Real-time data from mobile platforms to evaluate sustainable transportation infrastructure


By displacing gasoline and diesel fuels, electric cars and fleets reduce emissions from the transportation sector, thus offering important public health benefits. However, public confidence in the reliability of charging infrastructure remains a fundamental barrier to adoption. Using large-scale social data and machine-learning based on 12,720 electric vehicle (EV) charging stations, we provide national evidence on how well the existing charging infrastructure is serving the needs of the rapidly expanding population of EV drivers in 651 core-based statistical areas in the United States. We deploy supervised machine-learning algorithms to automatically classify unstructured text reviews generated by EV users. Extracting behavioural insights at a population scale has been challenging given that streaming data can be costly to hand classify. Using computational approaches, we reduce processing times for research evaluation from weeks of human processing to just minutes of computation. Contrary to theoretical predictions, we find that stations at private charging locations do not outperform public charging locations provided by the government. Overall, nearly half of drivers who use mobility applications have faced negative experiences at EV charging stations in the early growth years of public charging infrastructure, a problem that needs to be fixed as the market for electrified and sustainable transportation expands.

Access options

Rent or Buy article

Get time limited or full article access on ReadCube.


All prices are NET prices.

Fig. 1: US map of active charge station reviews.
Fig. 2: Model architecture for the CNN.
Fig. 3: Saliency heatmap for reviews with a domain-specific term ‘iced’.
Fig. 4: Predicted probability of negative sentiment in public and private spaces.
Fig. 5: Predicted probability of negative sentiment by geographical area and POI.

Data availability

We provide the weights of the trained deep-learning models. These datasets generated and/or analysed during the current study are available in the Figshare repository The raw data that support the findings of this study are available from the corresponding author upon request. These data may not be posted publicly due to privacy restrictions. For interested readers, an alternative open data API service with global EV charging infrastructure data is available from OpenChargeMap (, which is derived from a variety of public sources and contributions. Source Data are provided with this paper.

Code availability

All custom code and algorithm replication materials have been deposited on the Github repository using Zenodo version releases at


  1. 1.

    Market Data: EV Market Forecasts: Global Forecasts for Light Duty Plug-In Hybrid and Battery EV Sales and Populations: 2018–2030 (Navigant, 2019).

  2. 2.

    Inventory of US Greenhouse Gas Emissions and Sinks: 1990-2016 Document No. 430-R-18-003 (EPA, 2018).

  3. 3.

    The electric battery vehicle. Nature 144, 627 (1939).

  4. 4.

    Michalek, J. J. et al. Valuation of plug-in vehicle life-cycle air emissions and oil displacement benefits. Proc. Natl Acad. Sci. USA 108, 16554–16558 (2011).

    CAS  Article  Google Scholar 

  5. 5.

    Tessum, C. W., Hill, J. D. & Marshall, J. D. Life cycle air quality impacts of conventional and alternative light-duty transportation in the United States. Proc. Natl Acad. Sci. USA 111, 18490–18495 (2014).

    CAS  Article  Google Scholar 

  6. 6.

    Holland, S. P., Mansur, E. T., Muller, N. Z. & Yates, A. J. Are there environmental benefits from driving electric vehicles? The importance of local factors. Am. Econ. Rev. 106, 3700–3729 (2016).

    Article  Google Scholar 

  7. 7.

    Li, S., Tong, L., Xing, J. & Zhou, Y. The market for electric vehicles: indirect network effects and policy design. J. Assoc. Environ. Resour. Econ. 4, 89–133 (2017).

    Google Scholar 

  8. 8.

    Andreoni, J. Impure altruism and donations to public goods: a theory of warm-glow giving. Econ. J. 100, 464–477 (1990).

    Article  Google Scholar 

  9. 9.

    Kotchen, M. Green markets and private provision of public goods. J. Political Econ. 114, 816–834 (2006).

    Article  Google Scholar 

  10. 10.

    Warner, M. & Amir, H. Managing markets for public service: the role of mixed public–private delivery of city services. Public Adm. Rev. 68, 155–166 (2008).

    Article  Google Scholar 

  11. 11.

    Warner, M. & Hebdon, R. Local government restructuring: privatization and its alternatives. J. Policy Anal. Manag. 20, 315–336 (2001).

    Article  Google Scholar 

  12. 12.

    Carley, S., Krause, R. M., Lane, B. W. & Graham, J. D. Intent to purchase a plug-in electric vehicle: a survey of early impressions in large US cites. Transp. Res. Part D 18, 39–45 (2013).

