Perceived pressures, especially those that are viewed as detrimental or overpowering, naturally cause stress. Chronic stress exposure greatly raises the likelihood of mental health issues like anxiety, sadness, and irregular sleep patterns, even while short-term stress can be adaptive. Heart rate variability (HRV), which measures the dynamic fluctuations in the time intervals between heartbeats rather than presenting a static heart rate number, is one of the most informative physiological indicators of stress. This study presents a sophisticated convolution neural network (CNN)-based model for multi-class stress classification and investigates HRV as a reliable biomarker for stress detection. The model provides a more sophisticated understanding of stress patterns by dividing stress into three different states: no stress, interruption stress, and time pressure stress. When tested on the SWELL-KW dataset, the suggested solution outperformed current techniques in terms of accuracy. Additionally, by using variance analysis to improve the interpretability and accuracy of stress evaluation, this work highlights the crucial role that HRV characteristics play in stress identification.