International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 13 Issue: 04 | Apr 2026 www.irjet.net p-ISSN: 2395-0072
AI-Based Fatality Risk Hotspot and Cause-Shift Prediction System Using Accidental and Natural Death Data
Vaibhav M 1 , Sathvik M V 2 , Vamshi Krishna P S 3, Varun S4, Dr. Savitha G5
1 Student, Department of Computer Science Engineering R V Institute of Technology and Management, Bengaluru, India
2 Student, Department of Computer Science Engineering R V Institute of Technology and Management, Bengaluru, India
3 Student, Department of Computer Science Engineering R V Institute of Technology and Management, Bengaluru, India
4 Student, Department of Computer Science Engineering R V Institute of Technology and Management, Bengaluru, India
5 Associate Professor, Department of Computer Science Engineering R V Institute of Technology and Management, Bengaluru– 560076, Karnataka, India ***
Abstract - The increasing number of accidental and natural deaths reported each year highlights the need for intelligentsystemsthatcansupportearlyrisk detectionand preventive planning. Although large volumes of mortality data are collected by government agencies, these datasets are commonly used only for historical reporting and statistical summaries. This paper proposes an AI-Based FatalityRiskHotspotandCause-ShiftPredictionSystemthat transforms historical death records into predictive insights. The proposed framework applies machine learning techniques to estimate future fatality counts, classify dominant causes of death, and identify high-risk regions through clustering. Random Forest Regression is used for trend forecasting, Random Forest Classification is applied for cause prediction, and K-Means Clustering is used to group regions based on fatality severity. A web-based dashboardisdesignedtopresentresultsthroughinteractive charts and region-wise analysis. The system enables authorities to move from reactive reporting to proactive decision-making by supporting public safety planning, healthcare preparedness, and targeted interventions. Experimental observations indicate that the integrated approach improves interpretability and provides useful predictionsforpolicydevelopmentandresourceallocation.
Key Words: Artificial Intelligence, Fatality Prediction, Machine Learning, Risk Hotspots, Cause Analysis, Public Safety, Clustering
1.INTRODUCTION
Accidental and natural deaths remain a serious public concern, causing significant social and economic impact everyyearacrossdifferentregionsofIndia.Largevolumes of mortality data are collected by government agencies andstatisticaldepartments,coveringfactorssuchasstate, year, and cause of death [1], [5]. However, these records are often used only for annual reports and historical summaries, which limits their value for future planning.
Traditional analytical methods mainly describe past trendsandareless effective inpredictingupcomingrisks, identifying vulnerable regions, or detecting shifts in fatality patterns [3]. This creates a strong need for intelligent,data-driven,andcost-effectiveapproachesthat cansupportproactivedecision-making.
Artificial Intelligence, which combines computational models with data analysis techniques, has emerged as a powerful solution for extracting meaningful insights from structured datasets [5]. By leveraging machine learning algorithms and historical mortality records, AI systems canuncoverhiddenpatterns,forecastfuturefatalities,and classify dominant causes of death with higher efficiency than manual methods [2]. These capabilities allow administrators to recognize risk hotspots, allocate healthcare resources effectively, and plan preventive measuresbeforelossesincrease.
This paper aims to examine the role of AI and machine learning in transforming mortality analysis into a predictive framework. The following sections discuss the use of regression models for fatality forecasting, classification techniques for cause-shift detection, clustering methods for hotspot identification, and interactivedashboardsforsupportingpolicydecisionsand publicsafetymanagement[4].
2. LITERATURE REVIEW
The application of Artificial Intelligence in public data analysis has significantly changed the way large-scale datasets are interpreted and utilized. Instead of relying only on manual reports and descriptive statistics, modern systems use predictive models to generate future insights fromhistoricalrecords[1].Inmortalityanalysis,AIenables authorities to shift from reactive observation to proactive planning by forecasting risks, identifying vulnerable regions, and detecting changes in death patterns. The
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 13 Issue: 04 | Apr 2026 www.irjet.net p-ISSN: 2395-0072
major areas of research include fatality trend prediction, hotspot identification, cause classification, and intelligent decision-supportsystems.
