COMPREHENSIVE ANALYSIS OF ACCIDENTAL MORTALITY AND TRAFFIC ACCIDENTS IN INDIA (2012-2022): A BUSINESS INTELLIGENCE APPROACH

Abstract

This research project is a detailed Business Intelligence (BI) study covering the last eleven years (2012-2022) of accidental death and traffic accidents in India. Despite great efforts to address public health concerns and infrastructural developments, accidental deaths still continue to be a significant socio-economic issue in India, often claiming over 400,000 lives per year. The main aim of this study was to move away from traditional descriptive reporting of mortality and into a more advanced analytical framework that is predictive and prescriptive. This study will involve applying machine learning algorithms to identify the patterns of mortality, predict mortality rates in the near future and classify geographic areas by risk so that interventions can be targeted to the appropriate areas. The methodology involved the use of comprehensive data, including more than 218 aggregated state and temporal records, with the exception of the suicide data, which would not provide an analytic focus on preventable accidental mortality. A multi-phased analytical approach was used. In the descriptive phase, the analysis showed that traffic accidents are the main cause of death, accounting for 46.0% of the deaths from all causes (194,347 deaths were caused by traffic accidents in 2022 alone). The data also showed a large demographic vulnerability, with a ratio of 4.77 male deaths for every female death, and a significant proportion of deaths occurred in the working age population (18-45 years). Several machine learning models were created and tested during the predictive phase. A Linear Regression model successfully showed that the national death rate has a high correlation with natural and unnatural causal factors (R² = 0.86), and predict the national death rate well. At the same time, regional data points were clustered using K-Means clustering to form three risk segments.At the same time, the regional data points were clustered using K-Means clustering to form three risk segments. This concentration brought out stark geographic differences, such as the dangerous areas in Bihar and Punjab with almost 80% fatality rates, while the less risky areas had rates of 28.9%. Also, a Logistic Regression classification model was developed to act as an early warning system, which successfully classified high, medium or low risk years using demographic and causal indicators. The ultimate goal of the research is to provide prescriptive recommendations for policies. The use of the developed risk clusters proposes a data-driven and decentralized disaster management and road safety approach. Some of the key recommendations are that the identified "High-Risk" clusters be given more funds in emergency infrastructure budgets and that predictive models be used to define localized Key Performance Indicators (KPIs) for the National Road Safety Mission.