Real-time taxi demand prediction using recurrent neural network

Donggyun Ku, Sungyong Na, Jooyoung Kim, Seungjae Lee

Research output: Contribution to journalArticlepeer-review

6 Scopus citations


The study aims to predict the location of taxi users, based on an algorithm that was built using a map learning method, which is one of the techniques of deep learning. As the location data of taxi riders showed sequential characteristics over time, learning was performed using a recurrent neural network, which is suitable for predicting dynamic changes over time. The main data used in the analysis were the Seoul Metropolitan Government's taxi tachometer data. These data were collected over a span of six months, from February 2018 to July 2018. Seoul Metropolitan Government's building data and Seoul public transportation smart card data were used as secondary data sources to reflect taxi traffic characteristics. Deep learning results were reviewed using different accuracy values based on combinations of the data sources, such as taxi data only, taxi data and building data and taxi data and smart card data. As a result, the algorithm was able to accurately obtain the distribution of taxi passengers' boarding positions compared to the actual taxi riding pattern through statistical analysis. On the basis of these predictions, the asymmetric characteristics of taxi traffic in terms of transport planning and management can be solved.

Original languageEnglish
Pages (from-to)75-87
Number of pages13
JournalProceedings of the Institution of Civil Engineers: Municipal Engineer
Issue number2
StatePublished - Jun 2021


  • Statistical analysis
  • Transport management
  • Transport planning


Dive into the research topics of 'Real-time taxi demand prediction using recurrent neural network'. Together they form a unique fingerprint.

Cite this