Download Road Accident Severity Prediction: A Hybrid Machine Learning Approach and more Schemes and Mind Maps Computer Communication Systems in PDF only on Docsity!
i
Enhanced Decision Support System for Road
Accident Severity Forecast
A Project Report submitted in partial fulfilment of the
requirements for the award of the Degree of
BACHELOR OF TECHNOLOGY
IN
COMPUTER SCIENCE AND ENGINEERING with
Specialization
ARTIFICIAL INTELLIGENCE & MACHINE LEARNING
Submitted by
P.JYOSHNA 20U41A
M.M.PRATHYUSHA 20U41A
P.RAVI KIRAN REDDY 20U41A
K.MOHAN KUMAR 20U41A
Under the Esteemed guidance of
Mrs.REDDI SWAPNA
Assistant Professor, Department of CSM & CSD
DADI INSTITUTE OF ENGINEERING & TECHNOLOGY (AN AUTONOMOUS INSTITUTE) (Approved by A.I.C.T.E., New Delhi & Permanently Affiliated to JNTU GV) Accredited by NAAC with ‘A’ Grade and Inclusion u/s 2(f) & 12(B) of UGC Act An ISO 9001:2015, ISO 14001:2015 & ISO 45001:2018 Certified Institute. NH-16, Anakapalle – 531002, Visakhapatnam, A.P.
ii DADI INSTITUTE OF ENGINEERING & TECHNOLOGY (AN AUTONOMOUS INSTITUTE) (Approved by A.I.C.T.E., New Delhi & Permanently Affiliated to JNTU GV) Accredited by NAAC with ‘A’ Grade and Inclusion u/s 2(f) & 12(B) of UGC Act An ISO 9001:2015, ISO 14001:2015 & ISO 45001:2018 Certified Institute. NH-16, Anakapalle – 531002, Visakhapatnam, A.P.
CERTIFICATE
This is to certify that the project report entitled “ Enhanced Decision Support System for Road Accident Severity Forecast ” is being submittedbyP. JYOSHNA(20U41A4215),M.M. PRATHYUSHA (20U41A4227), P. RAVIKIRAN REDDY(20U41A4229),K. MOHAN KUMAR (20U41A4232). In partial fulfilment of the requirements for award of the Degree of Bachelor of Technology in COMPUTER SCIENCE AND ENGINEERING with Specialization ARTIFICIAL INTELLIGENCE & MACHINE LEARNING , Dadi Institute of Engineering & Technology(A), Anakapalle affiliated to JNTUGV, accredited by NAAC with 'A' grade is a record of bonafide work carried out by them under my guidance and supervision.
EXTERNAL EXAMINER
R. SWAPNA
(ASSISTANT PROFESSOR)
(PROJECT GUIDE)
A. VENKATESWARA RAO
(ASSOCIATE PROFESSOR)
(HEAD OF THE DEPARTMENT)
iv
DECLARATION
We hereby declare that the project entitled “Enhanced Decision Support System for Road Accident Severity Forecast” is submitted in partial fulfilment of the requirements for the award of Bachelor of Technology in Computer Science and Engineering with Specialization Artificial Intelligence & Machine Learning under esteemed supervision of Mrs. R. SWAPNA, Assistant Professor. This is a record of work carried out by us and results embodied in this project report have not been submitted to any other university for the award of any Degree. PROJECT ASSOCIATES
P.JYOSHNA 20U41A
M.M.PRATHYUSHA 20U41A
P.RAVIKIRANREDY 20U41A
K.MOHAN KUMAR 20U41A
v
ABSTRACT
Road accidents constitute a substantial threat to global public safety, necessitating comprehensive efforts to reduce their severity. The suggested Decision Support System combines powerful machine learning algorithms, historical accident data, real-time traffic and weather information, and geographical elements to effectively anticipate the severity of road accidents. The Decision Support System provides actionable information for policymakers, law enforcement agencies, and emergency responders by utilizing a vast dataset and cutting-edge predictive models. Through comprehensive evaluation and validation utilizing real-world accident data, the effectiveness and dependability of the proposed Decision support systems are displayed, emphasizing their potential to improve road safety initiatives and save lives. The research findings help to expand decision support systems in the field of road safety, paving the door for proactive measures to reduce accident severity and improve overall traffic management tactics. Keywords : Road safety decision-making, road accident prediction, statistical parametric models, machine learning.
