A Hybrid Deep Learning Model for Traffic Flow Prediction

Yavuz Canbay; Alper Talha Karadeniz

Makale detayları

A Hybrid Deep Learning Model for Traffic Flow Prediction

Yazarlar:

Yavuz Canbay | Alper Talha Karadeniz

Yayıncı:

Erciyes University

Ana Başlıklar:

Derin Öğrenme | Nöral Ağlar

Dergi:

Erciyes Üniversitesi Fen Bilimleri Enstitüsü Fen Bilimleri Dergisi Vol.40 Issue 3, pp.518-527

Belge Tipi:

Makale

Yayın Tarihi:

12/2024

Belge Dili:

İngilizce

Document ID (Org No.):

2001337248

pISSN:

1012-2354

Anahtar kelimeler:

Özet:

 Urbanization has led to a rise in traffic issues, which not only result in economic losses but also cause to environmental degradation, posing a threat to the quality of life for city residents. Identifying traffic issues and mitigating traffic congestion proactively is of utmost significance. Traffic flow prediction is analyzing historical and current traffic data to predict future traffic conditions. Precise predictions enable people to make suitable choices regarding travel routes and transportation methods. This reduces the duration of trip and enhances the level of satisfaction with traffic conditions. Traditional traffic prediction methods rely on statistical techniques or simple time series analysis. However, these analyses suffer to capture the complex spatial and temporal dependencies in traffic data. Recent advancements in artificial intelligence have enabled to create complex and data-driven prediction models. This paper introduces a novel hybrid model for predicting traffic flow using deep learning techniques. The performance of the proposed model was evaluated against some baseline algorithms on a real-world dataset. Regression metrics such as Mean Absolute Error MAE, Mean Square Error MSE, R-Square R², and Root Mean Square Error RMSE were employed to compare the performances of the models. In the experiments, it has been seen that the proposed hybrid model outperformed other models, with 5.1834 MAE, 54.4060 MSE, 7.3760 RMSE and 0.9923 R2 values. The results emphasize the potential of deep learning methods in the domain of traffic prediction and offer direction for future investigations.

 Urbanization has led to a rise in traffic issues, which not only result in economic losses but also cause to environmental degradation, posing a threat to the quality of life for city residents. Identifying traffic issues and mitigating traffic congestion proactively is of utmost significance. Traffic flow prediction is analyzing historical and current traffic data to predict future traffic conditions. Precise predictions enable people to make suitable choices regarding travel routes and transportation methods. This reduces the duration of trip and enhances the level of satisfaction with traffic conditions. Traditional traffic prediction methods rely on statistical techniques or simple time series analysis. However, these analyses suffer to capture the complex spatial and temporal dependencies in traffic data. Recent advancements in artificial intelligence have enabled to create complex and data-driven prediction models. This paper introduces a novel hybrid model for predicting traffic flow using deep learning techniques. The performance of the proposed model was evaluated against some baseline algorithms on a real-world dataset. Regression metrics such as Mean Absolute Error MAE, Mean Square Error MSE, R-Square R², and Root Mean Square Error RMSE were employed to compare the performances of the models. In the experiments, it has been seen that the proposed hybrid model outperformed other models, with 5.1834 MAE, 54.4060 MSE, 7.3760 RMSE and 0.9923 R2 values. The results emphasize the potential of deep learning methods in the domain of traffic prediction and offer direction for future investigations.

 Urbanization has led to a rise in traffic issues, which not only result in economic losses but also cause to environmental degradation, posing a threat to the quality of life for city residents. Identifying traffic issues and mitigating traffic congestion proactively is of utmost significance. Traffic flow prediction is analyzing historical and current traffic data to predict future traffic conditions. Precise predictions enable people to make suitable choices regarding travel routes and transportation methods. This reduces the duration of trip and enhances the level of satisfaction with traffic conditions. Traditional traffic prediction methods rely on statistical techniques or simple time series analysis. However, these analyses suffer to capture the complex spatial and temporal dependencies in traffic data. Recent advancements in artificial intelligence have enabled to create complex and data-driven prediction models. This paper introduces a novel hybrid model for predicting traffic flow using deep learning techniques. The performance of the proposed model was evaluated against some baseline algorithms on a real-world dataset. Regression metrics such as Mean Absolute Error MAE, Mean Square Error MSE, R-Square R², and Root Mean Square Error RMSE were employed to compare the performances of the models. In the experiments, it has been seen that the proposed hybrid model outperformed other models, with 5.1834 MAE, 54.4060 MSE, 7.3760 RMSE and 0.9923 R2 values. The results emphasize the potential of deep learning methods in the domain of traffic prediction and offer direction for future investigations.

