A Hybrid Machine Learning Approach To Detect Postpartum Depression (PDB)

Abstract

Postpartum depression, which occurs in the days and weeks following childbirth, can have serious impacts on both mothers and babies. Symptoms of postpartum depression include mood swings, exhaustion, and a sense of hopelessness, and it can lead to long-term mood disorders such as postpartum psychosis. In some cases, this condition can even lead to maternal and infant mortality. Traditional methods of detecting postpartum depression, such as face-to-face doctor consultations, can be time-consuming and may not be feasible for individuals in remote areas. To address this challenge, we propose using a Hybrid machine learning algorithm, which is ensemble the four algorithm those are decision trees, K-nearest neighbors, logistic regression and support vector machines. We trained our hybrid model using our dataset of over 1503 sample, which has 15 different features. We got our best accuracy with our hybrid model; the accuracy is 98.78% with less error than the other traditional machine learning algorithm. We got the second-best accuracy with random forest 98.34%. Based on these results, we conclude that our hybrid model show the best performance for detecting the postpartum depression (PPD), while other machine learning algorithm exhibits the lowest performance.