Wyniki wyszukiwania - BazTech

Ograniczanie wyników

Znaleziono wyników: 2

Liczba wyników na stronie

Wyniki wyszukiwania

Wyszukiwano:
w słowach kluczowych: algorytmy komputerowe

Sortuj według:

Ogranicz wyniki do:

Flight delay prediction based with machine learning

Hatıpoğlu Irmak, Tosun Ömür, Tosun Nedret

LogForum

2022

Vol. 18, no. 1

97--107

Background: The delay of a planned flight causes many undesirable situations such as cost, customer satisfaction, environmental pollution. There is only one way to prevent these problems before they occur, and that is to know which flights will be delayed. The aim of this study is to predict delayed flights. For this, the use of machine learning techniques, which have become widespread with the development of computer capacities and data storage systems, is preferred. Methods: Estimations are made with three up-to-date techniques XGBoost, LightGBM, and CatBoost techniques based on Gradient Boosting from machine learning techniques. The bayesian technique is used for hyper-parameter settings. In addition, the Synthetic Minority Over-Sampling Technique (SMOTE) technique is also used, as the majority of flights are on time and delayed flights, which constitute a minority class, may adversely affect the results. The results are analyzed and shared with and without SMOTE. Results: As a consequence of the application, which was run on a data set containing all of an international airline's flights [18148 flights] for a year, it was discovered that flights may be predicted with high accuracy. Conclusions: The application of machine learning techniques to anticipate flight delays is new, but it has a lot of potential. Companies will be able to avert problems before they develop if delays are correctly estimated, which can generate plenty of issues. As a result, concrete advantages such as lower costs and higher customer satisfaction will emerge. Improvements will be made at the most vulnerable place in the aviation business.

Zastosowanie algorytmu mrówkowego do wyznaczania maksymalnej grupy wzajemnie połączonych elementów

Schiff K.

Czasopismo Techniczne. Elektrotechnika

2008

R. 105, z. 1-E

63-78

W artykule zaprezentowano algorytmy mrówkowe wyznaczające największą klikę w grafie, za pomocą której modeluje się problem wyznaczania największego ze skupień wzajemnie połączonych elementów elektronicznych na płytce drukowanej w celu minimalizacji długości połączeń między nimi, a w konsekwencji minimalizacji ilości materiału zużytego na ich wytworzenie. W artykule zaprezentowano algorytm oparty na odmiennych aspektach zachowania się mrówek w porównaniu z dotychczas opracowanymi algorytmami. Główną różnicą między algorytmami jest faza eksploracji, która została wprowadzona w prezentowanym algorytmie. Opracowany algorytm porównano z Algorytmem 457 pod względem wyznaczanego wymiaru klik. Dokonano również porównania procedur lokalnego przeszukiwania (2,1)-wymiany i procedury opartej na metaheurystyce kolonii mrówek.

In this paper an ANT algorithm, which is used to find a maximum group of mutually connected electronic elements in order to minimize the total length of connections, is presented. The new algorithm differs from algorithms which have been presented in scientific papers until now. The main difference is a phase of ANT exploration which is absent in other ANT algorithms. Sizes of maximum clique indicated by ANT algorithm and the Algorithm 457 are compared. The influence of the local search was presented also and the (2,1)-exchange local procedure and the ANT procedure of local search was compared.