摘要
针对深度学习在精神疾病预测中模型多样性不足和准确度有限的问题,文章提出一种融合卷积神经网络(CNN)空间特征提取能力和长短时记忆网络(LSTM)时序信息处理能力的混合模型。通过10折交叉验证和多种评估指标测试,该CNN-LSTM模型在特定数据集上取得了显著效果,预测准确率达到97.65%,灵敏度为96.47%,特异性达98.82%。实验结果表明,该混合模型的性能明显优于单独使用CNN或LSTM模型,验证了其在精神疾病预测任务中的有效性和优越性。
In response to the problem of insufficient diversity and limited accuracy of deep learning models in predicting mental illness,this article proposes a hybrid model that combines the spatial feature extraction ability of convolutional neural networks(CNN)and the temporal information processing ability of long short-term memory networks(LSTM).Through 10 fold cross validation and multiple evaluation metrics testing,the CNN-LSTM model achieved significant results on a specific dataset,with a prediction accuracy of 97.65%,sensitivity of 96.47%,and specificity of 98.82%.The experimental results show that the performance of the hybrid model is significantly better than using CNN or LSTM models alone,verifying its effectiveness and superiority in mental illness prediction tasks.
作者
高亚婕
刘鑫源
袁子钦
李星玥
GAO Yajie;LIU Xinyuan;YUAN Ziqin;LI Xingyue(College of Engineering,Qufu Normal University,Rizhao,Shandong 276826,China)
基金
2024年山东省大学生创新训练项目(S202410446098)。