2.1 Fatality Trend Prediction:
Forecasting future fatality counts is one of the most important tasks in mortality analytics. Traditional statistical methods such as linear regression and timeseries models were initiallyused to estimate yearly death trends. While these methods perform adequately on simple datasets, they often struggle to capture non-linear relationships and interactions between multiple variables suchasregion,year,andcausecategory[2].
Machinelearning models, especially ensemble methods suchasRandomForestRegression,haveshownimproved performance in trend forecasting because they can learn complex patterns from historical data. These models are robust to noise, reduce overfitting, and provide more reliable predictions across diverse regions [3]. In healthcare and population studies, regression-based forecasting has been successfully applied to estimate diseaseburden,accidentfrequency,andresourcedemand
2.2 Hotspot Identification Using Clustering:
Another important area of research focuses on identifying high-risk locations where fatalities are consistently higher than average. Clustering algorithms are widely used for this purpose because they automatically group regions with similar fatality characteristicswithoutrequiringpredefinedlabels[4].
K-Means Clustering is one of the most commonly used techniques due to its simplicity and efficiency. By analyzingdeathcountsandrelatedfeatures,thealgorithm can divide regions into categories such as low-risk, medium-risk, and high-risk zones. Such grouping helps administrators prioritize interventions and allocate emergency resources where they are most needed [5]. Similar clustering approaches have also been used in crime mapping, epidemic surveillance, and disaster managementsystems.
2.3 Cause Classification and Cause-Shift Analysis:
Several studies have explored classification techniques for determining dominant categories within public datasets. In mortality analysis, classification models can predict whether accidental or natural causes are more likely to dominate in a specific region or future period. This is valuable because the preventive actions required for accidents differ significantly from those needed for health-relatednaturaldeaths[6].
Algorithms such as Decision Trees, Support Vector Machines,andRandomForestClassificationarefrequently usedforcategoryprediction.Amongthem,RandomForest performs effectively on structured datasets because it combines multiple decision trees and improves classification accuracy. Cause-shift analysis extends this idea by identifying transitions in dominant causes over time,helpingauthoritiesdetectemergingrisksearly[7]
2.4 Computational Drug Repositioning:
Recent research emphasizes the importance of integrating predictive models with visualization tools. Standalone predictions are less useful if decision-makers cannot interpret results quickly. Therefore, web dashboards and interactive systems are increasingly combined with AI models to present charts, regional comparisons,andreal-timerecommendations[8].
Frameworks such as Streamlit and Flask have enabled the development of lightweight decision-support platforms where users can select a region, enter future years, and instantly view predictions. These systems improveaccessibilityfornon-technical usersand make AI outputsmorepracticalforgovernanceandplanning
2.5 Research Gap:
Although previous studies have addressed forecasting, clustering, or classification separately, limited work combines all three capabilities in one unified framework for accidental and natural death data. Most existing systems focus only on historical reporting or a single analytical method. This paper addresses that gap by proposing an integrated AI system that combines fatality prediction, hotspot detection, cause-shift classification, anddashboard-basedvisualizationinasingleplatform[9].
3. OBJECTIVES
The primary objective of this research is to develop an intelligent analytical system that can transform historical mortality records into predictive insights for better planning and decision-making. The specific objectives of theproposedworkareasfollows:
To analyze accidental and natural death data collectedacrossdifferentregionsandyears.
To predict future fatality counts using machinelearningregressiontechniques.
To identify high-risk regions by applying clusteringalgorithmsonmortalitypatterns.
To classify dominant causes of death as accidental or natural using supervised learningmodels.
Todetectshiftsinfatalitycausesovertimefor earlyinterventionandpolicyplanning.
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 13 Issue: 04 | Apr 2026 www.irjet.net p-ISSN: 2395-0072
To design a web-based dashboard for interactive visualization of predictions and trends.
