Predicting Postpartum Depression Risk Using Cross Vector Spider Swarm Intelligence and Hypernet Gated Multi-Perceptron Neural Network

Abstract

One of the most critical mental diseases, which influences the health of mothers and newborn babies, is Postpartum Depression (PPD). The issue of predicting risk factors for PPD, based on the analysis of vast Personal Health Records (PHR), is highly problematic, which complicates traditional predictive systems. This paper introduces a forecasting model that combines Cross Vector Spider Swarm Intelligence (CVSWI) and a Hypernet Gated Multi-Perceptron Neural Network (HG-MPNN) to improve the early detection and control of PPD. The procedure begins with the collection of the PPD-PHR dataset and its pre-processing using a Z-score covariance filter to remove irrelevant data and enhance data quality. To calculate the PPD Impact Margin Rate, a decision tree approach is adopted to obtain a coherent understanding of the relationship between risk factors and the occurrence of PPD. The advantage of CVSWI is its ability to maximise the features it selects, which are likely to be essential predictors, while reducing dimensionality. The Active Scalar Pattern Mining Algorithm (ASPMA) is capable of identifying latent patterns associated with PPD. The suggested HG-MPNN model has been effective, with an accuracy value of 99.36, a precision of 1.00, a recall of 0.99, and an F1-score of 0.99 (which implies that the model can categorise the risk levels of PPD with limited false predictions).