Stress affects the physical and mental health of women significantly; hence accurate prediction models are essential in combating stress. In this review, we highlight the recent advances in stress prediction using deep learning (DL) in women and identify potential methodological pitfalls and challenges. This paper provides a review of state-of-the-art DL models applied to stress prediction in the literature. The focus is on the datasets referred, feature extraction techniques implemented, and evaluation measures used in the literature. Also, within this review, we emphasize the incorporation of multimodal physiological data, behavioral data, and environmental data into predictive algorithms. The paper further highlights major challenges, including data imbalance, generalization issues, and ethical concerns, along with possible solutions like data augmentation transfer learning and privacy-preserving techniques. In conclusion, the paper discusses future work to build personalized real-time stress prediction systems for women that utilize findings on physiological and psychological differences between sexes. Such a comprehensive review presents directions of research for the purpose of catering to researchers and practitioners to advance this field while utilizing better approaches through deep learning innovations.