 Kentleşme, trafik sorunlarının artmasına yol açmış, bu durum yalnızca ekonomik kayıplara yol açmakla kalmayıp aynı zamanda çevresel bozulmaya da neden olarak kent sakinlerinin yaşam kalitesini tehdit etmektedir. Trafik sorunlarını tespit etmek ve trafik sıkışıklığını proaktif olarak azaltmak son derece önemlidir. Trafik akışı tahmini, gelecekteki trafik koşullarını tahmin etmek için geçmiş ve güncel trafik verilerini analiz ediyor. Kesin tahminler, insanların seyahat rotaları ve ulaşım yöntemleri konusunda uygun seçimler yapmasına olanak sağlar. Bu, yolculuk süresini azaltır ve trafik koşullarından memnuniyet düzeyini artırır. Geleneksel trafik tahmin yöntemleri istatistiksel tekniklere veya basit zaman serisi analizine dayanır. Ancak bu analizler, trafik verilerindeki karmaşık mekansal ve zamansal bağımlılıkları yakalama konusunda sıkıntı yaşamaktadır. Yapay zekadaki son gelişmeler, karmaşık ve veriye dayalı tahmin modellerinin oluşturulmasına olanak sağlamıştır. Bu makalede, derin öğrenme tekniklerini kullanarak trafik akışını tahmin etmek için yeni bir hibrit model önerilmektedir. Önerilen modelin performansı, gerçek dünya veri kümesindeki bazı temel algoritmalara göre değerlendirilmiş ve modellerin performanslarını karşılaştırmak için Ortalama Mutlak Hata MAE, Ortalama Kare Hata MSE, R-Kare R² ve Ortalama Karekök Hata RMSE gibi regresyon metrikleri kullanılmıştır. Yapılan deneylerde önerilen hibrit modelin 5,1834 MAE, 54,4060 MSE, 7,3760 RMSE ve 0,9923 R2 değerleriyle diğer modellerden daha iyi performans gösterdiği görülmüştür. Sonuçlar, trafik tahmini alanında derin öğrenme yöntemlerinin potansiyelini vurgulamakta ve gelecekteki araştırmalar için yol gösterici olduğunu göstermektedir.

Makale detayları

A Hybrid Deep Learning Model for Traffic Flow Prediction

Yazarlar:

Yavuz Canbay | Alper Talha Karadeniz

Yayıncı:

Erciyes University

Ana Başlıklar:

Derin Öğrenme | Nöral Ağlar

Dergi:

Erciyes Üniversitesi Fen Bilimleri Enstitüsü Fen Bilimleri Dergisi Vol.40 Issue 3, pp.518-527

Belge Tipi:

Makale

Yayın Tarihi:

12/2024

Belge Dili:

İngilizce

Document ID (Org No.):

2001337248

pISSN:

1012-2354

traffic flow

prediction

deep learningtrafik akışı

tahmin

derin öğrenme

Sayı

Makale detayları

A Hybrid Deep Learning Model for Traffic Flow Prediction

Yazarlar: Yavuz Canbay | Alper Talha Karadeniz

Yayıncı: Erciyes University

Ana Başlıklar: Derin Öğrenme | Nöral Ağlar

Dergi: Erciyes Üniversitesi Fen Bilimleri Enstitüsü Fen Bilimleri Dergisi Vol.40 Issue 3, pp.518-527

Belge Tipi: Makale

Yayın Tarihi: 12/2024

Belge Dili: İngilizce

Document ID (Org No.): 2001337248

pISSN: 1012-2354

traffic flow

prediction

deep learningtrafik akışı

tahmin

derin öğrenme

Sayı

Yazarlar:

Yavuz Canbay | Alper Talha Karadeniz

Yayıncı:

Erciyes University

Ana Başlıklar:

Derin Öğrenme | Nöral Ağlar

Dergi:

Erciyes Üniversitesi Fen Bilimleri Enstitüsü Fen Bilimleri Dergisi Vol.40 Issue 3, pp.518-527

Belge Tipi:

Makale

Yayın Tarihi:

12/2024

Belge Dili:

İngilizce

Document ID (Org No.):

2001337248

pISSN:

1012-2354