever, but I believe this would be more useful for those people who use travel agents. Most of us on TF are die hard independents who enjoy the planning stage almost as much as the actual journey.
#3 daawgon has been a member since 17/4/2007. Posts: 1,091
I just had a look.... .so I picked two places in NZ (Tauranga, Wellington) and HCMC. Clicked on Recommend to get my recommendations - and it comes up with Mangere. Mangere? Of all of NZ it comes up with that? Seriously?! It's a lower socio-economic, working class area that doesn't offer much to travellers ... aside form the airport.
OK, so I tried it again. This time only selecting Tauranga as the place I visited. Recommendation was Noumea! Sorry but there really isn't ANY connection between Tauranga and Noumea. Why not at least other cities in NZ? It would be helpful to know what algorithm is being used...
Sorry, I just don't get it. I don't want to discourage you but I think some more serious work needs to be done to get it to provide meaningful info if you want it to be taken seriously. Also - the areas to select from (as places that I've visited) are seriously limited and appear to be quite random - so I don't feel that I can give an indication of 'recently visited places' that would even allow the algorithm to give me helpful info. Somtam's idea of having sites like TF download some info into it would be worth exploring.
You need to do some sensibility testing - do the results make sense?
Keep working on it - it may have potential. But my opinion is that you need to get it to a much better state before launching it.
My two cents worth
I chose Edmonton, Calgary and Montreal - and got San Jose, California. I see the similarities.
I chose Dallas and Phoenix - and got Honolulu, Hawaii...hit 'next' and got Henderson, USA. Huh?
I chose Anaheim, Los Angeles and San Francisco - and got San Diego, California...hit 'next' and got Los Angeles...hit 'next' and got Cripple Creek, USA. Where?
This is not inspiring. This is lame.
Better off wearing a blindfold and dropping a pencil on a map...or spinning a globe and stopping it with my finger.
Thanks everyone for constructive feedback. Let me explain the algorithm briefly. It is called collaborative filtering. As name suggests, recommendations are based on places you and other people visited. Based on the cities you chose it tries to find several people with similar tastes. Next it looks which cities they visited and you haven't. So if you get recommended Mangere, that means someone with similar experience visited it.
Obviously the quality depends on the number of users collaborating. I launched it a few weeks ago so there is not many users from NZ yet, but it will improve with every visit. That's the idea of machine learning.
I have many improvement ideas and one of them is select more adequate and attractive locations in each region. I will contact Travelfish admin to see if we can import your locations there.
Thank you for giving me suggesstion to where to go next time i'll remember your words always...................
#9 bhartirathor has been a member since 22/11/2013. Posts: 3