In the field of offline handwritten Chinese character evaluation, the stroke writing order (hereinafter referred to as stroke order) is one of the important evaluation indicators. However, research on the intelligent recognition of offline handwritten Chinese character stroke order is still relatively scarce. This paper proposes a method for the intelligent recognition of offline handwritten Chinese character stroke order using deep learning models and constructs a corresponding stroke order dataset. The method extracts features of each stroke in the handwritten Chinese character using MobileNetV2, and the extracted features are input into the encoder for sequence modeling. In the decoder, an attention mechanism is employed to calculate attention weights. During this process, the decoder dynamically focuses on different parts of the sequence based on the current hidden state and the encoded results from the encoder, emphasizing the key information of each stroke. In this way, the model can effectively capture the dependencies between different strokes, ultimately achieving more accurate stroke order recognition. Experiments conducted on the stroke order dataset show that the method achieves an accuracy of 86.21% and an average edit distance of 0.90, providing a certain research foundation and practical value for offline intelligent recognition of handwritten Chinese character stroke order.