    Article  Google Scholar 

  13. 13.

    Sovacool, B. K. & Hirsh, R. F. Beyond batteries: an examination of the benefits and barriers to plug-in hybrid electric vehicles (PHEVS) and a vehicle-to-grid (V2G) transition. Energy Policy 37, 1095–1103 (2009).

    Article  Google Scholar 

  14. 14.

    Helveston, J. P. et al. Choice at the pump: measuring preferences for lower-carbon combustion fuels. Environ. Res. Lett. 14, 084035 (2019).

    Article  Google Scholar 

  15. 15.

    Xu, Y., Çolak, S., Kara, E. C., Moura, S. J. & González, M. C. Planning for electric vehicle needs by coupling charging profiles with urban mobility. Nat. Energy 3, 484–493 (2018).

  16. 16.

    Asensio, O. I. Correcting consumer misperception. Nat. Energy 4, 823–824 (2019).

    Article  Google Scholar 

  17. 17.

    Williams, B. & DeShazo, J. Pricing workplace charging: financial viability and fueling costs. Transp. Res. Rec. 2454, 68–75 (2014).

    Article  Google Scholar 

  18. 18.

    Asensio, O. I. & Delmas, M. A. Nonprice incentives and energy conservation. Proc. Natl Acad. Sci. USA 112, E510–E515 (2015).

    CAS  Article  Google Scholar 

  19. 19.

    Asensio, O. I. & Delmas, M. A. The dynamics of behavior change: evidence from energy conservation. J. Econ. Behav. Organ. 126, 196–212 (2016).

    Article  Google Scholar 

  20. 20.

    Alexander, L., Jiang, S., Murga, M. & González, M. C. Origin–destination trips by purpose and time of day inferred from mobile phone data. Transp. Res. Part C 58, 240–250 (2015).

    Article  Google Scholar 

  21. 21.

    González, M. C., Hidalgo, C. A. & Barabasi, A.-L. Understanding individual human mobility patterns. Nature 453, 779–782 (2008).

    Article  Google Scholar 

  22. 22.

    Kim, Y. Convolutional neural networks for sentence classification. In Proc. 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP) (eds Moschitti, A. et al.) 1746–1751 (Association for Computational Linguistics, 2014).

  23. 23.

    LeCun, Y., Bengio, Y. & Hinton, G. Deep learning. Nature 521, 436–444 (2015).

    CAS  Article  Google Scholar 

  24. 24.

    Socher, R. et al. Recursive deep models for semantic compositionality over a sentiment treebank. In Proc. 2013 Conference on Empirical Methods in Natural Language Processing (eds Yarowsky, D. et al.) 1631–1642 (2013).

  25. 25.

    dos Santos, C. & Gatti, M. Deep convolutional neural networks for sentiment analysis of short texts. In Proc. COLING 2014, the 25th International Conference on Computational Linguistics: Technical Papers (eds Tsujii, J. & Hajic, J.) 69–78 (Dublin City University and Association for Computational Linguistics, 2014).

  26. 26.

    Yih, W.-t., Toutanova, K., Platt, J. C. & Meek, C. Learning discriminative projections for text similarity measures. In Proc. Fifteenth Conference on Computational Natural Language Learning (ed. Pradhan, S.) 247–256 (Association for Computational Linguistics, 2011).

  27. 27.

    Shen, Y., He, X., Gao, J., Deng, L. & Mesnil, G. Learning semantic representations using convolutional neural networks for web search. In Proc. 23rd International Conference on World Wide Web (ed. Chung, C.-W.) 373–374 (ACM, 2014).

  28. 28.

    Kalchbrenner, N., Grefenstette, E., & Blunsom, P. A convolutional neural network for modelling sentences. In Proc. 52nd Annual Meeting of the Association for Computational Linguistics (eds Toutanova, K. and Wu, H.) 655–665 (Association for Computational Linguistics, 2014).

  29. 29.

    Collobert, R. et al. Natural language processing (almost) from scratch. J. Mach. Learn. Res. 12, 2493–2537 (2011).

    Google Scholar 

  30. 30.

    Mikolov, T., Sutskever, I., Chen, K., Corrado, G. S. & Dean, J. in Advances in Neural Information Processing Systems (eds Burges, C. J. C. et al.) 3111–3119 (Curran Associates, Inc., 2013).

  31. 31.

    Hochreiter, S. & Schmidhuber, J. Long short-term memory. Neural Comput. 9, 1735–1780 (1997).

