Tapped Delay Lines for GP Streaming Data Classification with Label Budgets
Streaming data classification requires that a model be available for classifying stream content while simultaneously detecting and reacting to changes to the underlying process generating the data. Given that only a fraction of the stream is ‘visible’ at any point in time (i.e. some form of window interface) then it is difficult to place any guarantee on a classifier encountering a ‘well mixed’ distribution of classes across the stream. Moreover, streaming data classifiers are also required to operate under a limited label budget (labelling all the data is too expensive). We take these requirements to motivate the use of an active learning strategy for decoupling genetic programming training epochs from stream throughput. The content of a data subset is controlled by a combination of Pareto archiving and stochastic sampling. In addition, a significant benefit is attributed to support for a tapped delay line (TDL) interface to the stream, but this also increases the dimensionality of the task. We demonstrate that the benefits of assuming the TDL can be maintained through the use of oversampling without recourse to additional label information. Benchmarking on 4 dataset demonstrates that the approach is particularly effective when reacting to shifts in the underlying properties of the stream. Moreover, an online formulation for class-wise detection rate is assumed, where this is able to robustly characterize classifier performance throughout the stream.
KeywordsStreaming data classification Non-stationary Class imbalance Benchmarking
The authors gratefully acknowledge support from NSERC Discovery and CRD programs (Canada) and RUAG Schweiz AG (Switzerland) while conducting this research.
- 1.Atwater, A., Heywood, M.I.: Benchmarking Pareto archiving heuristics in the presence of concept drift: diversity versus age. In: ACM Genetic and Evolutionary Computation Conference, pp. 885–892 (2013)Google Scholar
- 2.Atwater, A., Heywood, M.I., Zincir-Heywood, A.N.: GP under streaming data constraints: a case for Pareto archiving? In: ACM Genetic and Evolutionary Computation Conference, pp. 703–710 (2012)Google Scholar
- 3.Behdad, M., French, T.: Online learning classifiers in dynamic environments with incomplete feedback. In: IEEE Congress on Evolutionary Computation, pp. 1786–1793 (2013)Google Scholar
- 4.Bifet, A., Read, J., Žliobaitė, I., Pfahringer, B., Holmes, G.: Pitfalls in benchmarking data stream classification and how to avoid them. In: Blockeel, H., Kersting, K., Nijssen, S., Železný, F. (eds.) ECML PKDD 2013, Part I. LNCS, vol. 8188, pp. 465–479. Springer, Heidelberg (2013) CrossRefGoogle Scholar
- 5.Cervantes, A., Isasi, P., Gagné, C., Parizeau, M.: Learning from non-stationary data using a growing network of prototypes. In: IEEE Congress on Evolutionary Computation, pp. 2634–2641 (2013)Google Scholar
- 6.Dempsey, I., O’Neill, M., Brabazon, A.: Foundations in Grammatical Evolution for Dynamic Environments. SCI, vol. 194. Springer, Heidelberg (2009) Google Scholar
- 7.Dempsey, I., O’Neill, M., Brabazon, A.: Survey of EC in dynamic environments (chap. 3). In: , pp. 25–54. Springer, Heidelberg (2009)Google Scholar
- 9.Fan, W., Huang, Y., Wang, H., Yu, P.S.: Active mining of data streams. In: Proceedings of SIAM International Conference on Data Mining, pp. 457–461 (2004)Google Scholar
- 14.Harries, M.: Splice-2 comparative evaluation: electricity pricing. Technical report, University of New South Wales (1999)Google Scholar
- 15.Heywood, M.I.: Evolutionary model building under streaming data for classification tasks: opportunities and challenges. Genet. Program. Evolvable Mach. (2015). doi: 10.1007/s10710-014-9236-y
- 16.Lindstrom, P., MacNamee, B., Delany, S.J.: Drift detection using uncertainty distribution divergence. Evol. Intel. 4(1), 13–25 (2013)Google Scholar
- 20.Vahdat, A., Atwater, A., McIntyre, A.R., Heywood, M.I.: On the application of GP to streaming data classification tasks with label budgets. In: ACM Genetic and Evolutionary Computation Conference: ECBDL Workshop, pp. 1287–1294 (2014)Google Scholar