Sürdürülebilir Kentsel Hareketlilik için Açıklanabilir Makine Öğrenmesi: Hava-Mevsim Koşullarında Capital Bikeshare Talebi

Abstract

Bisiklet paylaşım sistemleri, hava durumuna ve mevsimselliğe duyarlı talebi görünür kılan ayrıntılı kullanım kayıtları üretmektedir fakat talep tahmini üzerine yapılan çalışmaların çoğu, işletmecilerin planlama için ihtiyaç duyduğu şeffaflıktan çok tahmin doğruluğuna odaklanmaktadır. Bu çalışma, kamuya açık Capital Bikeshare veri setini (Washington, D.C., 2011-2012) kullanarak günlük ve saatlik düzeyde bisiklet paylaşım talebi için açıklanabilir makine öğrenmesi modelleri geliştirmekte ve doğrulamaktadır. Sentetik doldurma yapılmadan gerçekten eksik olan 165 saatlik kaydın korunduğu, tek bir nem sensörü anomalisinin düzeltildiği ve hedef değişkene sızıntıyı önlemek için casual ile registered kullanıcı sayılarının açıklayıcı değişken olarak dışlandığı bir veri kalitesi denetiminin ardından günlük ve saatlik modeller, sıkı biçimde kronolojik bir doğrulama şeması -beş genişleyen takvim tabanlı çapraz doğrulama katmanı ve 2012’nin son çeyreğini kapsayan bir test kümesi- altında karşılaştırılmıştır. Günlük düzeyde doğrusal gün trendi içeren bir Ridge regresyonu (çapraz doğrulama RMSE = 979.4; test kümesi R² = 0.458), saatlik düzeyde ise aynı trend terimini içeren bir XGBoost modeli (çapraz doğrulama RMSE = 73.4; test kümesi R² = 0.757) seçilmiştir. SHAP tabanlı açıklama, günlük düzeyde en belirleyici değişkenler olarak gün trendini, yılı ve sıcaklığı; saatlik düzeyde ise saati, gün trendini ve çalışma günü durumunu göstermiştir. Duyarlılık analizi, aşırı düşük talepli tek bir günün (Capital Bikeshare’in Hurricane Sandy öncesi hizmeti askıya aldığı 29 Ekim 2012) günlük düzeydeki toplam mutlak yüzdesel hatanın %82’sini oluşturduğunu, buna karşın ölçeğe duyarsız hata ölçütlerinin (sMAPE, WAPE) görece durağan kaldığını göstermiştir. Varyans şişirme faktörü tanılamaları, sıcaklıkla güçlü çoklu bağlantı nedeniyle hissedilen sıcaklığın doğrusal model tasarımlarından çıkarılmasını gerektirmiştir. Bulgular, hava durumu ve mevsimselliğe duyarlı, kronolojik olarak doğrulanmış talep tahmininin sürdürülebilir kentsel hareketlilik planlaması için şeffaf bir karar destek aracı olabileceğini desteklerken hizmet kesintisi olaylarının modellenmeyen bir hata kaynağı olduğunu ortaya koymaktadır.

Keywords

• Açıklanabilir Makine Öğrenmesi
• Bisiklet Paylaşım Talebi
• SHAP
• Kronolojik Doğrulama
• Sürdürülebilir Kentsel Hareketlilik

Explainable Machine Learning for Sustainable Urban Mobility: Capital Bikeshare Demand under Weather and Seasonal Conditions

