#
Using Open Source Data for Landing Time Prediction with Machine Learning Methods^{ †}

^{1}

^{2}

^{*}

^{†}

## Abstract

**:**

## 1. Introduction

## 2. Data Acquisition

#### 2.1. ADS-B Data

`traffic`library [10]. The following data cleaning was accomplished before modeling.

- An index to indicate observations of the same trajectory is added according to the ICAO call sign and timestamp.
- Trajectories outside the circles of 45, 100, 150, 200, and 250 nautical miles around Zurich Airport are excluded (entry points of these five circles are shown in the right subplot of Figure 1).
- Aircraft passing by but not landing at Zurich Airport are excluded.
- Trajectories with an entry point too close to the runway are excluded as well.

#### 2.2. Weather Data

#### 2.3. Feature Extraction

#### 2.4. Descriptive Analysis

## 3. Landing Time Prediction

#### 3.1. Methods and Comparison

`ranger`[13] is applied to fit a random forest method. Gradient boosting machines use a collection of weak learners, modeled by decision trees, to construct a strong prediction model. GBM uses an iterative method to update the model. For each step, the tangent of the cost function is fit by a decision tree and updates the model towards optimal. Package

`xgboost`[14] is applied to fit such a model. A feedforward neural network lets information in the feature space flow through functions, or layers, to map to the target value [15]. We add a penalized parameter to avoid overfitting since the neural network is too flexible and prone to overfitting. We use

`TensorFlow`[16] to construct a feedforward neural network with four hidden layers and ${L}^{2}$ norm regularization. To fit hyperparameters like the number of variables in a random forest model and penalty parameters in NN and GBM, we randomly split all observations as the training, validation, and test sets. The test set consists of 20% of all the observations, while the training and validation sets amount to the remaining 80%. The ratios of the training and validation sets are different for the three methods, 4:1 for RF and NN and as a hyperparameter for GBM. Hyperparameters are searched on a grid by the validation set, and the methods are evaluated on the test set. The methods are evaluated using the Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE):

#### 3.2. Error with Radius

#### 3.3. Feature Importance

## 4. Summary

## Funding

## References

- Kern, C.S.; de Medeiros, I.P.; Yoneyama, T. Data-driven aircraft estimated time of arrival prediction. In Proceedings of the 2015 Annual IEEE Systems Conference (SysCon) Proceedings, Vancouver, BC, Canada, 13–16 April 2015; pp. 727–733. [Google Scholar]
- Ayhan, S.; Costas, P.; Samet, H. Predicting estimated time of arrival for commercial flights. In Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, London, UK, 19–23 August 2018; pp. 33–42. [Google Scholar]
- Takacs, G. Predicting flight arrival times with a multistage model. In Proceedings of the 2014 IEEE International Conference on Big Data, Washington, DC, USA, 27–30 October 2014; pp. 78–84. [Google Scholar]
- Schultz, M.; Lorenz, S.; Schmitz, R.; Delgado, L. Weather impact on airport performance. Aerospace
**2018**, 5, 109. [Google Scholar] [CrossRef] - Dhief, I.; Wang, Z.; Liang, M.; Alam, S.; Schultz, M.; Delahaye, D. Predicting Aircraft Landing Time in Extended-TMA using Machine Learning Methods. In Proceedings of the ICRAT 2020 Conference, Tampa, USA, 15 September 2020. [Google Scholar]
- Le, T.H.; Tran, P.N.; Pham, D.T.; Schultz, M.; Alam, S. Short-Term Trajectory Prediction Using Generative Machine Learning Methods. In Proceedings of the ICRAT 2020 Conference, Tampa, USA, 15 September 2020. [Google Scholar]
- Hastie, T.; Tibshirani, R.; Friedman, J. The Elements of Statistical Learning: Data Mining, Inference, and Prediction; Springer Science+Business Media: New York, USA, 2013. [Google Scholar]
- Ayhan, S.; Samet, H. Aircraft trajectory prediction made easy with predictive analytics. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, San Francisco, CA, USA, 13–17 August 2016; pp. 21–30. [Google Scholar]
- De Leege, A.; van Paassen, M.; Mulder, M. A machine learning approach to trajectory prediction. In Proceedings of the AIAA Guidance, Navigation, and Control (GNC) Conference, Boston, MA, USA, 19–22 August 2013; p. 4782. [Google Scholar]
- Olive, X. Traffic, a toolbox for processing and analysing air traffic data. J. Open Sour. Softw.
**2019**, 4, 1518. [Google Scholar] [CrossRef] - Wang, Z.; Liang, M.; Delahaye, D. Automated data-driven prediction on aircraft Estimated Time of Arrival. J. Air Trans. Manag.
**2020**, 88, 101840. [Google Scholar] [CrossRef] - Hofert, M.; Kojadinovic, I.; Mächler, M.; Yan, J. Elements of Copula Modeling with R; Springer Nature: Cham, Switzerland, 2019. [Google Scholar]
- Wright, M.N.; Ziegler, A. ranger: A Fast Implementation of Random Forests for High Dimensional Data in C++ and R. J. Stat. Softw.
**2017**, 77. [Google Scholar] [CrossRef] - Chen, T.; Guestrin, C. XGBoost: A Scalable Tree Boosting System. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, San Francisco, CA, USA, 13–17 August 2016; pp. 785–794. [Google Scholar] [CrossRef]
- Goodfellow, I.; Bengio, Y.; Courville, A.; Bengio, Y. Deep Learning; MIT Press: Cambridge, MA, USA, 2016. [Google Scholar]
- Abadi, M.; Agarwal, A.; Barham, P.; Brevdo, E.; Chen, Z.; Citro, C.; Corrado, G.S.; Davis, A.; Dean, J.; Devin, M.; et al. TensorFlow: Large-Scale Machine Learning on Heterogeneous Systems. 2015. Available online: tensorflow.org (accessed on 8 October 2020).
- Grömping, U. Variable importance assessment in regression: Linear regression versus random forest. Am. Stat.
**2009**, 63, 308–319. [Google Scholar] [CrossRef] - Strobl, C.; Boulesteix, A.L.; Kneib, T.; Augustin, T.; Zeileis, A. Conditional variable importance for random forests. BMC Bioinform.
**2008**, 9, 307. [Google Scholar] [CrossRef] [PubMed]

