Detecting Early Markers of Post-Partum Depression in New Mothers: An Efficient LSTM-CNN Approach Compared to Logistic Regression

This study proposes an efficient model combining Long Short-Term Memory (LSTM) with Convolutional Neural Networks (CNNs) to detect early markers of post-partum depression among new mothers. We compare the performance of this model with Logistic Regression, a traditional method. Our approach involves leveraging the temporal dependencies in the data using LSTM and capturing spatial features using CNNs. We aim to provide accurate and timely identification of post-partum depression symptoms to enable early intervention and support for new mothers. Through comprehensive comparison with Logistic Regression, we demonstrate the superior performance of our proposed model in detecting early markers of post-partum depression. In this study, there are two groups: The Novel Long-Short Term Memory with Convolutional Neural Networks (LSTM-CNN) and another group is Logistic Regression, each with a sample size of 20. The mean accuracy of the LSTM-CNN is 77.26% and the Logistic Regression model is 72.36%, respectively. The independent sample t- test has a significance value $p=0.04 (\mathrm{p}\lt 0.05)$ indicating the study between the LSTM-CNN model and the Logistic Regression algorithm is statistically significant. Overall, deep learning algorithms have shown promise in predicting Postpartum depression and the LSTM-CNN method achieved a higher accuracy when compared to the Logistic Regression and other machine learning algorithms.