Abstract

Bike-sharing systems generate fine-grained ridership records that make weather- and season-sensitive demand visible, yet many prediction studies emphasize predictive accuracy over the transparency that operators need for planning. This study develops and validates explainable machine learning models for day- and hour-level bike-sharing demand using the publicly available Capital Bikeshare dataset (Washington, D.C., 2011-2012). Following a data-quality screening stage —which retained 165 genuinely absent hourly records without synthetic imputation, corrected a single humidity-sensor anomaly, and excluded casual and registered ridership counts as predictors to prevent target leakage— day- and hour-level models were compared under a strictly chronological validation scheme, using five expanding calendar-based cross-validation folds and a final-quarter 2012 hold-out. A Ridge regression with a linear day-trend term was selected at day level (cross-validation RMSE = 979.4; hold-out R² = 0.458), while an XGBoost model with the same trend term was selected at hour level (cross-validation RMSE = 73.4; hold-out R² = 0.757). SHAP-based explanation identified the day trend, year, and temperature as the leading day-level demand drivers, and hour-of-day, the day trend, and working-day status as the leading hour-level drivers. A sensitivity analysis showed that a single extreme low-demand day (29 October 2012, coinciding with Capital Bikeshare’s service suspension ahead of Hurricane Sandy) accounted for 82% of total day-level absolute percentage error, while scale-robust error measures (sMAPE, WAPE) remained comparatively stable. Variance-inflation diagnostics required excluding apparent temperature from linear model specifications due to strong collinearity with temperature. These findings support weather- and season-aware, chronologically validated demand forecasting as a transparent decision-support tool for sustainable urban mobility planning, while identifying service-disruption events as an unmodeled source of error.

Keywords

• Explainable Machine Learning
• Bike-Sharing Demand
• SHAP
• Chronological Validation
• Sustainable Urban Mobility

References

1. Baumanis, C., Hall, J., & Machemehl, R. (2023). A machine learning approach to predicting bicycle demand during the COVID-19 pandemic. Research in Transportation Economics, 100, 101276. https://doi.org/10.1016/j.retrec.2023.101276
2. Bean, R., Pojani, D., & Corcoran, J. (2021). How does weather affect bikeshare use? A comparative analysis of forty cities across climate zones. Journal of Transport Geography, 95, 103155. https://doi.org/10.1016/j.jtrangeo.2021.103155
3. Betkier, I., Oszczypała, M., & Dawid, W. (2025). Bike-sharing systems for sustainable cities: Demand prediction using graph neural networks and spatial aggregation. Sustainable Cities and Society, 128, 106479. https://doi.org/10.1016/j.scs.2025.106479
4. Biswas, M. E., Hossain, M. D., Huh, E.-N., & Sultana, T. (2025). Spatiotemporal data-driven hourly bike-sharing demand prediction using ApexBoost regression. International Journal of Data Science and Analytics, 20(7), 6259–6282. https://doi.org/10.1007/s41060-025-00820-0
5. Ermagun, A., & Pudleiner, K. (2026). Extreme weather effects on bikeshare ridership across climate regions. Journal of Transport Geography, 134, 104691. https://doi.org/10.1016/j.jtrangeo.2026.104691
6. Fanaee-T, H. (2013). Bike Sharing [Data set]. UCI Machine Learning Repository. https://doi.org/10.24432/C5W894
7. Fanaee-T, H., & Gama, J. (2014). Event labeling combining ensemble detectors and background knowledge. Progress in Artificial Intelligence, 2, 113–127. https://doi.org/10.1007/s13748-013-0040-3
8. Gammelli, D., Wang, Y., Prak, D., Rodrigues, F., Minner, S., & Pereira, F. C. (2022). Predictive and prescriptive performance of bike-sharing demand forecasts for inventory management. Transportation Research Part C: Emerging Technologies, 138, 103571. https://doi.org/10.1016/j.trc.2022.103571

