PsyPredict - An EVM based Postpartum Depression Predictor

Abstract

Postpartum depression, also referred to as PPD, a serious mental health issue that affects mothers after childbirth and is often undiagnosed because of its complex patterns of symptoms. This condition has drastic and prolonged effects on both the mother and her child. This makes a critical need for a method of its early detection. Existing PPD detections mostly rely on clinical tests or self-reports, bit these can be time consuming and lack early detection. This is addressed by various machine learning models that are used, these include Support Vector Machines, XGBoost, Neural Networks and Ensemble Voting Classifier. These analyze complex data to predict PPD with high accuracy. PsyPredict presents a machine learning framework which would help in early detection of postpartum depression by using a clinically structured data set comprising of more than 1500 postnatal survey responses with more than 10 psychological and behavioral indicators. Our model (Ensemble Voting Classifier) achieves approximately 97% accuracy. PsyPredict also culminates in a deployable dashboard which makes it capable of real time predication and clinical interpretation support. This bridges the gap between advance machine learning models and decision support. Through this study we demonstrate a thoughtful combination of data processing, model selection and treatability tools that can significantly enhance PPD detection.