{ "fetched_at": "2026-10-18T13:41:59Z", "source_url": "https://www.kaggle.com/competitions/flight/data", "pages": [ { "name": "Description", "content": "\n\n\n\n

Flight Quest Phase 1 Winners

\n\n
    \n
  1. \n \n\n
    \n Team Gxav &* used a mixture of gradient boosting and random forest models to predict gate and runway arrival times. With average errors of 4.2 and 3.2 minutes for gate and runway arrivals, respectively, this translates to 40% and 45% improvements over the standard industry benchmark estimates. Key to their success was careful feature selection with their final models using only 58 and 84 features for gate and runway arrivals, respectively, from the total 258 features they painstakingly constructed and optimized.\n
    \n
    2. \n\n
  2. \n \n\n
    \n Team As High As Honor used a two-step approach that combined the results of a generalized linear model that encoded intuition about important variables with refinements derived from a random forest model. The team capitalized on the success of the linear model to add the effects of multiple variables and cleanly resolve issues of missing data.\n
    \n
    3. \n\n
  3. \n \n \"\"\n \n \n

    3. Gabor Takacs

    \n     \n
    \n Team Taki used a six layer model relying on successive ridge regressions and gradient boosting\nmachines to model both gate and runway arrival times. This approach used 56 features extracted from the raw data, with all but two coming from the test day data.\n\n
    \n
    4. \n\n
  4. \n \n \"\"\n \n \n

    4. Sergey Kozub

    \n     \n
    \n Team Sun’s approach to predicting gate and runway arrival times relied on creating a derived data set with new variables encoding information about the aircraft, airport, airway, gate, hour, and flight path times. Important features used in this model include aircraft GPS position, ASDI flight plans the direction from which airplanes approached airport runways.\n
    \n
    5. \n
  5. \n \n \"\"\n \n \n

    5. Jacques Kvam

    \n     \n
    \n Jacques Kvam’s approach for predicting runway and gate arrival times used gradient boosting for a model using 10,000 trees and a whopping 1,102 features trained on 260,000 flights. Most significant among these included the distance between the final waypoint and the arrival airport. Many weather features were important as well including temporary vertical visibility and wind speed at the arrival airport.\n
    \n
    6. \n
  6. \n \n \"\"\n \n \n

    Honorable Mention: Matt Berseth

    \n \n
    \n Team __mtb__ used random forest and gradient boosted models to estimate runway and gate arrival times. The final solution included over 100 different individual models, each focused on a narrow set of features (i.e. wind/weather, flight plan, aircraft's current location, etc.). These individual models were blended together to generate the final estimates. The training data was created by randomly selecting eight cutoff times for each day in the training period. A separate cross validation data set was used to select hyper-parameters.\n
    \n
    7. \n\n
\n\n

\n\n\n\n

Think you can change the future of flight?

\n

Did you know airlines are constantly looking for ways to make flights more efficient? From gate conflicts to operational challenges to air traffic management, the dynamics of a flight can change quickly and lead to costly delays.

\n

There is good news. Advancements in real-time big data analysis are changing the course of flight as we know it. Imagine if the pilot could augment their decision-making process with “real time business intelligence,”—information available in the cockpit that would allow them to make adjustments to their flight patterns. 

\n

Your challenge, should you decide to accept it: 

\n

Use the different data sets found on this page under Get the Data to develop a usable and scalable algorithm that delivers a real-time flight profile to the pilot, helping them make flights more efficient and reliably on time.

\n
Tweet\n\n\n
\n\n

\"\"

", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Prizes", "content": "

Prizes will be awarded based on performance on the final evaluation data set as follows:

1st Prize $100,000
2nd Prize $50,000
3rd Prize $40,000
4th Prize $30,000
5th Prize $20,000

\r\n

LSU Prize $10,000

\r\n

Total $250,000

\r\n

LSU Prize
The Lean Startup conference prize will be awarded to the best performing submission by a team including a member registered with the LSU.  See the complete Flight Quest Rules for more details.  To register your team as eligible, send an email to LSUprize@kaggle.com before the submission deadline on March 11, 2013.  The email should have subject line \"LSU Prize\" and it should include the relevant member's full name, display name and the email address the member used when registering for the conference.