9. Huang, J., Zheng, Y., Lan, L., & Chen, J. (2025). MADENet: Explainable AI-driven bike-sharing demand forecasting for sustainable urban mobility. Journal of Advances in Engineering and Technology, 2(4), 1-17. https://ojs.apspublisher.com/index.php/jaet/article/view/662
10. Jaber, A., & Csonka, B. (2023). Investigating the temporal differences among bike-sharing users through comparative analysis based on count, time series, and data mining models. Alexandria Engineering Journal, 77, 1–13. https://doi.org/10.1016/j.aej.2023.06.087
11. Kabir, S. M. R., & Zisha, F. K. (2026). Spatial dependency patterns in weather-enhanced bike-share demand forecasting in Washington DC. Transport Findings. https://doi.org/10.32866/001c.161506
12. Kalambay, P., & Pulugurtha, S. S. (2022). A before-after impact assessment of COVID-19 outbreak on bike-sharing ridership in Washington, DC. Urban, Planning and Transport Research, 10(1), 181–203. https://doi.org/10.1080/21650020.2022.2077820
13. Kang, Z., Chang, X., Sun, H., & Guo, X. (2025). Real-time reposition management of bike-sharing systems: A synchronous predict-then-optimize approach. Transportation Research Part A: Policy and Practice, 201, 104678. https://doi.org/10.1016/j.tra.2025.104678
14. Kharaghani, H., Etemadfard, H., & Golmohammadi, M. (2023). Spatio-temporal analysis of precipitation effects on bicycle-sharing systems with tensor approach. Journal of Geovisualization and Spatial Analysis, 7(2), 30. https://doi.org/10.1007/s41651-023-00161-1
15. Li, Z. (2022). Extracting spatial effects from machine learning model using local interpretation method: An example of SHAP and XGBoost. Computers, Environment and Urban Systems, 96, 101845. https://doi.org/10.1016/j.compenvurbsys.2022.101845
16. Lu, K.-F., Liu, Y., & Peng, Z.-R. (2025). Assessing the impacts of transit systems and urban street features on bike-sharing ridership: A graph-based spatiotemporal analysis and prediction model. Journal of Transport Geography, 128, 104356. https://doi.org/10.1016/j.jtrangeo.2025.104356
17. Ma, Y., Ali, Y., Morton, C., & Haitao, H. (2026). Assessing spatial and temporal transferability of cycleway impact models for bike-share usage in London. Scientific Reports, 16(1), 23016. https://doi.org/10.1038/s41598-026-52211-5
18. Ntamwiza, J. M. V., Bwire, H., & Nkurunziza, A. (2026). Mapping a decade of bike-sharing research published between (2015–2025): A systematic and bibliometric review highlighting gaps and future trends. Transport Economics and Management, 4, 75–84. https://doi.org/10.1016/j.team.2026.03.002
19. Odoom, C., Boateng, A., Mensah, S. F., & Maposa, D. (2024). Modeling of the daily dynamics in bike rental system using weather and calendar conditions: A semi-parametric approach. Scientific African, 24, e02211. https://doi.org/10.1016/j.sciaf.2024.e02211
20. Polamarasetti, A. (2025). Explainable graph neural networks for optimized allocation in spatio-temporal bike sharing demand prediction. Concurrency and Computation: Practice and Experience, 37(21–22), e70202. https://doi.org/10.1002/cpe.70202
21. Uddin, M., Hwang, H.-L., & Hasnine, M. S. (2023). An interpretable machine learning framework to understand bikeshare demand before and during the COVID-19 pandemic in New York City. Transportation Planning and Technology, 46(4), 482–498. https://doi.org/10.1080/03081060.2023.2201280
22. Washington City Paper. (2012). Not even bikeshare can stand up to Sandy. https://washingtoncitypaper.com/article/456977/not-even-bikeshare-can-stand-up-to-sandy/
23. Wu, J., Feng, J., Fang, J., & Liu, H. (2026). Demand prediction for bike-sharing systems: A spatial and semantic modeling approach for enhanced accuracy and operational efficiency. Computers & Industrial Engineering, 213, 111775. https://doi.org/10.1016/j.cie.2025.111775
24. Xin, Q. (2026). Probabilistic bike-sharing demand forecasting under changing weather and seasonal regimes with transformer-based models. Transport Findings. https://doi.org/10.32866/001c.157499
25. Yang, H. (2026). Urban greenness mitigates the negative effect of extreme heat on docked bike-sharing demand: Evidence from Chicago. Sustainable Cities and Society, 149, 107850. https://doi.org/10.1016/j.scs.2026.107850
26. Zhao, Y., Du, B., Luo, M., & Wen, H. (2025). Robust spatio-temporal demand prediction for bike-sharing systems with dynamic hierarchical structure. CCF Transactions on Pervasive Computing and Interaction, 7(2), 229–245. https://doi.org/10.1007/s42486-024-00167-8
27. Zhu, J., Chen, J., Li, M., & Tu, W. (2026). How spatial scales enhance prediction: An interpretable multi-scale framework for bike-sharing demand prediction. Geo-spatial Information Science, 29(1), 474–488. https://doi.org/10.1080/10095020.2025.2522149
28. Zhu, X., Wang, X., Sun, L., & Miranda-Moreno, L. (2026). Assessing resilience of urban bike-sharing systems: Weather and land-use influences on demand dynamics. Cities, 176, 107152. https://doi.org/10.1016/j.cities.2026.107152