Sampling for subtrajectory clustering

In this thesis, we delve into an approximate solution for the NP-Hard problem of subtrajectory clustering. This problem involves organising subsegments of input trajectories into clusters based on their similarities, with each cluster represented by a pathlet, that is a sequence of spatial points. The approximate solution considered in this thesis is based on analysing a sample of the input. The sample is analysed by using a recently proposed greedy algorithm for subtrajectory clustering. Theoretical notations regarding the problem are discussed, alongside practical experimentation using real-world datasets. Our testing focuses on comparing the pathlets obtained from the original input data with those derived from sampled versions of the input data. Additionally, we evaluate a method that speeds up execution time but may decrease the precision of the output. Through these investigations, we seek to provide valuable insights into the efficacy and applicability of this approach in addressing the complexities of subtrajectory clustering problem.