Objective: Chronic non-specific low back pain (cNSLBP) is a common disease characterized by pain and functional limitations. Currently, the evaluation of its condition mostly relies on subjective scales and lacks objective indicators. This study aims to propose an objective evaluation method for cNSLBP based on pressure data by collecting patients' sitting pressure data and combining it with Swin Transformer neural network model. Method: A total of 100 patients with cNSLBP were recruited for the study, and pressure data were collected using a pressure-sensitive sensor array in five sitting positions: forward leaning, backward leaning, left leaning, right leaning, and sitting. These data were then converted into pressure images. A stress dataset containing multi-level features was constructed through data preprocessing and enhancement methods. Using Swin Transformer model to analyze pressure images, extract features and predict the severity of the disease, and compare it with traditional convolutional neural networks. Result: The Swin Transformer model achieved a mean absolute error (MAE) of 0.1368 on the cNSLBP patient pressure dataset, significantly better than traditional convolutional neural network models. Through data augmentation methods, the model exhibits strong generalization ability in limited sample sizes and can effectively distinguish the severity of patient conditions. Conclusion: The sit pressure data analysis method based on Swin Transformer can accurately evaluate the severity of cNSLBP patients, providing an objective and efficient auxiliary diagnostic method for clinical practice, and can better guide the clinical diagnosis and treatment of non-specific low back pain. Keywords chronic non-specific lower back pain; Sitting pressure; Disease assessment