Machine learning approaches for modeling reproductive patterns on count data

Reproductive patterns define the trends and behaviours related to childbirth, family size, and the timing of births within a group. They are affected by several variables, including social, economic, and cultural, along with biological components, providing them a crucial area of study in demography and health research. Analyzing reproductive patterns offers insights into societal health, family planning requirements, and population dynamics. The total number of children ever born (CEB) is a crucial metric in these patterns, reflecting a woman's lifelong fertility. This study analyses the factors affecting fertility among Andhra Pradesh's reproductive-age women, focusing on the socio-demographic impacts on the number of CEBs using information obtained from the fifth round of the National Family Health Survey (NFHS-5). This research applies machine learning techniques to identify key predictors of the number of CEB and compares model performance. Feature selection using the Boruta algorithm identified women’s age, fertility preference, wealth index, marital status, type of cooking fuel, religion, caste, and residence as the most influential variables. Associations between predictors and fertility outcomes were further explored using chi-square tests. The Support Vector Machine emerged as a precise prediction model. The findings provide insights for demographic research and support evidence-based reproductive health policy planning.