Short-term traffic flow prediction using seasonal ARIMA model with limited input data

Kumar, S. Vasantha; Vanajakshi, Lelitha · 2015 · OpenAlex-citations

DOI: 10.1007/s12544-015-0170-8

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the challenge of short-term traffic flow prediction in Intelligent Transportation Systems (ITS) where extensive historical databases are unavailable. While Seasonal Autoregressive Integrated Moving Average (SARIMA) models are recognized as highly accurate for traffic forecasting, existing literature typically requires large datasets spanning weeks or months for model development. This paper proposes a prediction scheme using SARIMA that relies on limited input data—specifically, flow observations from only three consecutive days—to predict traffic flow for the subsequent 24 hours. The research aims to demonstrate that such a minimal dataset is sufficient for generating accurate forecasts, thereby making SARIMA applicable in contexts where data storage or collection is constrained. The methodology involved collecting traffic flow data from a 3-lane arterial roadway in Chennai, India, using an automated Collect-R camera. Raw data, recorded at one-minute intervals, was aggregated into 10-minute intervals and converted to vehicles per hour, resulting in 144 data points per day. The model development utilized data from three consecutive days (September 20–22, 2012), while the following day (September 23, 2012) served as the validation set. The researchers applied seasonal differencing at a lag of 144 to account for daily seasonality and rendered the time series stationary. Model identification was performed by analyzing autocorrelation (ACF) and partial autocorrelation (PACF) functions, suggesting potential non-seasonal AR orders of 1, 2, or 3, combined with a seasonal MA order of 1. Model parameters were estimated using the maximum likelihood method in R software. The optimal model was selected based on the Akaike Information Criteria (AIC), with the SARIMA (2,0,0)×(0,1,1)144 model chosen due to its lowest AIC value of 4218.3. The results demonstrated that the proposed model achieved a Mean Absolute Percentage Error (MAPE) of 9.22% when predicting the 24-hour flow for the validation day. According to Lewis’ scale of interpretation, this error rate classifies the forecast as "highly accurate." The predicted flow values closely matched observed flows during both peak and off-peak hours. Comparative analysis showed that the SARIMA model outperformed simpler baseline methods; the historic average method yielded a MAPE of 10.53%, and the naive method resulted in a MAPE of 10.42%. The study also explored the effect of increasing input data from three to nine days, finding that while additional data slightly reduced error, the improvement was not significant enough to justify the increased data requirement. Furthermore, short-term predictions during morning and evening peak periods were successfully attempted using both historic and real-time data. The significance of this work lies in its demonstration that SARIMA models can be effectively deployed for ITS applications without the burden of maintaining large historical databases. By proving that three days of data are sufficient for high-accuracy 24-hour forecasts, the study expands the applicability of time-series analysis to locations with limited data infrastructure. The findings suggest that this lightweight prediction scheme is a viable, sustainable alternative to capital-intensive infrastructure strategies, offering a practical solution for real-time traffic management in metropolitan areas where data availability is a primary constraint.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• telematics crash prediction