基于AdaBoost算法的货车驾驶人安全倾向性分类

Truck driver safety tendency classification based on the AdaBoost algorithm

为了研究货车驾驶人驾驶行为的差异性,对不同货车驾驶人实行分类管理。通过车载OBD设备获取39位货车驾驶人在自然驾驶状态下的车辆行驶数据,采用均值滤波方法对数据进行平滑滤波处理,以消除车辆行驶过程中由于路面颠簸、发动机抖动等外界环境对数据的影响。选择最高车速、横向加速度峰值、行车方向加速度峰值、车速与发动机转速的最大相对比值作为货车驾驶人安全倾向性评价指标。在对数据进行K-means聚类分析的基础上,应用AdaBoost算法建立货车驾驶人安全倾向性分类模型,将货车驾驶人分为激进型驾驶人或保守型驾驶人。数据验证分类结果表明,基于AdaBoost算法的货车驾驶人安全倾向性分类模型的平均准确率可以达到98.74%,可有效区分激进型货车驾驶人及保守型货车驾驶人。

This paper is mainly to explore the differences or gaps of the different types of truck drivers in their driving behavior by classifying and managing different types of truck drivers from the chosen 39 truck drivers in the natural driving state so as to gain their vehicle driving data by driving the OBD(On-Board Diagnostics)vehicles.For such a goal,the data concenred on the said 39 truck drivers has been filtered via the mean filtering method by eliminating the external environment influential factorsto be caused by the road bumps and engine shake.But,the parameters,such as the maximum speed,the maximum lateral and longitudinal acceleration,the maximalized relative ratio of the vehicle speeds to the engine speed have actually been set up and controlled as their safety driving tendency evaluation indexes.On the other hand,K-means algorithms has been taken to cluster the said 39 truck drivers'driving data by dividing them into aggressive and conservative ones.Furthermore,the t test proves that there indeed exist significant differences on the mean parameters we have chosen,and their difference in the mean parameters tends to be statistically conspicuous.And,then,their driving safety tendency has been confirmed and determined by the AdaBoost algorithm of the classification model.And,in so doing,the results of K-means algorithms have been worked out to classify the said truck drivers as the sampling labels of the AdaBoost algorithm.In addition,the sampling data from the above truck drivers have been set up as the training set(26 samples)and the testing set(13 samples)of the AdaBoost algorithm,correspondingly and respectively.The data validation classification results demostrate that the average level of the truck drivers'safety tendency classification model based on AdaBoost algorithm has been worked out at about 98.74%,which can help to distinguish effectively the aggressive truck drivers from the rather conservative ones.And,then,the classification results of the drivers of the aggressive type prove to be prone to high-speed driving,overtaking,sudden acceleration,sudden deceleration,and unexpected braking speed and so on,and the conservative ones,on the contrary,are generally preferring to express stronger or more patient self-control driving with greater stability and reveal more disciplinary to the traffic rules.Therefore,different personalities and qualifications can be detected and categorized to evaluate the drivers as to their vehicle control and operation characteristic styles in the testing groups.