To support government agencies, healthcare departments, and safety authorities with data-drivenrecommendations.
4. DISCUSSION
4.1
Challenges & Limitations
Despite the growing importance of AI in mortality prediction and hotspot analysis, several limitations still affect the reliability and practical performance of these systems. These challenges arise in key areas such as fatality forecasting, cause classification, clustering accuracy, and data integration, often leading to reduced predictionqualityandhigherimplementationcomplexity.
In fatality forecasting, regression models may produce inaccurateestimateswhenhistoricaldatacontainssudden anomalies, incomplete records, or irregular reporting patterns [13]. The prediction process is generally expressed as , where includes variables such as region, year, and previous fatalities, while represents the expected future death count. If the training data does not properly capture changing real-world conditions, the estimated value of may deviate significantly from actual outcomes.
Similarly, classification models used for cause-shift prediction may face errors when accidental and natural deathpatternsoverlapsignificantly.Ifthetrainingdataset isimbalancedorcontainsnon-representativesamples,the classifiermaybecome biasedtoward majorityclassesand perform poorly on minority patterns [14], [20]. This can leadtoincorrectpredictionsregardingwhetheraccidental ornaturalcauseswilldominateinfutureperiods.
Hotspot detection using clustering techniques also presents challenges. Algorithms such as K-Means depend stronglyontheselectednumberofclustersandtheinitial centroidpositions.Theclusteringobjectiveistominimize the total within-cluster variance using the function
∑ , where denotes a cluster and represents its centroid. Poor initialization or unsuitable values of may lead to unstable grouping results or incorrect separation of high-risk and low-risk regions[15].
Another major limitation is data integration. Mortality datasets collected from multiple sources may contain inconsistent labels, missing values, duplicate entries, and varying reporting standards. These issues require extensive preprocessing before model training and may reducethedependabilityofthefinalpredictions[15].
Overall,thesechallengeshighlighttheneedforbalanced datasets, robust preprocessing, continuous model validation, and the inclusion of external factors such as demographics or environmental conditions. Addressing these limitations is essential for transforming AI-based mortality systems into reliable tools for real-world decision-making.
4.2 Emerging Role of AI and ML
Artificial Intelligence (AI) and Machine Learning (ML) are becoming powerful solutions for many limitations found in traditional mortality analysis systems, such as forecasting errors, biased classification results, weak hotspot detection, and difficulties in handling incomplete or heterogeneous datasets [13]. Conventional statistical methods often struggle with large and dynamic datasets, whereasAImodelscanprocesscomplexrelationshipsand generate more accurate predictions from historical records.
Particularly in fatality forecasting and cause-shift prediction, AI techniques improve speed, scalability, and predictive performance by learning hidden patterns across multiple years and regions. Advanced learning models can analyze structured mortality data, detect subtle trends, and adapt to changing patterns more effectivelythanrule-basedsystems[16],[17].
By using ensemble learning and multitask frameworks, AImodelscansimultaneouslypredictfuturedeathcounts, classify dominant causes, and estimate risk levels while capturing relationships between different variables. This improves model robustness on imbalanced datasets and reduces errors caused by noisy or incomplete records [14], [19]. Higher predictive accuracy enables early identification of high-risk regions and supports timely policyinterventions.
Automated Machine Learning (AutoML) further simplifies model development by selecting algorithms, tuning hyperparameters, and optimizing performance with minimal manual effort. This increases accessibility for non-expert users and improves the efficiency of deploying predictive systems for public administration [14], [19]. Deep learning approaches can also enhance pattern recognition when larger datasets become available, especially for long-term trend forecasting and anomalydetection.
Overall, AI and ML not only address the weaknesses of traditional analytical methods but also create opportunities for integrated intelligent systems that combine forecasting, classification, clustering, and realtime visualization. Such advancements can significantly strengthen data-driven governance, healthcare preparedness,andpublicsafetyplanning[16],[17].