**Figure 1.**(

**Left**) Sixty sample trajectories approaching Zurich Airport between 5 October 2019 and 31 October 2019, used for landing time prediction in this study. (

**Right**) Entry points of aircraft when 45, 100, 150, 200, and 250 nautical miles away from the runways. The red point in the middle denotes the reference point 47.45 latitude and 8.56 longitude of Zurich Airport.

**Figure 2.**The difference in the track angle between the entering point on the circle and the landing point. (

**Left**) Trajectories with less landing time; (

**right**) trajectories with more landing time.

**Figure 3.**K-means clustering and landing time by clustering. (

**Left**) Clustered entry points with the radius as 45 NM; (

**right**) density plot of landing time grouped by K-means clusters.

**Figure 8.**Density plot of errors on the test set depending on the radius by the RF, GBM, and NN methods.

**Table 1.**Summary of features, relevant for the aircraft landing time prediction. GFS, Global Forecast System.

Category | Feature | Description |
---|---|---|

Trajectory | latitude and longitude | position of aircraft when entering the circle |

geopotential altitude | position of aircraft when entering the circle | |

ground speed | ground speed of aircraft when entering the circle | |

track | track of aircraft when entering the circle | |

vertical rate | vertical rate of aircraft when entering the circle | |

cos.angle | cosine of the angle between tracks entering the circle and landing | |

Cluster | cluster | K-means cluster of aircraft when entering the circle |

Traffic density | density 15 min before | number of aircraft departing or landing 15 min before entering |

density 15 min after | number of aircraft departing or landing 15 min after entering | |

density 15 min BA | number of aircraft departing or landing 15 min Before/After entering | |

density 30 min before | number of aircraft departing or landing 30 min before entering | |

density 30 min after | number of aircraft departing or landing 30 min after entering | |

density 30 min BA | number of aircraft departing or landing 30 min before/after entering | |

density 60 min before | number of aircraft departing or landing 60 min before entering | |

density 60 min after | number of aircraft departing or landing 60 min after entering | |

density 60 min BA | number of aircraft departing or landing 60 min before/after entering | |

Weather | temperature, wind speed | GFS analysis data at (longitude = 8, latitude = 47, pressure = 1000 hPa) |

and relative humidity | and time closest to the time when aircraft entering the circle | |

Seasonality | hour of a day | 0–24 h of the day |

parts of a week | labeled as Monday to Sunday |

**Table 2.**Accuracy evaluation results (RMSE and MAE) of three machine learning methods in order to predict the aircraft landing time. GBM, Gradient Boosting Machine.

Radius (NM) | Methods | RMSE (min) | MAE (min) |
---|---|---|---|

45 | RF | 3.21 | 2.39 |

45 | GBM | 3.16 | 2.42 |

45 | NN | 4.18 | 3.22 |

100 | RF | 3.59 | 2.70 |

100 | GBM | 3.44 | 2.60 |

100 | NN | 5.01 | 3.72 |

150 | RF | 4.16 | 2.92 |

150 | GBM | 4.04 | 2.88 |

150 | NN | 5.65 | 4.00 |

200 | RF | 4.53 | 3.22 |

200 | GBM | 4.27 | 3.12 |

200 | NN | 7.02 | 4.99 |

250 | RF | 4.78 | 3.32 |

250 | GBM | 4.75 | 3.41 |

250 | NN | 10.64 | 8.61 |

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Chen, G.; Rosenow, J.; Schultz, M.; Okhrin, O.
Using Open Source Data for Landing Time Prediction with Machine Learning Methods. *Proceedings* **2020**, *59*, 5.
https://doi.org/10.3390/proceedings2020059005

**AMA Style**

Chen G, Rosenow J, Schultz M, Okhrin O.
Using Open Source Data for Landing Time Prediction with Machine Learning Methods. *Proceedings*. 2020; 59(1):5.
https://doi.org/10.3390/proceedings2020059005

**Chicago/Turabian Style**

Chen, Gong, Judith Rosenow, Michael Schultz, and Ostap Okhrin.
2020. "Using Open Source Data for Landing Time Prediction with Machine Learning Methods" *Proceedings* 59, no. 1: 5.
https://doi.org/10.3390/proceedings2020059005