Research And Application Of Unsupervised Clustering Algorithm In Driving Hotspot

Clustering is one of the most important technologies of data mining,which is used to discover unknown classification in data set. In addition, the clustering algorithm can not only be used as a tool to find the distribution of deep-level information data in database but also that can be used as a preprocessing step in data mining. Clustering analysis is to achieve such a goal that will division the data object into different clusters that the same cluster has high similarity but the similarity of data object in different clusters is low. In order to solve the problems of driving hotspot, this paper studied the traditional cluster analysis technique, driving hotspots which can reflect the user’s traffic laws are the zone of frequent vehicle.At first, some related concepts of unsupervised clustering techniques are given in this paper. The chief point of the paper is the reseatch on K-Means which is based on partitioning. K-Means algorithm has some advantages including O(n) time complexity and it is easy to use and can work well with large data set. Next the focus of this article is as follows, this paper proposes an improved algorithm KMSDR,which aims at to improve the three shortcomings of the number of clusters k, initial cluster centers and sensitive to isolated points. Goal of the algorithm is to find the cluster center, while ensuring that the large similarity in the same cluster and small similarity between different clusters. Algorithm uses improved Max-Min distance to select the new centers, according to the center distance threshold to determine whether the center is isolated; proposes Dis(S,k) distance function for getting the most suitable cluster number k and the approach of using data object in cluster instead of the mean center to remove the isolated points which can improve the quality of clustering. Improved algorithm KMSDR is proved that it has advantage in accuracy and efficiency by the simulation results on three sets of data. Finally, the improved algorithm KMSDR applied to the traffic hotspot issue of traffic data analysis system, and complete display function of the page for the traffic hotspot, and design and implement the functionality of ranging and surrounding retrieval.Experiments verify that improved algorithm which has strong stabiliyt can accurately find driving hotspots and it also has been improved on time performance. Driving hotspots may not only reflect the activity patterns of vehicle so that users know their driving habits, but also provide convenience for the user’s travel.