\r\n


", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Evaluation", "content": "

You are predicting both the runway arrival time and the gate arrival time, so you have two predictions for each airplane.

\r\n

Your predictions will be evaluated based on the root mean squared error (RMSE) between the predicteds and the truth:

\r\n

$$ \\textrm{RMSE} = \\sqrt{\\frac{1}{n} \\sum_{i=1}^{n} (y_i - \\hat{y}_i)^2}$$

\r\n

We will separately compute the RMSE for gate arrival times and runway arrival times, and use a weighted average of the two:

\r\n

$$ \\textrm{RMSE_Final} = (0.75 \\times \\textrm{RMSE_Gate}) + (0.25 \\times \\textrm{RMSE_Runway}) $$

\r\n
 
", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "rules", "content": "

Rules

\r\n

The complete Official GE Flight Quest Rules are available in PDF form.

\r\n

Click here to view or download the complete Official GE Flight Quest Rules.

\r\n

Contest ends at 11:59:59 p.m. UTC on March 13, 2013.  Entrants must submit at least one entry and final algorithm codebase before 11:59:59 p.m. UTC on February 14, 2013, to be eligible.  Must be 18 years of age or older to participate.  Free GEQuest website account or Kaggle account required.  No purchase necessary.  Void where prohibited.

", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Submission Instructions", "content": "

Arrival time predictions for all flights are submitted to Kaggle in a single comma-separated value (CSV) file. The submission file contains 3 columns:

\r\n\r\n

Sample submission files (such as estimated_arrival_benchmark.csv) for the public leaderboard set have been uploaded to the data page.  A sample submission for the final evaluation set will be released with that set.

\r\n

During the time the public leaderboard is open, you may submit predictions on the public leaderboard set. In order to be eligible for prizes, you must make at least one submission on the public leaderboard set.

\r\n

Model Submission

\r\n

When you submit an Entry, you also will be able to upload your model to the Website. Your final model must contain all code and parameter settings necessary to evaluate your models on new data, and include a README file with instructions on how to do so. The purpose of this is to ensure this Competition is fair and no answers were looked up for the Final Evaluation Set period. This model submission is required to be eligible to win a prize. This final model will be used solely for the purpose of verifying eligibility for prizes. You may upload either the archive itself or a SHA-256 hash of the archive.

\r\n

Kaggle will check the final submissions from the Entrants at the top of the Final Evaluation Leaderboard to make sure that Entrant’s final submission was created using that Entrant’s final model submission. If a potential winner did not use the final model codebase to create the final submission Entry, that potential winner will be disqualified and Kaggle will evaluate the final submission of the next Entrant on the Leaderboard as a potential winner.

\r\n

For an entry to be eligible to win, any open source or third party code or other materials used in connection with that Entry must be limited to code that GE will be entitled to commercially use and exploit, and any open-source-related or third party licensor requirements and restrictions must be disclosed as part of the model submission.

\r\n

Your model must be structured so that it makes each test day's predictions based on no information in the final evaluation test data other than the information from that day, which will be in an appropriately named folder.

", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Timeline", "content": "

November 29, 2012
Competition Launch; Initial Training Set Released

\r\n

Entrants begin analyzing Initial Training Set.

\r\n

December 18, 2012
Public Leaderboard Set Released; Leaderboard Activated

\r\n

Entrants submit entries developed using the Initial Training Set.  Entries are scored against the Public Leaderboard Test Set.

\r\n

January 11, 2013
Augmented Training Set, Part 1 Released

\r\n

Entrants further develop entries using the Initial Training Set and the Augmented Training Set, Part 1.  Entries continue to be scored against the Public Leaderboard Test Set.

\r\n

February 8, 2013
Public Leaderboard frozen. Participating teams must have made at least one submission to the public leaderboard by this date.

\r\n

February 9, 2013
Augmented Training Set, Part 2 Released
Public Leaderboard test data (with ground truth arrival times) is released

\r\n

Entrants further develop models using all available data; Entrants choose their best model and submit their model (see Submission Instructions).

