Early Detection of Postpartum Depression Using a Hybrid Fuzzy C-Means Clustering and Random Forest Model

Abstract

Postpartum Depression is an important and crucial mental health condition that highly affects a large proportion of women after childbirth. If this goes unnoticed or late detection is there then it might lead to adverse outcomes for both the mother and the child. The complexity as well as the subjectivity of psychological and behavioural factors make its early diagnosis very challenging. To address this issue, this study proposes a hybrid intelligent framework that combines Fuzzy C-Means clustering technique with a Random Forest classifier for early detection of postpartum depression risk. The proposed approach takes Fuzzy C Means Clustering into consideration for modelling the uncertainty and assigns a partial membership value to the individuals across selected risk categories (low, medium and high risk). These fuzzy-derived risk labels are then used for training a Random Forest model. Then we introduce a hybrid scoring mechanism to combine the probabilistic output of the Random Forest with fuzzy membership values, enhancing both prediction accuracy as well as interpretability. The experimental results demonstrate that the proposed hybrid method is yielding an accuracy of 91.84%. It outperforms the individual Fuzzy C-Means clustering model and the Random Forest classifier. The paper also presents a comparative analysis with the recent state-of-the-art methods to confirm that the proposed approach offers competitive accuracy while maintaining lower computational complexity and higher interpretability. The proposed framework provides a practical and scalable solution for early screening of postpartum depression which has the potential to be integrated into a clinical decision support system to facilitate timely intervention and improve the maternal health.