In order to study and predict the freeway traffic safety and realize the traffic flow in the nonlinear big data environment, based on deep learning, the long-short-time memory (LSTM) model based on recurrent neural network is proposed. The traffic flow is predicted and the predicted value of traffic flow is compared with the actual value at different times. The mean absolute percentage error of LSTM prediction model is tested and compared with the error of time proximity, periodicity, and trend. At the same time, the naive Bayes algorithm is used to carry out image recognition processing for attributes such as license plate number and vehicle color to conduct vehicle matching. The data processing, training process, and model realization of the model are studied, and the accuracy of the naive Bayesian algorithm is tested. The results show that the predicted value of the traffic flow prediction model based on LSTM is not much different from the actual value. The average prediction error for the period from May 7, 2018 to May 9, 2018 is approximately 13.8%. When the time series is 6, the error of the prediction model based on LSTM is 10.72%, and the prediction errors of the three sequences of time proximity, periodicity, and trend are 15.66%, 17.59%, and 20.67%, respectively. Considering the three sequences comprehensively, the prediction model can achieve good prediction effect. The accuracy of the vehicle matching model based on naive Bayes is about 82.7%, which can meet the requirements of the system. Therefore, it can be concluded that the LSTM traffic flow prediction model based on deep learning and the image recognition vehicle matching model based on naive Bayes can realize the traffic safety prediction of freeway, which has great practical significance.

为了研究和预测高速公路的交通安全性，实现非线性大数据环境下的交通流量，基于深度学习，提出了一种基于递归神经网络的长短时记忆模型。对交通量进行预测，并将交通量的预测值与不同时间的实际值进行比较。测试了LSTM预测模型的平均绝对百分比误差，并将其与时间接近度，周期性和趋势误差进行了比较。同时，采用朴素贝叶斯算法对车牌号，车辆颜色等属性进行图像识别处理，进行车辆匹配。研究了模型的数据处理，训练过程和模型实现，并测试了朴素贝叶斯算法的准确性。结果表明，基于LSTM的交通流量预测模型的预测值与实际值相差不大。从2018年5月7日到2018年5月9日的平均预测误差约为13.8％。当时间序列为6时，基于LSTM的预测模型的误差为10.72％，时间接近度，周期性和趋势这三个序列的预测误差分别为15.66％，17.59％和20.67％。综合考虑这三个序列，预测模型可以获得良好的预测效果。基于朴素贝叶斯的车辆匹配模型的准确性约为82.7％，可以满足系统的要求。因此，