Volume: 13 Issue: 04 | Apr 2026 www.irjet.net
4.3 Impact of Personalized and Region-Specific Decision Support
By enabling highly customized planning strategies, AIdriven mortality systems have the potential to transform publicadministration from a one-size-fits-all model intoa region-specific decision framework [18]. Instead of applying the same preventive measures everywhere, authorities can use localized predictions to design actions basedontheriskprofileofeachregion.
By combining multiple data sources such as mortality records, demographic indicators, healthcare availability, and environmental conditions, intelligent systems can forecast region-specific fatality risks, identify vulnerable populations,andoptimizeinterventionstrategies.Thiscan improve efficiency while reducing unnecessary expenditureanddelayedresponses[15],[18].
For example, urban areas with rising accidental deaths may require stronger traffic monitoring, industrial safety audits, or emergency response infrastructure. Regions showing increasing natural deaths may benefit from improved hospitals, disease surveillance, and awareness programs. Such targeted planning ensures that resources areallocatedwheretheycreatethegreatestimpact.
In the future, these systems can be expanded with AIdrivenrecommendationsthatautomaticallysuggestpolicy actions based on predicted risks. Integration with realtime data streams, GIS mapping, and population analytics can further enhance responsiveness and improve administrativeoutcomes[18].
Overall, the broader impact includes cost reduction through smarter resource allocation, better preparedness through early warnings, and improved governance through evidence-based planning. With responsible data use and continuous model refinement, AI-based mortality analytics can become a cornerstone of safer and more efficientpublicsystemsinthecomingyears[18].
4.4 Applications and Practical Significance
TheproposedAI-BasedFatalityRiskHotspotandCauseShift Prediction System has wide practical applications in governance,healthcare,andpublicsafety.Bytransforming historical mortality data into predictive insights, the system helps institutions make informed decisions rather thanrelyingonlyonpastreports.
Onemajorapplicationisingovernmentpolicyplanning. Authorities can identify regions with rising fatality trends and introduce preventive regulations, awareness campaigns, or infrastructure improvements before the situation worsens [21]. Predictive analytics can also support budget planning by directing resources toward high-prioritydistricts.
Inpublicsafetymanagement,thesystemcanbeusedto monitor accident-prone areas and implement corrective measures such as traffic control, industrial inspections, andemergencypreparednessprograms.Ifcertainregions consistently appear as high-risk hotspots, targeted interventions can be carried out to reduce future losses [22].
Another important application is healthcare resource allocation. Forecasts of natural deaths can help hospitals and health departments prepare beds, staff, medicines, and emergency facilities in advance. Early planning improvesresponseefficiencyandreducespressureduring peakdemandperiods[23].
The proposed model can also contribute to smart city administrationbyintegratingwithdigitaldashboards,GIS platforms, and urban planning systems. Risk maps and future forecasts can guide safer city design, road management,anddisasterreadinessstrategies[24].
Overall, the system serves as a practical decisionsupport tool that converts raw data into meaningful actionsforsociety.
4.5 Future Scope
Although the current system demonstrates the usefulness of AI in mortality analytics, several enhancements can further improve its real-world value. One important future direction is the integration of realtimedatasourcessuchashospitalrecords,trafficsystems, weather feeds, and emergency response data. Real-time updateswouldmakepredictionsmoredynamicandtimely [25].
The use of deep learning models can also be explored forcapturinghighlycomplextemporalpatternsandlargescale datasets. Techniques such as Long Short-Term Memory (LSTM) networks and hybrid neural models may improvelong-termforecastingaccuracy[26].
Another valuable enhancement is the addition of GISbasedheatmapsandgeospatialintelligence.Visualhotspot mapscanhelpadministratorsinstantlyidentifyvulnerable areasandmonitorregionalchangesmoreeffectively[27].
Future systems may also include population, economic, and environmental indicators to create richer predictive models. Factors such as population density, healthcare access, pollution levels, and climate conditions can improveriskestimationaccuracy[28].