    CAS  Article  Google Scholar 

  32. 32.

    Yin, W., Kann, K., Yu, M. & Schütze, H. Comparative study of CNN and RNN for natural language processing. Preprint at (2017).

  33. 33.

    Li, J., Chen, X., Hovy, E. & Jurafsky, D. Visualizing and understanding neural models in NLP. In Proc. 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (eds Knight, K. et al.) 681–691 (Association for Computational Linguistics, 2016).

  34. 34.

    Tixier, A. J.-P. Notes on deep learning for NLP. Preprint at (2018).

  35. 35.

    Song, B., Lee, C., Yoon, B. & Park, Y. Diagnosing service quality using customer reviews: an index approach based on sentiment and gap analyses. Serv. Bus. 10, 775–798 (2016).

    Article  Google Scholar 

  36. 36.

    Wang, H., Can, D., Kazemzadeh, A., Bar, F. & Narayanan, S. A system for real-time Twitter sentiment analysis of 2012 US presidential election cycle. In Proc. ACL 2012 System Demonstrations (ed. Zhang, M.) 115–120 (Association for Computational Linguistics, 2012).

  37. 37.

    Hopkins, D. J. & King, G. A method of automated nonparametric content analysis for social science. Am. J. Political Sci. 54, 229–247 (2010).

    Article  Google Scholar 

  38. 38.

    Bail, C. A. The cultural environment: measuring culture with big data. Theory Soc. 43, 465–482 (2014).

    Article  Google Scholar 

  39. 39.

    Ulibarri, N., Scott, T. A. & Perez-Figueroa, O. How does stakeholder involvement affect environmental impact assessment? Environ. Impact Assess. Rev. 79, 106309 (2019).

    Article  Google Scholar 

  40. 40.

    Board, T. R. & Council, N. R. Overcoming Barriers to Deployment of Plug-in Electric Vehicles (National Academies Press, 2015).

  41. 41.

    Gerardo Zarazua de Rubens, L. N. & Sovacool, B. K. Dismissive and deceptive car dealerships create barriers to electric vehicle adoption at the point of sale. Nat. Energy 3, 501–507 (2018).

    Article  Google Scholar 

  42. 42.

    Rezvani, Z., Jansson, J. & Bodin, J. Advances in consumer electric vehicle adoption research: a review and research agenda. Transp. Res. Part D 34, 122–136 (2015).

    Article  Google Scholar 

  43. 43.

    Plug-In Electric Vehicle Deployment Policy Tools: Zoning, Codes, and Parking Ordinances Technology Bulletin (DOE, 2015).

  44. 44.

    National Plug-In Electric Vehicle Infrastructure Analysis (DOE, 2017).

  45. 45.

    DeShazo, J. R. Improving incentives for clean vehicle purchases in the United States: challenges and opportunities. Rev. Environ. Econ. Policy 10, 149–165 (2016).

    Article  Google Scholar 

  46. 46.

    DeShazo, J., Sheldon, T. L. & Carson, R. T. Designing policy incentives for cleaner technologies: lessons from California’s plug-in electric vehicle rebate program. J. Environ. Econ. Manag. 84, 18–43 (2017).

    Article  Google Scholar 

  47. 47.

    Zhang, Y. & Wallace, B. C. A sensitivity analysis of (and practitioners’ guide to) convolutional neural networks for sentence classification. In Proc. Eighth International Joint Conference on Natural Language Processing (eds Kondrak, G. & Watanabe, T.) 253–263 (Asian Federation of Natural Language Processing, 2017).

  48. 48.

    Srivastava, N., Hinton, G., Krizhevsky, A., Sutskever, I. & Salakhutdinov, R. Dropout: a simple way to prevent neural networks from overfitting. J. Mach. Learn. Res. 15, 1929–1958 (2014).

    Google Scholar 

  49. 49.

    Karpathy, A., Johnson, J. & Fei-Fei, L. Visualizing and understanding recurrent networks. In International Conference on Learning Representations 2016 Workshop (International Conference on Learning Representations, 2016).

  50. 50.

    Cohen, J. A coefficient of agreement for nominal scales. Educ. Psychol. Meas. 20, 37–46 (1960).

    Article  Google Scholar 

  51. 51.

    Papke, L. E. & Wooldridge, J. M. Econometric methods for fractional response variables with an application to 401(k) plan participation rates. J. Appl. Econ. 11, 619–632 (1996).

    Article  Google Scholar 

  52. 52.

