1,721,063 research outputs found
Using interactions and dynamics for mining groups of moving objects from trajectory data
Full text linkAdvances in tracking technology enable the gathering of spatio-temporal data in the form of trajectories, which when analysed can convey useful knowledge. In particular, discovering groups of moving objects is a valuable means for a wide class of problems related to mobility. The task of group mining has been investigated by considering mostly the spatial closeness and similarity of the trajectories, while little attention has been paid to the relationships between the trajectories and time-changing nature of the trajectories. The relationships may provide evidence of interactions between the moving objects. The time-changing nature may provide evidence of dynamics of the movements. Therefore, interactions and dynamics can be sources of information to be considered in order to discover new forms of groups. Motivated by this, we introduce the concept of crews and propose a method to discover crews. A crew gathers moving objects with similar interactions and similar dynamics. The proposed method relies on i) new movement parameters, which explicitly consider interactions and dynamics, and ii) a distance-free clustering algorithm, which groups objects based on the similarity of the movement parameters. We conduct extensive experiments, which include a quantitative evaluation of the quality of the crews and comparison with alternative solutions
Mining Temporal Associations Between Air Pollution and Effects on the Human Health
No full textThe task of monitoring and improving the urban air quality has attracted a great deal of interest both from national governments and scientific communities. In order to implement policies for the environmental protection, the recent
urban planning decisions are often based on the results produced by several research
fields. An important research direction aims at understanding the pollution phenomenon by means of data mining approaches, which support decision makers with
information extracted directly from data. In this work we investigate the effect of air pollution on human health by taking into account the temporal variability of environmental data. Since the repercussions of air pollution on humans are perceived
only after a certain lapse of time, we propose to discover temporal associations
which relate a change at time tj of the population health conditions with a change
at time ti (tj>ti) of the polluting emissions. Information conveyed by discovered temporal associations could be exploited both to support policies for environmental
protection and to adopt strategies for the reduction of human health risks
Samples_for_NoDistance
No full textThis dataset contains the samples builts on 10 iterations with the version of the systems which use no weighting schem
Samples_for_DescriptiveandCollectiveDistances
No full textThis dataset contains the sampels builts on 10 iterations with the version of the system which combines Descriptive Distance and Collective Distance as weighting schem
Original datasets with the full training data
No full textThis dataset contains the original real-world data network
Samples_for_DescriptiveDistance
No full textThis dataset contains the samples built on 10 iterations with the version of the system which uses a Descriptive Distance weighting schem
Samples_for_CollectiveDistance
No full textThis dataset contains the samples built on 10 iterations with the version of the system which uses the collective distance weighting schem
Dataset_folder
No full textThe folder has the seventeen network datasets we used for a network regression task. The folder contains the files used for a 5-cross validation experiment replicated by 5 trials. The datasets are organized in 4 categories of files: namefile_arffTest_N (the testing file for the trial N-th), namefile_arffTrain_M (the training file for the trial M-th), namefile_arffTrain_P_arffSampleQ (the sample for Q-th trial at P-th fold). Also, a file of the distances of the edges is associated to each dataset
Dataset_folder
No full textThe folder has the seventeen network datasets we used for a network regression task. The folder contains the files used for a 5-cross validation experiment replicated by 5 trials. The datasets are organized in 4 categories of files: namefile_arffTest_N (the testing file for the trial N-th), namefile_arffTrain_M (the training file for the trial M-th), namefile_arffTrain_P_arffSampleQ (the sample for Q-th trial at P-th fold). Also, a file of the distances of the edges is associated to each dataset
- …
