DEMAND FORECASTING FOR BIKE SHARING SYSTEMS BY MACHINE LEARNING: A CASE STUDY IN KONYA

Gözde Can Atasagun; Ece Çetin Yağmur

Article Details

DEMAND FORECASTING FOR BIKE SHARING SYSTEMS BY MACHINE LEARNING: A CASE STUDY IN KONYA

Authors:

Gözde Can Atasagun | Ece Çetin Yağmur

Publisher:

Eskişehir Osmangazi University

Main Topics:

Endüstri Mühendisliği

Journal:

Eskişehir Osmangazi Üniversitesi Mühendislik ve Mimarlık Fakültesi Dergisi Vol.32 Issue 3, pp.1470-1484

Document Type:

Article

Publication Date:

12/2024

Document Language:

Turkish

Document ID (Org No.):

2001333098

pISSN:

2630-5712

Keywords:

Abstract:

 As the concept of sustainability increases day by day, eco-friendly techniques are researched instead of traditional transportation vehicles. One of these innovative solutions is the bike-sharing system, which provides carbon-emission-free transportation. The system greatly reduces the carbon footprint and also contributes to environmental sustainability by preventing urban traffic congestion. Bike-sharing systems, which are frequently preferred because they are economically affordable and accessible, have been widely used in many countries recently. An effective demand forecasting procedure should be developed to balance supply and demand. In the study for this purpose, bike sharing data for the years 20222023 in Konya, which has the longest bicycle path in Turkey, is examined. First, the data are summarized with descriptive statistics on an hourly and daily basis, and then the variables used in the forecasting models are defined. Multiple Linear Regression MLR, Ridge Regression RR, Elastic Net EN, Random Forest RF, Decision Trees DT, Support Vector Machines SVM, and Artificial Neural Networks ANN are used to estimate the actual demand. The hourly and daily demand forecasting models and the machine learning algorithms used are compared according to the coefficient of determination R2, mean absolute error MAE, mean square error MSE, and root mean square error RMSE. The model in which daily data is used is superior to the hourly model in all performance measures examined. When the algorithms are compared, the best technique for the hourly model is found to be ANN, while it is determined to be RF for the daily model.

 As the concept of sustainability increases day by day, eco-friendly techniques are researched instead of traditional transportation vehicles. One of these innovative solutions is the bike-sharing system, which provides carbon-emission-free transportation. The system greatly reduces the carbon footprint and also contributes to environmental sustainability by preventing urban traffic congestion. Bike-sharing systems, which are frequently preferred because they are economically affordable and accessible, have been widely used in many countries recently. An effective demand forecasting procedure should be developed to balance supply and demand. In the study for this purpose, bike sharing data for the years 20222023 in Konya, which has the longest bicycle path in Turkey, is examined. First, the data are summarized with descriptive statistics on an hourly and daily basis, and then the variables used in the forecasting models are defined. Multiple Linear Regression MLR, Ridge Regression RR, Elastic Net EN, Random Forest RF, Decision Trees DT, Support Vector Machines SVM, and Artificial Neural Networks ANN are used to estimate the actual demand. The hourly and daily demand forecasting models and the machine learning algorithms used are compared according to the coefficient of determination R2, mean absolute error MAE, mean square error MSE, and root mean square error RMSE. The model in which daily data is used is superior to the hourly model in all performance measures examined. When the algorithms are compared, the best technique for the hourly model is found to be ANN, while it is determined to be RF for the daily model.

 As the concept of sustainability increases day by day, eco-friendly techniques are researched instead of traditional transportation vehicles. One of these innovative solutions is the bike-sharing system, which provides carbon-emission-free transportation. The system greatly reduces the carbon footprint and also contributes to environmental sustainability by preventing urban traffic congestion. Bike-sharing systems, which are frequently preferred because they are economically affordable and accessible, have been widely used in many countries recently. An effective demand forecasting procedure should be developed to balance supply and demand. In the study for this purpose, bike sharing data for the years 20222023 in Konya, which has the longest bicycle path in Turkey, is examined. First, the data are summarized with descriptive statistics on an hourly and daily basis, and then the variables used in the forecasting models are defined. Multiple Linear Regression MLR, Ridge Regression RR, Elastic Net EN, Random Forest RF, Decision Trees DT, Support Vector Machines SVM, and Artificial Neural Networks ANN are used to estimate the actual demand. The hourly and daily demand forecasting models and the machine learning algorithms used are compared according to the coefficient of determination R2, mean absolute error MAE, mean square error MSE, and root mean square error RMSE. The model in which daily data is used is superior to the hourly model in all performance measures examined. When the algorithms are compared, the best technique for the hourly model is found to be ANN, while it is determined to be RF for the daily model.

