A HYBRID CNN–LSTM-BASED PREDICTIVE MODEL FOR NUTRITIONAL DEFICIENCY (IN WOMEN’S) DETECTION USING QUANTITATIVE MODELLING AND DEEP FEATURE LEARNING ON CLINICAL TIME-SERIES DATA

Women who aren't getting enough nutrition is still a public health problem, especially in places with few resources where early diagnosis and treatment are needed to avoid long-term health problems. Using real-life clinical datasets, this study shows a deep learning-based diagnostic framework for automatically finding nutrient deficiencies. Specifically, it looks at Vitamin A and Iron deficiencies in women. A CNN, LSTM network, and a hybrid (CNN–LSTM) model will be designed and tested as part of the proposed method. The dataset went through a lot of preprocessing, feature selection, and transformation to make it work with deep learning architectures. We saw that the CNN–LSTM model was more accurate at classifying things than other models like SVM, KNN, and Naïve Bayes. It got 93.2% for Iron Deficiency and 92.5% for Vitamin A Deficiency. The hybrid model also did a better job of generalising and overfitting, as shown by the training-validation loss curves. The suggested deep learning framework provides a scalable, non-invasive, and highly accurate way to find nutritional deficiencies early on. It can be used in real-time diagnostic systems or built into platforms for monitoring public health.