Announcement

Collapse
No announcement yet.

Second guessing destination airport -Why?

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Oblivian
    replied
    Example of a 'last known route' reused I mention which ends in what you guys see based on closest receiver/airport.

    It did do gry-chc earlier. But now it's a new leg but already populated in an assumed reverse direction.

    As the callsign = registration (like most small aircraft) the destination is moot unless left as ??? And then populated after the fact if filed route not available.

    Although it has been filed with at least one of the international route suppliers obviously. Headed to Chatham island off the East coast.

    Sent from my XT1092 using Tapatalk

    Leave a comment:


  • apgphoto
    replied
    Originally posted by Tyke's Aero blog View Post
    Dear Paul, please don't raise these minor issues! My analytics software says there's been a 16% improvement in data quality so you are clearly wrong!
    73.6% Of All Statistics Are Made Up

    Leave a comment:


  • Tyke's Aero blog
    replied
    Originally posted by Mike View Post
    But thanks we appreciate all data corrections!
    Crikey! Recognition and appreciation at last! Are you feeling alright Mike? Not bumped your head on something and feeling a bit off color, no?

    Leave a comment:


  • Mike
    replied
    Originally posted by Tyke's Aero blog View Post
    Dear Paul, please don't raise these minor issues! My analytics software says there's been a 16% improvement in data quality so you are clearly wrong!
    Once again, route improvements have absolutely nothing to do with aircraft data.

    Leave a comment:


  • Mike
    replied
    Originally posted by apgphoto View Post
    Trying to second guess GA flights is pointless they are always going to be wrong, the routes bot tries to give our EMS helicopters routes. Routes should be kept to real routes to try and avoid ridiculous suggested destinations. Fully agree its an impossible job with multiple callsigns for the same routes.
    As I wrote in my previous post https://forum.flightradar24.com/thre...ll=1#post95179 the matching computer don't know if a flight is GA, cargo, passenger or charter flight. It knows the callsign, nothing more. For you when seeing text 'EMS' or 'Ambulance' it's very easy to know it's one rescue mission. For the computer doing the matching it's just a callsign like any other, nothing more.

    Originally posted by apgphoto View Post
    Thread drift but the bot thats importing US data needs the sack, I just sampled four aircraft registered since July 1st three had old information and one just had the registration imported. When I updated the TBM-9** yesterday I noticed that only reg and hex were imported on those.
    No one needs to be sacked, it works just as expected. First step is to import Hex and Reg (that is why we don't want Brazilian, German och Chinese aircraft on US hex codes) and the next step is to populate correct data. Over 100k hex codes have been imported this week and of these over 36k have been populated so far.

    Leave a comment:


  • Mike
    replied
    Originally posted by Tyke's Aero blog View Post
    The improvements mostly come from people such as myself and other prolific members spending our time correcting the trash that is sent to you from other sources.
    96% is the number of correct routes. Route improvements have nothing to do with aircraft data improvements. But thanks we appreciate all data corrections!

    Leave a comment:


  • Tyke's Aero blog
    replied
    Originally posted by apgphoto View Post
    Thread drift but the bot thats importing US data needs the sack, I just sampled four aircraft registered since July 1st three had old information and one just had the registration imported. When I updated the TBM-9** yesterday I noticed that only reg and hex were imported on those.

    More thread drift the junk ICAO codes are still in the aircraft add drop down despite me posting them numerous times
    Dear Paul, please don't raise these minor issues! My analytics software says there's been a 16% improvement in data quality so you are clearly wrong!

    Leave a comment:


  • apgphoto
    replied
    Trying to second guess GA flights is pointless they are always going to be wrong, the routes bot tries to give our EMS helicopters routes. Routes should be kept to real routes to try and avoid ridiculous suggested destinations. Fully agree its an impossible job with multiple callsigns for the same routes.

    Thread drift but the bot thats importing US data needs the sack, I just sampled four aircraft registered since July 1st three had old information and one just had the registration imported. When I updated the TBM-9** yesterday I noticed that only reg and hex were imported on those.

    More thread drift the junk ICAO codes are still in the aircraft add drop down despite me posting them numerous times

    Leave a comment:


  • Tyke's Aero blog
    replied
    Originally posted by Mike View Post
    There are currently 16 people with access to the database and 11 of them have been logged in last week.
    Where were these 16 people when Maxi was on vacation for 2 weeks? Answer me that please.

    Leave a comment:


  • Tyke's Aero blog
    replied
    Originally posted by Mike View Post
    When I in the database thread posted that we are improving quality you call it trash. When I post that we have improved route quality from 80% to 96% you call it trash again and write we aiming for quantity over quality. 80->96% in 1 year is much better quallity if you ask me. But whatever we post you just call it trash. Grow up.
    The improvements mostly come from people such as myself and other prolific members spending our time correcting the trash that is sent to you from other sources.