 Sürdürülebilirlik kavramının her geçen gün öneminin arttığı çağımızda geleneksel ulaşım araçlarının yerine çevreye daha duyarlı alternatif yöntemler araştırılmaktadır. Bu yenilikçi çözümlerden birisi karbon emisyonsuz ulaşımı sağlayan paylaşımlı bisiklet kullanımıdır. Paylaşımlı bisiklet kullanımı, ulaşımda karbon ayak izini büyük ölçüde azaltmakta, ayrıca şehir içi trafik yoğunluğunu önleyerek çevresel sürdürülebilirliğe katkı sağlamaktadır. Ekonomik açıdan uygun fiyatlı ve erişilebilir olması nedeniyle de sıklıkla tercih edilen paylaşımlı bisikletler son yıllarda pek çok ülkede yaygın bir şekilde kullanılmaktadır. Paylaşımlı bisiklet sistemlerinden en yüksek faydayı sağlamak adına arz ve talep dengesi kurmak için etkin bir talep tahmin prosedürü geliştirilmelidir. Bu amaçla yapılan çalışmada Türkiyenin en uzun bisiklet yoluna sahip olan Konya ilinde 2022-2023 yıllarına ait paylaşımlı bisiklet kullanım verileri incelenmiştir. Öncelikle verilere ait saatlik ve günlük bazda tanıtıcı istatistikler verilerek veriler özetlenmiş, ardından tahmin modellerinde kullanılan değişkenler tanımlanmıştır. Paylaşımlı bisiklet kullanımına olan talebin tahmininde Çoklu Doğrusal Regresyon ÇDR, Ridge Regresyon RR, Elastik Net EN, Rassal Orman RO, Karar Ağaçları KA, Destek Vektör Makineleri DVM ve Yapay Sinir Ağları YSA kullanılmıştır. Geliştirilen günlük ve saatlik talep tahmin modelleri ve kullanılan makine öğrenme algoritmaları determinasyon katsayısı R2, ortalama mutlak hata MAE, ortalama hata karesi MSE ve kök ortalama hata karesi RMSE performans ölçütlerine göre karşılaştırılmıştır. Günlük verilerin değişken olarak kullanıldığı model, incelenen tüm performans ölçütlerine göre saatlik verilerin kullanıldığı modele göre üstünlük sağlamıştır. Algoritmalar karşılaştırıldığında saatlik veriler için en iyi performansa sahip teknik YSA iken, günlük veriler için ise en iyi performansı veren teknik RO olarak belirlenmiştir.

Article Details

DEMAND FORECASTING FOR BIKE SHARING SYSTEMS BY MACHINE LEARNING: A CASE STUDY IN KONYA

Authors:

Gözde Can Atasagun | Ece Çetin Yağmur

Publisher:

Eskişehir Osmangazi University

Main Topics:

Endüstri Mühendisliği

Journal:

Eskişehir Osmangazi Üniversitesi Mühendislik ve Mimarlık Fakültesi Dergisi Vol.32 Issue 3, pp.1470-1484

Document Type:

Article

Publication Date:

12/2024

Document Language:

Turkish

Document ID (Org No.):

2001333098

pISSN:

2630-5712

Bike-sharing systems

Machine learning

Artificial neural networks

Random forest

Demand forecastingPaylaşımlı bisiklet sistemleri

Makine öğrenmesi

Yapay sinir ağları

Rassal orman

Talep tahmini

Issues

Article Details

Authors: Gözde Can Atasagun | Ece Çetin Yağmur

Publisher: Eskişehir Osmangazi University

Main Topics: Endüstri Mühendisliği

Journal: Eskişehir Osmangazi Üniversitesi Mühendislik ve Mimarlık Fakültesi Dergisi Vol.32 Issue 3, pp.1470-1484

Document Type: Article

Publication Date: 12/2024

Document Language: Turkish

Document ID (Org No.): 2001333098

pISSN: 2630-5712

Issues

Authors:

Gözde Can Atasagun | Ece Çetin Yağmur

Publisher:

Eskişehir Osmangazi University

Main Topics:

Endüstri Mühendisliği

Journal:

Eskişehir Osmangazi Üniversitesi Mühendislik ve Mimarlık Fakültesi Dergisi Vol.32 Issue 3, pp.1470-1484

Document Type:

Article

Publication Date:

12/2024

Document Language:

Turkish

Document ID (Org No.):

2001333098

pISSN:

2630-5712