一种基于用户异步轨迹的身份识别智能方法

针对传统的轨迹身份识别存在的特征选择主观性强、精度有限等问题,本文提出了一种融合双向循环神经网络模型(ConvGRU-Bidir)。首先采用一维卷积和一维池化压缩轨迹数据,提取高维特征;然后采用双向GRU,分别从时间正序和时间逆序学习轨迹特征,最终实现用户身份ID识别。研究采用GeoLife轨迹数据集,来自122名用户的10837个轨迹样本参与模型训练及测试。结果表明,本文提出的模型对于异步轨迹数据的身份识别精度达97.28%,相比现有方法精度至少提高30%,由此证明了深度学习在此类问题上的可行性和有效性。

An intelligent identification method of user identity based on asynchronous trajectory

Traditional trajectory identification methods have limit in feature selection and accuracy. Therefore, this paper proposes a mixed neural network of convolutional neural networks and recurrent neural networks(ConvGRU-Bidir).Firstly, the one-dimensional CNN and one-dimensional pooling neural networks will compress trajectory data and extract high-dimensional features. Then, the bidirectional GRU learns trajectory features from time positive and reverse simultaneously. Finally, the model can recognize users' ID number. This paper uses the GeoLife trajectory dataset to train and test the model, which contains 10837 trajectory samples from 122 users. The results show that the model has an identification accuracy of 97.28% for asynchronous trajectory data, which has improved by at least 30% compared with the existing methods, which proves deep learning's availability and effectiveness in such problems.