Audit Risk Assessment Model Construction by Incorporating Deep Learning Techniques

While risk-oriented auditing has gradually emerged as a new trend in current internal auditing, the absence of specific risk assessment methods in related internal auditing standards has made it a common challenge for many enterprise internal auditing organizations and personnel to effectively implement and conduct risk assessments. This research focuses on integrating deep learning technology to construct an audit risk assessment model, first describing the problem of audit risk assessment and then validating it using the CNN-LSTM method. We select the appropriate indicators for audit risk assessment assignment after screening the sample data to ensure their qualitative and quantitative comparability. We construct a CNN-LSTM audit risk assessment model based on the above data. We use two input methods, the convolutional neural network and the long- and short-term memory network, to enhance the assessment model’s learning ability. We then choose 12 enterprise companies as data cases, with Company A serving as the representative for the CNN-LSTM-based assessment model’s performance simulation test and risk level assessment. We compare the four models, and the CNN-LSTM model’s AUC exceeds 0.5 in the confusion matrix and ROC curve graph, confirming the usefulness of CNN-LSTM in audit risk assessment.