\r\n

February 14, 2013
Model Submission Deadline (must include code or hash of code)

\r\n

March 4, 2013
Final Evaluation Set Released

\r\n

Entrants apply their previously submitted final model to the Final Evaluation Set to generate and submit their final predictions.

\r\n

March 11, 2013
Final Predictions Submission Deadline

", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "data-description", "content": "

 The Flight Quest data is temporally split into five datasets that will be released over the course of the competition. The names and dates of the flight data contained in each dataset are as follows:

\r\n\r\n

For each day in the test period (first for the public leaderboard, and later for the final evaluation), we will select a random time (uniformly chosen between 9am EST and 9pm EST) and select all of the flights in the air at that cutoff time. You will be provided with relevant data for each day that would be available at the chosen cutoff time.

\r\n

Your model must be structured so that it makes each test day's final test data set predictions based on no information in the final evaluation test data other than the information from that day, which will be in an appropriately named folder. (Reworded for clarification on 11/30/2012. See forum for explanation.

\r\n

The relevant data is divided into folders by day. For these purposes, a \"day\" includes all of the time from 1amPST/4amEST/9amUTC on that day until the same time on the next day. So Nov. 11, 2012 is considered to be from 9am UTC on Nov. 11, 2012 until 9am UTC on Nov. 12, 2012.

For a fuller description of the data, visit the Flight Quest Data wiki (information on the wiki is solely for background information and does not form part of the official documentation for Flight Quest).

Data provided by FlightStats, Inc., and other sources.

", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Competition.sidebar.bottom", "content": "
\r\n
\r\n

Deadlines

\r\nview timeline »
\r\n\r\n
\r\n
\r\n
\r\n

Prize Pool

\r\nview prizes »
\r\n

1st $100,000
2nd $50,000
3rd $40,000
4th $30,000
5th $20,000 

\r\n
LSU Prize ?
\r\n

$10,000

\r\n
", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Competition.title.after", "content": "
\"\"
\r\n
\r\n

Flight Quest

\r\n

Make flying more efficient

\r\n
\r\n
\"\" in partnership with \"\"
\r\n", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Faq", "content": "

What are the Industrial Internet Quests?

\r\n

The Industrial Internet Quests are a challenge to developers, data scientists and designers to create the best new algorithms and design of applications to reduce air travel delays and deliver the ultimate patient experience. The Quests will span several months and submissions will be reviewed and winners selected by Healthcare experts for the Hospital Quest and judged based on automated rules for the Flight Quest. Winners will receive a total prize pool of $600,000 in cash rewards as well as the opportunity to have their submission turned into a working simulation, with the ultimate goal of creating an onboard flight management application and hospital patient experience application, supported by GE.

\r\n

Proposals will be submitted online via www.gequest.com

\r\n

The Flight and Hospital quests, launched in collaboration with Alaska Airlines and Ochsner Health System, are part of an ongoing series of Industrial Internet challenges. GE is partnering with Kaggle, a leader in big data challenges, to execute the Quests. 

\r\n

 

\r\n

Why is GE issuing these Industrial Internet Quests?

\r\n

GE believes that the Industrial Internet represents the next frontier of systemic technical advances and innovations that enable breakthrough levels of efficiency and productivity. By creating an open platform and engaging with the developer community and data scientists to develop Industrial Internet expertise, we will be able to eliminate waste and increase efficiency to invent the next generation of industrial services.

\r\n

 

\r\n

The Quests:

\r\n

 

\r\n

Who is eligible to enter?

\r\n

Entries are open to anyone 18 years old or older, and all legally formed entities, except a bankrupt person or organization or a criminal offender under any national penal system.  The Quests are global.

\r\n

To enter, visit gequest.com from November 29, 2012.

\r\n

 

\r\n

Are GE employees eligible to enter?

\r\n

Officers, directors, employees and advisory board members (and their immediate families and members of the same household) of GE, Kaggle Inc. and their respective affiliates, subsidiaries, contractors, agents, judges and advertising and promotion agencies (collectively, the \"Competition Entities\") also are not eligible to participate in the Competition


\r\n

 

\r\n

Are there any age parameters?

