基于历史轨迹的个性化路线推荐算法的研究

Research on Personalized Route Recommendation Algorithm Based on Historical Trajectory

针对各导航软件快速给出私家车出行最佳路线却无法满足用户个人偏好的问题,论文基于2017年11月全国8186台私家车的出行轨迹,提出一种个性化路线推荐的优化方法。通过混合高斯模型拟合偏好分布,分析相关性找出与当前用户偏好和出行时间最为相近的轨迹,最后从中选择一条最符合用户需求的路线。与DAI方法相比,推荐路线匹配度高于80%的推荐轨迹高出了47%,推荐准确度也有了大幅度的提高,准确度对比平均值达到1.102。

Aiming at the problem that each navigation software quickly gives the best route for private cars but fails to meet the user's personal preferences,this article proposes a personalized route recommendation optimization method based on the travel tra⁃jectories of 8,186 private cars nationwide in November 2017.The mixed Gaussian model is used to fit the preference distribution,and the correlation is analyzed to find the trajectory that is closest to the current user preference and travel time.Finally,a route is selected that best meets the user's needs.Compared with the DAI method,the recommended trajectory with the recommended route matching degree higher than 80%is 47%higher,and the recommendation accuracy has also been greatly improved.The average ac⁃curacy comparison has reached 1.102.