    Leave a comment:


  • Mike
    replied
    Originally posted by Tyke's Aero blog View Post
    Would that "big team of developers" be roughly the same size as the "big team of database editors" ?
    There are currently 16 people with access to the database and 11 of them have been logged in last week. It doesn't really matter if there are 10-20-30 or 50 people as the database currently contains 1.100.452 hex codes and that is number that is just too big to handle manually. Just like routes we are working on solutions to automatically update database to both increase speed and quality. Just to give you some numbers last week the database editors have added/updated about 900 rows of data, at the same time our new robot has added/updated data for 36.000 rows. Just today the robot has added more than 15.000 new US registered aircraft. And yes I know, you will call it trash.

    Leave a comment:


  • Mike
    replied
    Originally posted by Tyke's Aero blog View Post
    A lot of those sources are providing you with complete trash though. The site seems to have gone in a different direction over this past few years where you now seem to aiming for quantity over quality, and it shows (and not in a good way).
    When I in the database thread posted that we are improving quality you call it trash. When I post that we have improved route quality from 80% to 96% you call it trash again and write we aiming for quantity over quality. 80->96% in 1 year is much better quallity if you ask me. But whatever we post you just call it trash. Grow up.

    Leave a comment:


  • Mike
    replied
    Originally posted by airnrail View Post
    I don't doubt 96%+ for RPT traffic ie larger airlines flying the same route day after day. But there is much more aviation going on than that. While it would be unreasonable to expect accurate data for private flights that often do not even lodge a flight plan, there are countless other commercial flights where plans are always lodged even for very short flights. Many of these, like the ones to Ballera that I quoted in an earlier post, operate regularly just not to a rigid timetable. They also oten have fixed fight numbers. The error I quoted with the destination shown as a Turkish airport is not recent - it has been evident for at least two years. I would have thought that such an error could have been corrected in that time. Please remember that there seems to be no way to notify these errors to FR24.
    The number is actuallt around 60-70% globally and this is where we started 5-6 years ago. Since then we have been working on improving the quality and we have now reached about 96-99%. As we track about 150.000 - 180.000 flights per day. Making it possible to change occasional flights wouldn't even make it noticable in the statistics, and it would only mess things up for the computer trying to learn the routes.

    Leave a comment:


  • Mike
    replied
    Matching routes is very, very, very complex. Maybe it sounds easy to match UAL100 to UA100, but there are thousands if not millions of exceptions which makes life hard for anyone trying to get the routes correct. For a person in most cases it can look obvious that a route is not correct but for a computer it's not as easy. For a human watching a screen and knowing knowing same basic facts about the aircraft or the flight or pattern or livery it's easy to know were this flight should be heading even if some of the data is missing or incorrect, but it's not as easy for a computer.

    One very basic problems is that the computer don't know if the flight is a passenger flight, a cargo flight, a charter flight, a test flight, a position flight, if a flight plan has been filed, if it's an IFR or VFR flight, etc. For a computer that only can see a callsign, and possibly can look upp that the aircraft is a Cessna 208, it doesn't help much in understanding what type of flight this is. By not knowing that makes it very hard to know which rules should be used to match the route.

    On top of that there are many other things that make matching complicated including alphanumeric callsign (RYR17GR, EZY9HJ and so one), airlines using "incorrect" callsigns for example KLM710 for flight KL235, or WZZ1234 for flight W62255 (who came up with this idea?), incorrect callsigns SBS123 instead of SAS123 (fat finger syndrome), usage of registration instead of assigned callsign (SZ-YAA instead of LNK8850), usage of same flight number for 3-4-5 or even 6 flights. Another growing problem is all the airlines doing flights for other airlines and mixing callsigns (one flight is LOT callsign, next flight is Nordica callsign), flights using IATA code instead of ICAO code as callsign (for example Australian doctors using AM-callsign). To make things even more funny, airlines in US airlines have started to use random flight numbers instead of reusing same number for same route.

    We have a team that is working on improving the quality of route matching and we are also working on getting as many data sources as well. We are tracking the progress and we can see that we are making improvments every month, but we will for sure never have 100% data quality.

    Leave a comment:


  • airnrail
    replied
    Originally posted by Mike View Post
    Yes, I know that you think that everything we do is trash, shit, bad and sucks. But we have a big team of developers working daily on improving data quality and we are today matching up to 96-99.5% of flights (depending on region), with correct route, compared to about 82-92% one year ago.
    I don't doubt 96%+ for RPT traffic ie larger airlines flying the same route day after day. But there is much more aviation going on than that. While it would be unreasonable to expect accurate data for private flights that often do not even lodge a flight plan, there are countless other commercial flights where plans are always lodged even for very short flights. Many of these, like the ones to Ballera that I quoted in an earlier post, operate regularly just not to a rigid timetable. They also oten have fixed fight numbers. The error I quoted with the destination shown as a Turkish airport is not recent - it has been evident for at least two years. I would have thought that such an error could have been corrected in that time. Please remember that there seems to be no way to notify these errors to FR24.

    Leave a comment:

Working...
X