    Papke, L. E. & Wooldridge, J. M. Panel data methods for fractional response variables with an application to test pass rates. J. Econ. 145, 121–133 (2008).

    Article  Google Scholar 

  53. 53.

    Ramalho, E. A., Ramalho, J. J. & Murteira, J. M. Alternative estimating and testing empirical strategies for fractional regression models. J. Econ. Surv. 25, 19–68 (2011).

    Article  Google Scholar 

  54. 54.

    2010 Census Urban and Rural Classification and Urban Area Criteria (US Census, 2010).

  55. 55.

    Conneau, A., Schwenk, H., Barrault, L. & Lecun, Y. Very deep convolutional networks for text classification. In Proc. 15th Conference of the European Chapter of the Association for Computational Linguistics Vol. 1 (eds Lapata, M. et al.) 1107–1116 (Association for Computational Linguistics, 2017).

  56. 56.

    Asensio, O. I. & Ha, S. Trained CNN and LSTM model on EV charging station reviews for sentiment analysis (Figshare, 2020);

Download references


This research was supported by a grant from the National Science Foundation (CPS award no. 1931980), the Civic Data Science REU programme at Georgia Tech (NSF award no. IIS-1659757), the Anthony and Jeanne Pritzker Family Foundation, the Sustainable LA Grand Challenge. We are grateful to E. Zegura and C. Le Dantec for discussions. For valuable research assistance, we thank M. E. Burke, S. Dharur, S. Oh and D. Marchetto. Special thanks to N. Hajjar.

This research was also supported in part through research cyberinfrastructure resources and services provided by the Partnership for an Advanced Computing Environment (PACE) at the Georgia Institute of Technology, Atlanta, Georgia, USA.

Author information




O.I.A. directed the research and wrote the paper; A.D., E.W. and K.A. developed code, analysed data and wrote the paper; K.A., C.H. and S.H. implemented algorithms and performed experiments; S.H. investigated model validation and interpretability. All authors reviewed the manuscript.

Corresponding author

Correspondence to Omar Isaac Asensio.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data

Extended Data Fig. 1 Distribution of CNN classifier predictions for 1,000 model runs.

The mean test accuracy for 1,000 runs is 84.6% with a S.D. of 0.79. Source data

Extended Data Table 1 Descriptive statistics, public and private.

Counts of machine classified reviews of binary sentiment by public and private ownership. 2,256 reviews were submitted in locations where it was impossible to discern whether it was public or private. Source data

Extended Data Table 2 Main results.

FRM results for the review rate and negativity score.

Extended Data Table 3 Descriptive statistics, urban and rural.

Counts of machine classified reviews of binary sentiment by geographic area type as defined by U.S. Census designations. Source data

Extended Data Table 4 Probability of negative sentiment for 18 core-based statistical areas in the United States.

Results of t-tests for free and paid stations by public and private ownership in 18 CBSAs in the United States. Source data

Extended Data Table 5 Probability of negative sentiment for top 20 U.S. states.

Results of t-tests for free and paid stations by public and private ownership in the top 20 states by number of reviews. Source data

Extended Data Table 6 Balance of training data.

Counts of positive and negative reviews by two human annotators (κ=0.84). Source data

Supplementary information

Supplementary Information

Supplementary discussion, Fig. 1, Tables 1–7 and references.

Source data

Source Data Fig. 1

Contains state ID and number of reviews.

Source Data Fig. 3

Contains the weights of the trained CNN model used to create the heatmap figure

Source Data Fig. 4

Contains location ID, Public or Private designations, Free or Paid Designations

Source Data Fig. 5

Contains location ID, U.S. Census area type, point of interest, and predicted sentiment

Source Data Extended Data Fig. 1

Contains 1000 CNN performance runs with accuracy in percentage

Source Data Extended Data Table 1

Contains location ID, public or private designations, predicted sentiment

Source Data Extended Data Table 3

Contains location ID, U.S. Census area type, and predicted sentiment per review

Source Data Extended Data Table 4

CBSA, free or paid designations, public or private designations, and predicted sentiment per review

Source Data Extended Data Table 5

Public or private designations, free or paid designations, state, predicted sentiment per review

Source Data Extended Data Table 6

Contains review ID and human expert labels positive or negative

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Asensio, O.I., Alvarez, K., Dror, A. et al. Real-time data from mobile platforms to evaluate sustainable transportation infrastructure. Nat Sustain 3, 463–471 (2020).

Download citation