MISiS: Application for forecasting flight delays

The main articles are:

• Mobile apps
• Neural networks (neural networks)
• Data mining Data mining

2023: Creating an App

Vyacheslav Pachkov, a student at MISIS University, created a smartphone application that predicts flight delays as a graduation qualification work. Representatives of NUST MISIS reported this to TAdviser on August 18, 2023.

Russia has created an application for predicting flight delays

The results of development testing showed that the model trained on big data made forecasts that differed from the true values ​ ​ of delays by 12 minutes, which, according to university representatives, indicates high accuracy and wide prospects for the development of this topic.

As you know, departure delays are a major problem for both airlines and passengers, as they result in significant time and money losses. Major factors that contribute to airline delays include weather conditions, congestion, type and age of air traffic, and aircraft maintenance issues.

Modern neural networks can help people insure themselves in the event of a flight transfer or cancellation and plan their next actions. According to NITU MISIS, various algorithmachine learning, such as a multilayer perceptron model, Bayesian modeling, decision tree, cluster classification or random forest, are suitable for predicting delays. They can assess both the likelihood and severity of flight delays, which may prove invaluable in developing better airline planning and maintenance strategies.

"Machine learning methods do an excellent job in the field of air transportation. Determining the most important and informative features is a key step in developing an effective model for predicting flight delays. It seems to me that if airlines or airports invest in this area to improve the accuracy and reliability of predictions, this will demonstrate concern for customers, increase the positive reputation of a responsible carrier and, of course, allow you to choose the most effective risk management strategy, "Vyacheslav Pachkov is convinced.

According to him, the development is based on an artificial neural network capable of processing complex input data and conducting nonlinear classification or regression, it is called the Multilayer Perceptron (MLP) model. MLP can simulate more complex functions and dependencies between input and output. The input layer receives the feature vector, the hidden layers process the data, and the output layer generates predictions. Neurons between layers are connected by scales that determine the degree of influence of each neuron on other neurons. The learning process continues until a certain stop criterion is reached, such as the minimum value of the loss function or the error stabilization on the validation set.

The model uses 9 input features: the time between arrival and departure from the departure airport; expected arrival time at the destination airport; range of flight; departure airport; arrival airport; aircraft type; temperature; probability of precipitation; time of year. As specified in NITU MISIS, a total of about a million records were collected, representing flight information for the last year from API the interfaces of flight tracking services FlightAware and FlightStats from, Russia,, Canada,, Great Britain, and France, Germany Australia Japan USA providing a significant amount of data and geographical diversity. WeatherAPI, OpenWeatherMap and WeatherUnderground were the sources for the collection of meteorological data.

This datacet was used as the basis for training and subsequent testing of the machine learning model. After measuring performance on a test dataset, the converted model was integrated into an iOS application to demonstrate how to work on real data. The application uses the developed model to perform predictions on a mobile device.

With further training at NUST MISIS, Vyacheslav Pachkov plans to refine the application, increasing the accuracy of the model and optimizing the layers of the neural network to speed up work on weak mobile devices.

See also

• Passenger air travel