\r\n

Entries are open to anyone 18 years old or older.

\r\n

 

\r\n

What makes GE qualified to issue the Quests?

\r\n

At GE, we’re working on exciting and promising Industrial Internet technologies to help create more efficient air travel and a frictionless hospital experience. As part of this, we are looking to engage with developers and data scientists to create an open platform, cultivate a community and develop a widespread expertise in the Industrial Internet.  GE has partnered with Alaska Airlines and Ochsner Health System to help develop new technologies. 

\r\n

 

\r\n

Does GE plan to announce additional Industrial Internet Quests in the future?

\r\n

GE plans to launch a series of “Industrial Internet Quests” in 2013. These quests would embrace the goals of open collaboration with our customers, developers, data scientists and designers to help GE in transforming various industries across Healthcare, Aviation, Rail, Oil & Gas and Power.

\r\n


Entry and judging process:

\r\n


What is the entry and judging process for Flight Quest?

\r\n

The Flight Quest will take place over two phases. In the first phase, participants will be asked to create a predictive model that increases aviation efficiency. The first phase of the Flight Quest begins on November 29, 2012 and submissions will be taken for two months. A third month will be used to further analyze the algorithms and identify the winning predictive model.

\r\n

For the second phase of the Quest, entrants will use the winning predictive algorithm to create a simulation, with the ultimate goal being the creation of an onboard real-time flight management system.

\r\n

The Flight Quest will use Kaggle’s live competition platform to submit and analyze predictive models.

\r\n

 

\r\n

What information will GE provide for the Flight Quest?

\r\n

Entrants will receive an initial two week data set of flight information sourced from the National Airspace System. A further two months of data will be released in stages during the Quest.  The information will include: flight number, origin, destination, take-off time, arrival time, latitude and longitude at frequent interim waypoints along the journey, and weather and wind data. Based on this information, entrants are challenged to design an algorithm for a flight management system that increases predictive efficiency. Data will be provided by FlightStats, Inc. and others.

\r\n

 

\r\n

Can an individual submit more than one entry?

\r\n

Yes, multiple entries are permitted. Each participant will select their best entry for final evaluation.

\r\n

 

\r\n

Can individuals collaborate to submit a single entry?

\r\n

Yes, multiple individuals or entities may collaborate to submit a single entry.  In the Flight Quest, all team members may submit entries on behalf of the team.

\r\n

 

\r\n

Is there a contact name or number for people with questions?
All of the information for the Industrial Internet Quests will be available at gequests.com. The specific terms for the Flight and Hospital Quests can be found at www.gequest.com/flight and www.gequest.com/hospital respectively.

\r\n

 

\r\n

Are you working with any outside partners, or firms to manage the entry process?

\r\n

To execute the Flight Quest, GE has partnered with Kaggle, the leader in big data challenges and analysis.

\r\n

 

\r\n

Can entrants enter other similar challenges?
Yes.

\r\n

 

\r\n

Who will comprise the judging panel?

\r\n

The Flight Quest will be scored by Kaggle’s live leaderboard.

\r\n

 

\r\n

How much money will be awarded?

\r\n

The combined winners from two phases of Flight Quest will receive a total prize pool of $500,000 from GE.

\r\n

For the Hospital Quest, the winners will receive a total prize pool of $100,000 from GE. 

\r\n

 

\r\n

Are there specific countries you are targeting?

\r\n

This is a global challenge and we encourage and welcome entries from everyone who meets the eligibility requirements.

\r\n

 

\r\n

Where can I find more answers?

\r\n

Each competition will have Forums for live discussion and addressing questions that arise during the course of the Quests.  These Forums can be viewed at www.gequest.com/flight and www.gequest.com/hospital.

", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Ge Product", "content": "

Fuel & Carbon Solutions - GE Aviation

\r\n

GE is leading the next wave of productivity revolution with its \"industrial internet\" solutions that will transform global industrial. One such solution is GE's Fuel & Carbon Solutions.