Finally, deploying the platform as a mobile and cloudbased application can increase accessibility for field officers,disastermanagementteams,andlocalauthorities. With continuous improvement, the proposed framework
International Research
Volume: 13 Issue: 04 | Apr 2026 www.irjet.net p-ISSN: 2395-0072
can evolve into a national-scale intelligent safety monitoringsystem.
4.6 Performance Evaluation and Model Validation
To ensure that the proposed system is reliable for realworld use, proper model evaluation is essential. Different machine learning tasks require different performance metrics. For fatality count prediction, regression models arecommonlyassessedusingMeanAbsoluteError(MAE), Root Mean Square Error (RMSE), and the coefficient of determination ( ). These metrics measure how closely predicted values match actual death counts. Lower MAE and RMSE values indicate better forecasting accuracy, while a higher value reflects stronger explanatory power[29].
For cause-shift classification, performance can be evaluated using Accuracy, Precision, Recall, and F1-Score. Accuracy measures the overall correctness of predictions, while Precision indicates how many predicted positive cases were correct. Recall evaluates how effectively the model identifies relevant cases, and F1-Score balances both Precision and Recall. These metrics are especially useful when accidental and natural death classes are unevenlydistributed[30].
Clustering models are often validated using measures such as the Silhouette Score and inertia values. The Silhouette Score indicates how well data points fit within their assigned clusters compared to other clusters. A higherscoresuggestsclearerseparationbetweenlow-risk, medium-risk, and high-risk regions. Inertia measures the compactnessofclustersandisminimizedduringK-Means optimization[31].
Cross-validation techniques can further improve trustworthiness by testing the model on multiple traintest splits instead of a single dataset partition. This reduces the risk of overfitting and ensures that performanceremainsstableacrossdifferentsamples[32].
4.7 Ethical and Social Considerations
While AI-based mortality prediction offers many benefits, ethical considerations must also be addressed. Public datasets may contain sensitive demographic or health-related information, making data privacy an importantconcern.Properanonymization,securestorage, and controlled access mechanisms are necessary before deployingsuchsystems[33].
Another issue is algorithmic bias. If historical data reflects reporting gaps or unequal representation of regions, the model may produce unfair predictions that disadvantage certain communities. Regular auditing,
balanced datasets, and transparent evaluation processes arerequiredtoreducesuchrisks[34].
Thesystemshouldalsobeviewedasadecision-support tool ratherthana replacementforhumanjudgment.Final policy decisions must involve domain experts, healthcare officials,andadministratorswhocaninterpretpredictions within real-world contexts. Responsible use of AI ensures that technology supports fairness, accountability, and publicwelfare.
5. CONCLUSIONS
The increasing availability of mortality data creates a valuable opportunity to improve public safety and administrative planning through intelligent analytics. However, traditional reporting systems mainly describe past events and provide limited support for forecasting future risks. This research addresses that gap by proposing an AI-Based Fatality Risk Hotspot and CauseShiftPredictionSystemusingaccidentalandnaturaldeath data.
The proposed framework combines multiple machine learning techniques to perform different analytical tasks within a single platform. Random Forest Regression is used to estimate future fatality counts, Random Forest Classification is applied to predict dominant causes of death,andK-MeansClusteringisusedtoidentifylow-risk and high-risk regions. By integrating these methods with an interactive dashboard, the system transforms raw historicalrecordsintomeaningfulandaccessibleinsights.
The study demonstrates that AI can significantly improve the usefulness of mortality datasets by enabling proactiveratherthanreactivedecision-making.Predictive insights can help governments strengthen safety measures, improve healthcare readiness, allocate resources efficiently, and plan targeted interventions for vulnerableregions.
Althoughchallengessuchasdataquality,imbalance,and changing external conditions still exist, continuous model improvement and richer datasets can further enhance system reliability. Future integration with real-time data, geospatial tools, and advanced deep learning models can makesuchsystemsevenmoreeffective.
In conclusion, the proposed work highlights how Artificial Intelligence can play an important role in transforming mortality analysis into a smart decisionsupportframework.Withresponsibleimplementationand ongoing refinement, AI-based predictive systems can contributetosafercommunities,strongergovernance,and betterlong-termplanning.
