Identifying traffic congestion and solving them by using predictive models has been ongoing research in intelligent transportation scenarios. However, it is improper that such scenarios can be judged on the basis of mean traffic intensity and mean traffic speed. This paper works on this aspect and uses data mining approaches to derive the aggregation metrics of traffic intensity data from the city of Madrid. This work uses a novel similarity measure by utilizing the results of the Wilcoxon Signed Rank test across 2018 locations to discover similarities. We propose a Genetic Algorithm on the results of the Wilcoxon test for forming communities based on the aggregation metrics. This work also compares and evaluates the performance of the proposed algorithm against standard distance measures and other state-of-the-art approaches. Forfinding the optimal number of possible communities in the data, we have taken the help of Davies - Bouldin Test. Our experimental results show the effectiveness of the Genetic Algorithm using various parameters, such as number of dissimilar points within a cluster, minimum number of dissimilar data points between clusters and overall based on Modified Silhouette coefficient. Furthermore, we find that our method is able to distribute the data points in a more uniform manner across formed communities in comparison to other approaches considered in this work.
Bibliographical noteVersion archived for private and non-commercial use with the permission of the author/s and according to publisher conditions. For further rights please contact the publisher.
- evolutionary clustering
- fuzzy c-means
- genetic algorithms
- intelligent transportation systems
- K-means clustering
- traffic classification