\r\n

Through a comprehensive, collaborative approach, GE Fuel & Carbon Solutions goes beyond the powerplant to provide airlines with insight into their operations, policies and processes. Our experts will work with our customers team to identify and select fuel savings projects, implementation and ongoing support to ensure fuel savings are sustainable.

\r\n

GE Aviation's Fuel & Carbon Solutions uses proprietary algorithms to evaluate airline operations data and identify opportunities that could reduce fuel consumption by more than 2 percent. If an average-sized airline (70 aircraft, $1 billion annual fuel costs) implemented Fuel & Carbon Solutions to achieve 2 percent improvement in fuel consumption, their annual fuel expenditure would decrease by $20 million and reduce 60 metric tons of CO2 emissions - which is equivalent to removing more than 10,000 passenger vehicles from the road or planting more than 1.3 million trees.

\r\n

http://www.geaviation.com/services/optimization/fuel-and-carbon-solutions.html

\r\n

\"\" 

", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Winners", "content": "\r\n\r\n

Flight Quest Phase 1 Winners

\r\n
    \r\n
  1. \r\n

    1st Prize: Team Gxav &*

    \r\n

    Xavier Conort, Cao Hong, Clifton Phua, Ghim-Eng Yap and Kenny Chua (Singapore)

    \r\n

    Team Gxav &* used a mixture of gradient boosting and random forest models to predict gate and runway arrival times. With average errors of 4.2 and 3.2 minutes for gate and runway arrivals, respectively, this translates to 40% and 45% improvements over the standard industry benchmark estimates. Key to their success was careful feature selection with their final models using only 58 and 84 features for gate and runway arrivals, respectively, from the total 258 features they painstakingly constructed and optimized. Q&A with Xavier

    \r\n\r\n
    2. \r\n
  2. \r\n

    2nd Prize: Team As High As Honor

    \r\n

    Jonathan Peters (Preston, UK) and Paweł Jankiewicz (Warsaw, Poland)

    \r\n

    Team As High As Honor used a two-step approach that combined the results of a generalized linear model that encoded intuition about important variables with refinements derived from a random forest model. The team capitalized on the success of the linear model to add the effects of multiple variables and cleanly resolve issues of missing data.

    \r\n\r\n
    3. \r\n
  3. \r\n

    3rd Prize: Team Taki

    \r\nGabor Takacs(Gyor, Hungary)\r\n

    Team Taki used a six layer model relying on successive ridge regressions and gradient boosting machines to model both gate and runway arrival times. This approach used 56 features extracted from the raw data, with all but two coming from the test day data.

    \r\n\r\n
    4. \r\n
  4. \r\n

    4th Prize: Team Sun

    \r\n

    Sergey Kozub (Kursk, Russia)

    \r\n

    Team Sun’s approach to predicting gate and runway arrival times relied on creating a derived data set with new variables encoding information about the aircraft, airport, airway, gate, hour, and flight path times. Important features used in this model include aircraft GPS position, ASDI flight plans and the direction from which airplanes approached airport runways.

    \r\n\r\n
    5. \r\n
  5. \r\n

    5th Prize: Team Jacques Kvam

    \r\n

    Jacques Kvam (Livermore, CA, USA)

    \r\n

    Jacques Kvam’s approach for predicting runway and gate arrival times used gradient boosting for a model using 10,000 trees and a whopping 1,102 features trained on 260,000 flights. Most significant among these included the distance between the final waypoint and the arrival airport. Many weather features were important as well including temporary vertical visibility and wind speed at the arrival airport.

    \r\n\r\n\r\n
  6. \r\n

    Honorable Mention: Team __mtb__

    \r\n

    Matt Berseth (Jacksonville, FL, USA)

    \r\n

    Team __mtb__ used random forest and gradient boosted models to estimate runway and gate arrival times. The final solution included over 100 different individual models, each focused on a narrow set of features (i.e. wind/weather, flight plan, aircraft's current location, etc.). These individual models were blended together to generate the final estimates. The training data was created by randomly selecting eight cutoff times for each day in the training period. A separate cross validation data set was used to select hyper-parameters.

    \r\n\r\n
    7. \r\n
", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false } ] }