The invention discloses a self-adaptive traffic event detection method based on an improved random forest. The method comprises the following steps: acquiring data that a vehicle enters and exits twoadjacent ETC portals in a detection period T; calculating traffic parameters; generating a balance sample set based on k-means clustering; training a random forest model by utilizing the balance sample set; inputting the real-time traffic parameters as input vectors into the established random forest model and outputting an event detection result; and calculating a false detection rate according to a detection result, and performing model adaptive adjustment according to the false detection rate. According to the method, k-means clustering sampling is introduced to generate a balance sample, arandom forest is generated according to the balance sample, speed dispersion characteristics are considered, flow, average speed and speed dispersion in a detection period are used as input vectors,the generated random forest model is input, real-time event detection is carried out, and finally, adaptive adjustment is performed on the algorithm according to the historical false detection rate, thereby realizing real-time accurate detection of the traffic event.
本发明公开了一种基于改进随机森林的自适应交通事件检测方法,该方法包括:获取检测周期T内车辆出入相邻两个ETC门架的数据;计算交通参数;基于k‑means聚类生成平衡样本集;利用平衡样本集训练随机森林模型;将实时交通参数作为输入向量输入建立好的随机森林模型并输出事件检测结果;根据检测结果计算误检率,然后根据误检率进行模型自适应性调整。本发明通过引入k‑means聚类抽样产生平衡样本,根据平衡样本生成随机森林,并且考虑了速度离散特征,以检测周期内的流量、平均速度和速度离散度作为输入向量,输入所生成的随机森林模型,进行实时事件检测,最后根据历史误检率对算法进行自适应调整,从而实现了交通事件的实时准确检测。
Adaptive traffic event detection method based on improved random forest
一种基于改进随机森林的自适应交通事件检测方法
2021-03-26
Patent
Elektronische Ressource
Chinesisch
Traffic abnormal data anomaly detection method based on improved robust random forest
Europäisches Patentamt | 2024
|Europäisches Patentamt | 2023
|Europäisches Patentamt | 2022
|Random forest-based emergent traffic incident space-time deduction method
Europäisches Patentamt | 2023
|