confefrence presentation.pptxconfefrence presentation.pptx

confefrence presentation.pptx

1.     Traffic Congestion Prediction Using Machine Learning: Amman City Case Study By Areen Arabiat Supervisor Prof. Omar Al- Momani   Co-Supervisor Dr. Mohammad Hassan  

2.     Outline : Introduction Problem Statement Research objectives Research Scope Machine Learning Approaches and classifiers Research Methodology Results and Implementation 9.Results 10. Conclusion  

3. Introduction The cities grow in size, and people move from rural to urban areas, the number of vehicles may grow rabidly and potentially causing traffic congestion. The most common definition of traffic congestion is when demand for transport exceeds the capacity of roads. Traffic congestion can be classified into two types: recurrent congestions, which are usually caused by a mobility demand , and intermittent congestions, which are usually caused by a lack of mobility. that surpasses the road network's capacity , including nonrecurring traffic jams caused by incidents or special events.

4. The intelligent transportation systems (ITS) is used to improve the efficiency and capacity of transportation networks by prediction traffic congestion. Despite its relevance, traffic congestion prediction receives far less research than traffic flow prediction. The role of predicting traffic congestion is critical in the implementation of ITS, Traffic congestion forecasting will also provide traffic information ahead of time for the decision makers. Traffic congestion prediction has created a growing research area, especially in the field of Machine Learning (ML) Introduction

5. Problem Statement Traffic congestion is one of the most serious issues threatening societies , and it is constantly worsening as a result of rapid urbanization and significant increase in the number of vehicles on the road. This issue has significant economic, health, and environmental implications , in addition to road users‘ dissatisfaction with the quality of transportation . This study aims to use machine learning to detect traffic congestion and assist drivers in avoiding congested areas. This can be accomplished by deploying some algorithms of ML and artificial intelligence on the collected data related to 8th Circle approaches in Amman city.

6. Research objectives 1. To build a dataset for traffic data at the 8th Circle approaches. 2. To predicate the traffic congestion at The 8th Circle approaches using The Logistic Regression, K-nearest neighbors algorithm (KNN), Decision Trees (DT), Random Forest, SVM, and multilayer perceptron (MLP) classifiers. 3. To determine the best ML classifiers for predicating traffic congestion.

7. Research Scope

8. Machine Learning Approaches

9. Machine Learning classifiers

10. Research Methodology

11. Results and Implementation First Experiment ( Zahran street (7th Circle)) ,App (1,2,3) Second Experiment (King Abdullah Street (Al- sha’b circle)) ,App(4,5,6) Third Experiment (Ata Ali ( Wadi al Seer Street)),App(7,8,9) Fourth Experiment (King Abdullah (Airport Street)) ,App(10)

12. WEKA Tool

13. Performance Matrices   Predicted Normal Congested Actual Normal True Positive (TP) False Negative (FN) Congested False Positive (FP) True Negative (TN)    

14. Results Accuracy Accuracy is a measure of how well a model performs across all targets.   RandomTree SVM Random Forest KNN multilayer perceptron Logistic Regression Zahran street (7th Circle) 97.7% 99.4% 97.5% 98.0% 94.5% 96.6% King Abdullah (Airport Street) 96.3% 99.1% 96.9% 98.5% 95.7% 97.1% Ata Ali (Wadi al Seer Street) 97.2% 99.6% 97.1% 98.3% 96.3% 96.6% King Abdullah Street (Al-sha’b circle) 97.2% 99.7% 96.2% 98.7% 95.4% 97.6%

15. Results Precision RandomTree SVM Random Forest KNN multilayer perceptron Logistic Regression Zahran street (7th Circle) 98.4% 99.4% 98.7% 98.6% 97.0% 99.2% King Abdullah (Airport Street) 95.2% 98.8% 97.2% 98.3% 96.9% 97.8% Ata Ali (Wadi al Seer Street) 100.0% 100.0% 100.0% 100.0% 99.9% 100.0% King Abdullah Street (Al- sha’b circle) 96.9% 99.8% 97.0% 99.0% 98.4% 97.7% The precision is determined by the ratio of correctly identified Positive (congested )to total Positive (either congested or uncongested).

16. Results Sensitivity The Sensitivity is computed as the ratio of correctly identified Positive values to the total number of Positive values Random Tree SVM Random Forest KNN multilayer perceptron Logistic Regression Zahran street (7th Circle) 98.4% 99.8% 97.8% 98.6% 95.3% 96.0% King Abdullah (Airport Street) 97.2% 99.3% 96.5% 98.1% 94.2% 96.3% Ata Ali (Wadi al Seer Street) 97.2% 99.6% 97.1% 98.3% 96.4% 96.6% King Abdullah Street (Al-sha’b circle) 97.6% 99.5% 95.4% 98.5% 92.5% 97.5%

17. Results F-measure RandomTree SVM Random Forest KNN multilayer perceptron Logistic Regression Zahran street (7th Circle) 98.4% 99.6% 98.3% 98.6% 96.2% 97.6% King Abdullah (Airport Street) 96.2% 99.1% 96.8% 98.2% 95.6% 97.0% Ata Ali (Wadi al Seer Street) 98.6% 99.8% 98.5% 99.1% 98.1% 98.3% King Abdullah Street (Al-sha’b circle) 97.2% 99.7% 96.2% 98.8% 95.3% 97.6% The F-measure is determined by taking the harmonic mean of precision and sensitivity and assigning equal weighting to each.

18. Conclusion Object 1: To collect traffic data at the 8th Circle approaches: -Datasets were collected , divided and per-process into four datasets for each approach. -The dataset was prepared by cleaning the data (removing noises and missing values) and making it suitable for ML classifiers which increases the accuracy of prediction. Object 2: To predicate the traffic congestion at The 8th Circle approaches using The Logistic Regression, KNN, DT, Random Forest, SVM, and MLP and ML classifiers. four Experiments was conducted using The Logistic Regression, KNN, DT, Random Forest, SVM, and MLP classifiers.

19. Conclusion Object 3: To determine the best ML classifiers for predicate traffic congestion. The obtained results show that the SVM classifier is the best among other classifiers to predict traffic congestion .

20. Future Work This work has focused on the traffic congestion prediction on the 8th Circle approaches of Amman city. the 8th Circle is one of eight series of traffic circles that run east-west through Amman city. In future work, we look to predicate the traffic congestion at each traffic circle, using ML classifiers.

21. Thank You 