【摘要】 用户行为数据的复杂性和多样性使得现有算法难以准确提取区分用户身份的特征，进而影响检测系统的准确性，易产生误判或漏判。因此，提出基于TextCNN模型的实时身份检测算法。引入TextCNN模型搭建实时身份检测框架，以此为基础，应用TextCNN模型——输入层、嵌入层、卷积层、池化层、全连接层与输出层，计算PC端和WEB端用户行为数据特征数值，并确定其对应的融合权重。融合这两端的数据特征后，制定用户身份检测规则，执行制定规则即可以实现用户身份的实时检测。实验结果显示：应用提出算法在PC端行为数据特征提取方面取得了显著成效。算法提取的特征与实际结果高度一致，特征提取完整度最高达到95%,且用户身份检测的F1分数最高为0.92,证明了该算法具有出色的应用性能。更多还原

【Abstract】 The complexity and diversity of user behavior data make it difficult for existing algorithms to accurately extract the characteristics that distinguish user identity, which further affects the accuracy of the detection system, and is prone to miscarriage or omission of judgment. Therefore, a real-time identity detection algorithm based on TextCNN model is proposed. The TextCNN model is introduced to build a real-time identity detection framework. On this basis, the TextCNN model-input layer, embedded layer, convolution layer, pooling layer, full connection layer and output layer-is applied to calculate the characteristic values of user behavior data on the PC side and the WEB side, and determine their corresponding fusion weights. After fusing the data characteristics of these two ends, the user identity detection rules are formulated, and the real-time detection of user identity can be realized by executing the formulated rules. The experimental results show that the proposed algorithm has achieved remarkable results in the feature extraction of PC behavior data. The features extracted by the algorithm are highly consistent with the actual results. The integrity of feature extraction is up to 95%, and the F1 score of user identity detection is up to 0.92, which proves that the algorithm has excellent application performance.更多还原