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 13 Issue: 04 | Apr 2026 www.irjet.net p-ISSN: 2395-0072
REFERENCES
[1] L. Breiman, “Random Forests,” Machine Learning, vol. 45, no. 1, pp. 5–32, Oct. 2001, doi: 10.1023/A:1010933404324.
[2] J. MacQueen, “Some Methods for Classification and Analysis of Multivariate Observations,” in Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability, Berkeley, CA, USA, vol. 1, pp. 281–297, 1967.
[3] F. Pedregosa et al., “Scikit-learn: Machine Learning in Python,”JournalofMachineLearningResearch,vol.12,pp. 2825–2830,Oct.2011.
[4] J. Han, M. Kamber, and J. Pei, Data Mining: Concepts and Techniques, 3rd ed. Waltham, MA, USA: Morgan Kaufmann,2011.
[5] World Health Organization, “WHO Mortality Database: Global Health Estimates and Mortality Trends,” Geneva, Switzerland,2024.
[6]GovernmentofIndia,MinistryofHomeAffairs,“Annual Statistical Report on Accidental and Natural Deaths in India,”NewDelhi,India,2024.
[7] IndiaAI, “Accidental and Natural Death Statistics DatasetRepository,”GovernmentofIndiaDigitalPlatform, 2025.
[8] T. Chen and C. Guestrin, “XGBoost: A Scalable Tree Boosting System,” in Proceedings of the 22nd ACM SIGKDDInternationalConferenceonKnowledgeDiscovery and Data Mining, San Francisco, CA, USA, pp. 785–794, Aug.2016,doi:10.1145/2939672.2939785.
[9] G. James, D. Witten, T. Hastie, and R. Tibshirani, An IntroductiontoStatisticalLearningwithApplicationsinR, 2nded.NewYork,NY,USA:Springer,2021.
[10]M.KuhnandK.Johnson,AppliedPredictiveModeling. NewYork,NY,USA:Springer,2013.
[11]S.Raschka andV.Mirjalili, PythonMachineLearning: Machine Learning and Deep Learning with Python, ScikitLearn, and TensorFlow, 3rd ed. Birmingham, U.K.: Packt Publishing,2019.
[12] I. Goodfellow, Y. Bengio, and A. Courville, Deep Learning.Cambridge,MA,USA:MITPress,2016.
[13] A. Géron, Hands-On Machine Learning with ScikitLearn, Keras, and TensorFlow, 3rd ed. Sebastopol, CA, USA:O’ReillyMedia,2022.
[14]T.Hastie,R.Tibshirani,andJ.Friedman,TheElements of Statistical Learning: Data Mining, Inference, and Prediction,2nded.NewYork,NY,USA:Springer,2009.
[15] P. Tan, M. Steinbach, and V. Kumar, Introduction to DataMining,2nded.Boston,MA,USA:Pearson,2018.
[16] A. K. Jain, M. N. Murty, and P. J. Flynn, “Data Clustering: A Review,” ACM Computing Surveys, vol. 31, no. 3, pp. 264–323, Sept. 1999, doi: 10.1145/331499.331504.
[17] N. Bhatia and V. Dixit, “Survey of Clustering Algorithms for Big Data: Taxonomy and Empirical Analysis,” International Journal of Computer Applications, vol.79,no.2,pp.15–22,Oct.2013.
[18]R.Kohavi,“AStudyofCross-ValidationandBootstrap for Accuracy Estimation and Model Selection,” in Proceedings of the 14th International Joint Conference on Artificial Intelligence (IJCAI), Montreal, Canada, pp. 1137–1143,1995.
[19] D. M. W. Powers, “Evaluation: From Precision, Recall and F-Measure to ROC, Informedness, Markedness and Correlation,” Journal of Machine Learning Technologies, vol.2,no.1,pp.37–63,2011.
