陆宝红+陈晨+曹曦文+陈眺+曾洁+史昇+宋雪桦

摘要：

车辆加速性能是衡量驾驶员对车辆驾驶舒适性的一个标准。传统的车辆加速性能是通过判断加速踏板开度衡量的，加速踏板出厂时设置为不能自动调整。采用改进的C4.5算法对车辆加速性能分类，实现自动调整加速踏板开度。首先通过泰勒中值定理对C4.5算法进行简化，然后对车联网数据进行特征提取，生成判断加速性能的决策树分类规则，测试特征提取的并行化运行效率。通过特征提取后的数据集验证了改进的C4.5算法效率和准确率。测试结果表明，改进算法在不降低分类准确率的前提下，有效提高了分类效率。

关键词：

车联网；HBase；MapReduce；C4.5

DOIDOI：10.11907/rjdk.172155

中图分类号：TP391

文献标识码：A文章编号文章编号：1672-7800（2018）001-0184-04

Abstract：Vehicle acceleration performance is a measure of the drivers comfort in driving a vehicle and improving driving comfort is a matter of concern. The traditional vehicle acceleration performance is determined by judging the accelerator pedal opening degree， and the accelerator pedal can not be adjusted automatically when the factory is set up. Therefore， the paper adopts the improved C4.5 algorithm to automatically adjust the accelerator pedal opening degree to the vehicle acceleration performance classification. Firstly， the algorithm of C4.5 is simplified by Taylors mean value theorem， and then the feature extraction of car network data is carried out to generate decision tree classification rules to judge the acceleration performance， parallel operation efficiency of test feature extraction， and finally the data extracted by feature Set to verify the efficiency and accuracy of the improved C4.5 algorithm. The test results show that the improved algorithm effectively improves the efficiency of classification and the classification accuracy is not reduced.

Key Words：car networking； HBase； MapReduce； C4.5

0引言

分类问题在数据挖掘领域研究和应用较为广泛[1]。决策树算法分类速度快、精度高、生成的分类规则易于理解。其中，决策树C4.5算法最为经典，具有以下优点：分类规则易于理解，准确率较高；基本解决了ID3存在的多值偏向性问题；可以处理连续数值型属性；可以处理缺失值；能够在决策树构造过程中进行剪枝[2]，但C4.5存在构造决策树效率较低的缺陷。

车辆的加速性能以及踏板的开度使用情况反映了驾驶的舒适性，提高加速性能的关键是如何通过整车参数判断加速性能不足，以及油门踏板加速过深或加速过浅。通过设置阈值来判断加速性能的传统方法无法证明其合理性，阈值由于环境及人为因素影响不好准确界定。……