Predicting Traffic Data in GIS using Different Neural Network Methods
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 predicting traffic flow in metropolitan areas, specifically focusing on forecasting future traffic conditions based on historical data. While previous research has largely concentrated on traffic control and reduction, this paper aims to fill a gap by comparing the effectiveness of different neural network algorithms for time-series traffic prediction. The motivation stems from the significant impact of traffic on urban life, including pollution and economic burden, and the need for accurate forecasting to aid in route selection and urban planning. The research was conducted in Tehran, Iran, covering its 22 municipal districts. Traffic data were collected using a web-based system interfacing with the Neshan map platform, which categorizes traffic intensity into four levels (light, semi-heavy, heavy, and very heavy) represented by numerical codes. Data were extracted for nine consecutive days (September 5–13, 2022) to serve as input for training models to predict traffic on the tenth day (September 14). Four neural network methods were employed: Basic Neural Network, Feed-forward Levenberg-Marquardt, Conjugate Gradient, and Bayesian Neural Network. For the latter three methods, the data were split into 70% training, 15% validation, and 15% test sets. Performance was evaluated using Mean Squared Error (MSE) and regression values across all 24 hours of the day. The results indicate that the Feed-forward Levenberg-Marquardt method achieved the highest prediction accuracy at 81.59%, followed closely by the Bayesian Neural Network (81.55%) and the Conjugate Gradient method (81.50%). The Basic Neural Network performed the least accurately, with a 75% accuracy rate. In terms of regression values, which measure the proximity between input and output data, the Basic Neural Network showed an approximate 80% correlation. In contrast, the Feed-forward Levenberg-Marquardt, Conjugate Gradient, and Bayesian Neural Network methods exhibited lower regression values of 69.69%, 69.71%, and 69.87%, respectively. Despite the lower regression scores for the advanced methods, their lower MSE values confirmed their superior predictive capability compared to the basic model. The study concludes that the Feed-forward Levenberg-Marquardt algorithm is the most effective method for predicting traffic time series among those tested. This finding suggests that advanced optimization algorithms within neural networks can significantly improve the accuracy of traffic forecasting models. The implications for the field include the potential integration of these high-accuracy prediction models into Geographic Information Systems (GIS) and urban traffic management tools, aiding in better route planning and traffic control strategies in large cities.
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.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | Crossref | — | — | 1 | 2026-06-25 |
| archive | success | canonical_url | — | — | 1 | 2026-06-26 |
| extract | success | cached | — | — | 5 | 2026-06-26 |
| clean | success | clean | — | — | 1 | 2026-06-25 |
| chunk | success | chunk | — | — | 1 | 2026-06-25 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-25 |
| promote | success | — | — | — | 1 | 2026-06-25 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 4 | 2026-06-26 |
| tag | success | vector_similarity | — | — | 6 | 2026-06-25 |
| verify | success | — | — | — | 1 | 2026-06-26 |
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.