Improved Statistical Approach to Analyze Multivariate Women’s Fertility Dataset for Better Prediction

The integration of Artificial Intelligence (AI) has significantly transformed healthcare, particularly in reproductive medicine for treatment planning and disease diagnosis. However, the effectiveness of AI models relies heavily on access to high-quality data, and many researchers face limitations due to inadequate datasets, which hampers the commercialization of AI systems. To address this issue, we conducted a survey among women at an infertility clinic in West Bengal, India, following WHO recommendations, and involved clinicians in patient assessments. Our analysis of women’s fertility included 17 lifestyle-related features, resulting in a dataset of 70 samples categorized as fertile or infertile. The primary aim was to augment this limited dataset by introducing a novel statistical probability-based method to generate up to 700 synthetic samples. We evaluated the effectiveness of this method using six conventional AI tools: support vector machine, decision tree, naïve Bayes, random forest, multi-layer perceptron, and bagging. Our findings indicate a significant performance improvement with the larger sample size, with the support vector machine serving as the benchmark. This technique offers a valuable solution for researchers dealing with small multivariate datasets, enhancing opportunities for more robust applications in reproductive medicine.