vii
- INTRODUCTION LIST OF CONTENTS
- LITERATURE SURVEY
- SYSTEM ANALYSIS
- 3.1 Existing System
- 3.1 Proposed System
- SYSTEM REQUIREMENTS
- SYSTEM STUDY
- 5.1 Feasibility Study
- 5.1.1 Economical Feasibility
- 5.1.2 Technical Feasibility
- 5.1.3 Social Feasibility
- SYSTEM DESIGN
- 6.1 System Flow Chart
- 6.2 UML Diagrams
- 6.2.1 Class Diagram
- 6.2.2 Sequence Diagram
- 6.2.3 Use Case Diagram
- 6.2.4 Activity Diagram
- IMPLEMENTATION
- 7.1 Introduction
- 7.2 Modules
- 7.3 Algorithms
- 7.4 Coding
- 5 Results
- 8.1 Introduction 8. SYSTEM TESTING
- 8.2 Testing Methodologies
- 8.2.1 Unit Testing
- 8.2.2 Integration Testing
- 8.2.3 Functional Testing
- 8.2.4 System Testing
- 8.2.5 White Box Testing
- 8.2.6 Black Box Testing
- 8.2.7 Acceptance Testing
- 8.3 Test Case
- CONCLUSION&FUTURE SCOPE
- REFERENCES
viii
2
CHAPTER 1
INTRODUCTION
Intelligent Road Traffic Incident Severity Forecast, often referred to as IRTISF, is a cutting-edge technological solution that has revolutionized the way we approach road safety and traffic management. In an era marked by the rapid growth of urban centers and the increasing complexity of transportation networks, the need for accurate and timely predictions of traffic incident severity has never been greater. IRTISF leverages advanced data analytics, machine learning algorithms, and real-time information from various sources such as traffic cameras, sensors, and historical incident data to provide proactive insights into the potential outcomes of traffic incidents. It goes beyond merely identifying accidents and congestion; it empowers transportation authorities, emergency responders, and commuters with the ability to anticipate the severity of these incidents. Road accidents are a significant public health concern worldwide, causing millions of deaths and injuries annually. Despite considerable efforts to improve road safety, the severity of accidents remains a persistent challenge. Forecasting the severity of road accidents is crucial for implementing effective preventive measures, allocating resources efficiently, and reducing the overall impact on society understanding the factors contributing to road accident severity is complex, as it involves a myriad of variables, including driverbehavior, road conditions, vehicle characteristics, and environmental factors. By analyzing historical accident data and employing advanced predictive modeling techniques, researchers and policymakers aim to develop robust forecasting models capable of accurately predicting the severity of road accidents. One of the primary objectives of forecasting road accident severity is to identify high-risk areas and vulnerable populations. By leveraging data-driven approaches, such as machine learning algorithms and
3 spatial analysis techniques, researchers can pinpoint geographical locations and demographic groups with elevated accident severity rates. This information is invaluable for implementing targeted interventions and prioritizing resources to mitigate the impact of accidents on communities .moreover, forecasting road accident severity enables stakeholders to anticipate trends and patterns over time, facilitating proactive decision-making and policy development. By identifying emerging risk factors and evolving trends, policymakers can implement preemptive measures to address potential threats to road safety before they escalate into larger-scale issues. Furthermore, accurate forecasting of road accident severity can inform the development of innovative technologies and interventions aimed at preventing accidents and minimizing their consequences. From advanced driver assistance systems and vehicle automation technologies to improved road infrastructure and enhanced emergency response protocols, forecasting models play a crucial role in guiding the design and implementation of effective interventions. In conclusion, forecasting road accident severity is a multifaceted endeavor that requires a comprehensive understanding of the underlying factors and dynamic nature of road safety. By leveraging advanced analytics and innovative technologies, stakeholders can develop predictive models capable of accurately anticipating accident severity and guiding proactive interventions. Addressing the challenges associated with forecasting road accident severity is essential for enhancing road safety and minimizing the societal impact of accidents on communities worldwide Road traffic accident (RTA) is churning the world with Killing thousands and bringing demolition of property In a day withoutdiscrimination but did not give much Attention to mitigate the severity. However, it is one of The life-threatening incidents in the world cause of death And property damage. Identifying the primary road traffic Accident factors will help to provide an appropriate
5 and random Forest algorithm Proposed to predict target specific road accident severity. The proposed approach compared with individual classifiers to measure the performance of the developed model. Accuracy, precision, specificity, and recall used to compare The new approach and conventionaltechniques (SVM,KNN, LR, and RF). The new approach composed of the following phases: (I) removing disturbing noise and filling Missing data using mean for numericvariables and mode for the categorical variable, (II) splitting the dataset into Training and test dataset, (III) creating new feature using Clustering, (IV) training classifiers, (V) finally evaluating the Performance of individual classifiers. Moreover, the proposed approach compared with a deep neural network to evaluate further with another state of the art classifier Techniques. The evaluation outcome showed the proposed Better performs than other classifiers based on classification and performance metrics.
