Stress is a widespread issue that affects people everywhere. Several things, including difficulty at job, money troubles, marital problems, and health issues, can contribute to it. Anxiety, depression, and cardiovascular illnesses are just a few of the physical and mental health issues that stress can cause. In order to properly manage stress, it is crucial to understand what causes it in people. Researchers and clinicians can encourage the development of new approaches to deal with the problematic effects of a sustained stress response by identifying and managing stress in its early phases. We propose to use sleep dataset in order to predict the stress levels as compared to non-sleep mode data that has attributes such as snoring range, respiration rate, body temperature, limb movement rate, blood oxygen levels, eye movement, number of hours slept, and heart rate with stress levels classified from 0 (low/normal) to 4 (extreme/high). We compare the performance of Logistic regression, XGB classifier, Random Forest classifier, Naive Bayes, Decision tree, LightGBM, CatBoost (Categorical Boosting), Gradient Boosting Regressor, SVM, ANN, and KNN models to predict the stress levels for sleep mode data. We found that only SVM and Naïve bayes gave a maximum accuracy of 85% stress prediction accuracy using non-sleep mode data whereas many models such as Logistic regression, Naive Bayes, LightGBM, CatBoost, SVM, and KNN predicted with 100% accuracy for sleep mode data.