基于异常点检测的心理健康辅助诊断方法

Approach of assisted diagnosis for mental health based on outlier detection

采用异常点检测算法研究心理健康辅助诊断任务,提出并设计一种基于异常点检测的心理健康辅助诊断方法,有效识别心理沙盘数据中的异常样本。在构建心理健康辅助诊断模型过程中,分析数据特性,提取与用户心理健康状况高度相关的特征,构建虚拟心理沙盘数据集;使用4种传统异常点检测算法,识别沙盘数据集中异常样本,设计融合策略,集成不同算法检测结果,提高异常样本检测精准性和效率,辅助人类专家进行精确诊断;对模型预测性能和结果进行详细分析,结合基线模型进行对比评价。试验结果表明,基于异常点检测的心理健康辅助诊断方法在沙具使用相似度、距离度量、聚类性能等3项指标上获得较好性能。

Towarding the assisted diagnosis for mental health research problem,outlier detection was employed to propose and design an outlier detection-based approach for mental health assisted diagnosis,which could effectively identify outliers from large amounts of virtual mental sandbox samples.During constructing the assisted diagnostic model for mental health,the data characteristics were analyzed,and meanwhile the features highly correlated with the user's mental health status were extracted to establish a virtual psychological sandplay dataset.Four traditional outlier detection algorithms were utilized to identify abnormal samples in the sandplay dataset.A fusion strategy was designed,and the detection results of different algorithms were integrated to enhance the accuracy and efficiency of anomaly sample detection,to assist human experts in making precise diagnoses.The performance and result of the proposed approach were analyzed in details through comparing with the baseline models.The experimental results showed that compared with other models,the assisted diagnosis for mental health based on outlier detection approach had better performance in three indexes such as similarity of sand tool usage,distance of mean vector and clustering performance.