6
LITERATURE SURVEY
8 SVM with RBF kernel gave better accuracy (94%) than MLP (64%). The study showed driving with high speed after drunk was the main reason for accident occurrence. Wahab and Jiang carried out crash accidents on Ghana dataset using MLP, PART, and Simple CART intending to evaluate classifiers and to identify the major factors for motorcycle crash. Authors used Weka tools to compare and analyze datasets and Info Gain Attribute Eval applied to see the most influential variable for motorcycle crash in Ghana. As a result simple CART model showed better accuracy than other classification models. Kumar et al. implemented k-means and Association Rule data mining approaches to identify the frequency of accident severity locations and to extract hidden information. From the total 158 locations; 87 of them were selected after removing accident location frequency count less than 20. Then k-means were applied to cluster into three groups, Number of clusters are determined by gap statistics. To get rules, they used minimum support of 5 percent. As a result, curved and slop on the hilly surface were revealed as accident prone locations. Authors worked on the FARS data-set using data mining techniques to combat death and injury severity during 2007.After prepossessing the study applied clustering Association rule and Nave Bayes to get trends of fatal accidents in the USA. The study explained and identify human and collusion types were the main cause of the fatality rate. Other studies conducted using clustering and classification techniques to predict an accurate model in Iran. The research mainly focused on combining k means clustering with self-organizing maps to get better classification accuracy than ANN and ANFIS. The author’s preference model better performs than the single classifiers. Al Mam look et al. used AdaBoost, Nave Bayes, Logistic Regression and random forest to get determinant factors and to identify high risky highways for Michigan traffic Agencies. Performance measurement ROC, AUC, Precision and recall and F1-score were applied to evaluate models. The Study showed random forest
9 outperforms other classifiers with an accuracy of 75.5%. Tiwari et al. conducted a data mining approach to analyze causality class traffic accidents. The authors implemented clustering like K-modes and SOM and classification techniques like NB, DT, and SVM. As a result, better accuracy was presented on cluster dataset over classification. The existing study on road accident severity in Ethiopia see. These stated works concerned mostly on road accident analysis and pedestrian severity in Ethiopia using Statistical methods. on the other hand some studies employed a data mining techniques (Decision Tree and MLP) on Weka tool focusing mainly on driver responsibility. Another study employed J48 and PART a data mining algorithm on driver and vehicle information considering as a major risk in accident severity on Weka tool. Other related work in the country, Beshah studied to identify the key road way related variables for accident severity in Ethiopia. Authors used a data mining approach (Decision Tree, Naive Bayes and KNN) to develop a decision rule to improve road safety. Their focus has been analyzing driver and pedestrian crashes without giving more attention to the influence of machine learning accuracy for better identification of major risks influencing in road accident in Ethiopia. At this time there is a great need for increasing road safety prevention study due to the growth of crashes. There is still room for improvement in the prediction accuracy of RTA in the case of Ethiopia to improve prediction accuracy in road safety. Therefore ,we tried to develop a new hybrid approach to classify road accident severity by combining or collaborating clustering with classification, which will give remarkable classification results in road accident prediction. Clustering minimizes the sample dataset in the cooperation. Classification predicts road traffic severity. In this vein, clustering provides indirect cooperation for classification to extract hidden information from the training set to improve classifier performance.
11
CHAPTER 3
SYSTEM ANALYSIS
3.1 Existing system
The existing system for assessing road severity typically involves several factors, including road condition, traffic volume, speed limits, and accident data. Roads are often categorized into different classes or levels of severity based on these factors. This classification helps transportation authorities prioritize maintenance and safety improvements. The Road accidents data analysis to spot the main factors surrounded a road and traffic accident. There are a lot of Data Mining Algorithms which are available to find out the association between independent variables in a hug data. The most popular and commonly used algorithm is association rule mining. This can be used to detect the significant associations between the data stored in the large database. Predictive and FP growth algorithm are the most common association rule mining methods which are used. The results obtained from these data mining approach can help understand the most significant factor or often repeating patterns. The generated patterns are identifies the most dangerous roads in terms of road accidents and necessary measures can be taken to avoid accidents in those roads. Data analysis has been widely used in research with literature there of employing different graphical representations and statistical analyses, to perform preliminary investigation on datasets.
3.1.1. Drawbacks
● Accuracy is Low. ● Efficiency also low
12
3.2 Proposed system
A machine learning-based approach to predict road accident severity based on various contributing factors and differing a accident zones based on the accidents occur in that place. Several machine learning algorithms, including Decision Trees, Random Forest, Extra Trees and Gradient Boosting, are implemented and evaluated using cross- validation to identify the most effective model for the task .By predicting accident severity, emergency services can be better prepared to handle potential incidents, potentially reducing the number of fatalities and minimizing the impact of accidents on road infrastructure and public health.
3.2.1 Advantages
● Accuracy is Very high. ● Efficiency increased
