Polykarpos Meladianos

Evaluating the Effect of Time Series Segmentation on STARIMA-Based Traffic Prediction Model

As the interest for developing intelligent transportation systems increases, the necessity for effective traffic prediction techniques becomes profound. Urban short-term traffic prediction has proven to be an interesting yet challenging task. The goal is to forecast the values of appropriate traffic descriptors such as average travel time or speed, for one or more time intervals in the future. In this paper a novel and efficient short-term traffic prediction approach based on time series analysis is provided. Our idea is to split traffic time series into segments (that represent different traffic trends) and use different time series models on the different segments of the series. The proposed method was evaluated using historical GPS traffic data from the city of Berlin, Germany covering a total period of two weeks. The results show smaller traffic prediction error, in terms of travel time, with respect to two basic time series analysis techniques in the relevant literature.

A Degeneracy Framework for Graph Similarity

The problem of accurately measuring the similarity between graphs is at the core of many applications in a variety of disciplines. Most existing methods for graph similarity focus either on local or on global properties of graphs. However, even if graphs seem very similar from a local or a global perspective, they may exhibit different structure at different scales. In this paper, we present a general framework for graph similarity which takes into account structure at multiple different scales. The proposed framework capitalizes on the wellknown k-core decomposition of graphs in order to build a hierarchy of nested subgraphs. We apply the framework to derive variants of four graph kernels, namely graphlet kernel, shortest-path kernel, Weisfeiler-Lehman subtree kernel, and pyramid match graph kernel. The framework is not limited to graph kernels, but can be applied to any graph comparison algorithm. The proposed framework is evaluated on several benchmark datasets for graph classification. In most cases, the corebased kernels achieve significant improvements in terms of classification accuracy over the base kernels, while their time complexity remains very attractive.

Kernel Graph Convolutional Neural Nets

Graph kernels have been successfully applied to many graph classification problems. Typically, a kernel is first designed, and then an SVM classifier is trained based on the features defined implicitly by this kernel. This two-stage approach decouples data representation from learning, which is suboptimal. On the other hand, Convolutional Neural Networks (CNNs) have the capability to learn their own features directly from the raw data during training. Unfortunately, they cannot handle irregular data such as graphs. We address this challenge by using graph kernels to embed meaningful local neighborhoods of the graphs in a continuous vector space. A set of filters is then convolved with these patches, pooled, and the output is then passed to a feedforward network. With limited parameter tuning, our approach outperforms strong baselines on 7 out of 10 benchmark datasets.