[20] P. J. Rousseeuw, “Silhouettes: A Graphical Aid to the Interpretation and Validation of Cluster Analysis,” Journal of Computational and Applied Mathematics, vol. 20, pp. 53–65,Nov.1987,doi:10.1016/0377-0427(87)90125-7.
[21] S. Hochreiter and J. Schmidhuber, “Long Short-Term Memory,” Neural Computation, vol. 9, no. 8, pp. 1735–1780,Nov.1997,doi:10.1162/neco.1997.9.8.1735.
[22] Y. LeCun, Y. Bengio, and G. Hinton, “Deep Learning,” Nature, vol. 521, pp. 436–444, May 2015, doi: 10.1038/nature14539.
[23] C. M. Bishop, Pattern Recognition and Machine Learning.NewYork,NY,USA:Springer,2006.
[24] S. Russell and P. Norvig, Artificial Intelligence: A Modern Approach, 4th ed. Hoboken, NJ, USA: Pearson, 2021.
[25] J. Brownlee, Machine Learning Mastery with Python: UnderstandYourData,CreateAccurateModels,andWork Projects End-to-End. Melbourne, Australia: Machine LearningMastery,2020.
[26] Ministry of Road Transport and Highways, Government of India, “Road Accidents in India – Annual Report,”NewDelhi,India,2024.
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 13 Issue: 04 | Apr 2026 www.irjet.net p-ISSN: 2395-0072
[27] National Crime Records Bureau, “Accidental Deaths and Suicides in India,” Ministry of Home Affairs, GovernmentofIndia,NewDelhi,India,2024.
[28] Ministry of Health and Family Welfare, Government of India, “National Health Statistics Report,” New Delhi, India,2024.
[29] Smart Cities Mission, Government of India, “Urban Safety, Risk Monitoring and Smart Planning Report,” New Delhi,India,2024.
[30] World Bank, “Population Density, Development Indicators and Urban Risk Statistics,” Washington, DC, USA,2024.
[31] European Commission, “Ethics Guidelines for Trustworthy Artificial Intelligence,” Brussels, Belgium, 2021.
[32] OECD, “Recommendation of the Council on Artificial Intelligence,”Paris,France,2023.
[33] United Nations, “Global Population and Mortality Trends Report,” Department of Economic and Social Affairs,NewYork,NY,USA,2024.
[34] D. C. Montgomery, E. A. Peck, and G. G. Vining, Introduction to Linear Regression Analysis, 5th ed. Hoboken,NJ,USA:Wiley,2012.
[35] J. Fox, Applied Regression Analysis and Generalized Linear Models, 3rd ed. Thousand Oaks, CA, USA: Sage Publications,2015.
[36]T.M.Mitchell,MachineLearning.NewYork,NY,USA: McGraw-Hill,1997.
[37]A.Ng,MachineLearningYearning:TechnicalStrategy forAIEngineers,DeepLearning.AI,2018.
[38] K. P. Murphy, Machine Learning: A Probabilistic Perspective.Cambridge,MA,USA:MITPress,2012.
[39] R. Szeliski, Computer Vision: Algorithms and Applications,2nded.Cham,Switzerland:Springer,2022.
[40] M. Wooldridge, An Introduction to MultiAgent Systems,3rded.Hoboken,NJ,USA:Wiley,2020.
[41] I. H. Witten, E. Frank, and M. A. Hall, Data Mining: Practical Machine Learning Tools and Techniques, 4th ed. Cambridge,MA,USA:MorganKaufmann,2016.
[42] A. Géron, “Practical Deep Learning for Structured Data Systems,” O’Reilly Media Technical Series, Sebastopol,CA,USA,2023.
[43]J.Leskovec,A.Rajaraman,andJ.D.Ullman, Mining of Massive Datasets, 3rd ed. Cambridge, U.K.: Cambridge UniversityPress,2020.
[44]G.Shmueli,N.R.Patel,andP.C.Bruce,DataMiningfor Business Analytics: Concepts, Techniques, and ApplicationsinPython.Hoboken,NJ,USA:Wiley,2017.