Time-Series in Hyper-parameter Initialization of Machine Learning Techniques Tom´aˇ s Horv´ ath 1,4(B ) , Rafael G. Mantovani 2 , and Andr´ e C. P. L. F. de Carvalho 3 1 Faculty of Informatics, ELTE - E¨otv¨os Lor´and University, P´azm´anyP´ eter s´ et´any 1/C, Budapest 1117, Hungary tomas.horvath@inf.elte.hu 2 Federal Technology University - Paran´a, Campus of Apucarana, Rua Marc´ ılio Dias, 635 - Jardim Para´ ıso, Apucarana, PR 86812-460, Brazil rafaelmantovani@utfpr.edu.br 3 Institute of Mathematical and Computer Sciences, University of S˜ao Paulo, Avenida Trabalhador S˜ao Carlense, 400 - Centro, S˜ao Carlos, SP 13566-590, Brazil andre@icmc.usp.br 4 Institute of Computer Science, Faculty of Science, Pavol Jozef ˇ Saf´ arik University, Jesenn´a 5, 040 01 Koˇ sice, Slovakia tomas.horvath@upjs.sk Abstract. Initializing the hyper-parameters (HPs) of machine learning (ML) techniques became an important step in the area of automated ML (AutoML). The main premise in HP initialization is that a HP set- ting that performs well for a certain dataset(s) will also be suitable for a similar dataset. Thus, evaluation of similarities of datasets based on their characteristics, named meta-features (MFs), is one of the basic tasks in meta-learning (MtL), a subfield of AutoML. Several types of MFs were developed from which those based on principal component analysis (PCA) are, despite their good descriptive characteristics and relatively easy computation, utilized only marginally. A novel approach to HP initialization combining dynamic time warping (DTW), a well- known similarity measure for time series, with PCA MFs is proposed in this paper which does not need any further settings. Exhaustive experi- ments, conducted for the use-cases of HP initialization of decision trees and support vector machines show the potential of the proposed app- roach and encourage further investigation in this direction. Keywords: Automated ML · Metalearning · PCA · DTW 1 Introduction The growing popularity of machine learning (ML) in various application domains and the shortage of data scientists has raised the demand for automated ML (AutoML) [7]. A special focus of AutoML is on configuring the hyper-parameters c  Springer Nature Switzerland AG 2021 H. Yin et al. (Eds.): IDEAL 2021, LNCS 13113, pp. 246–258, 2021. https://doi.org/10.1007/978-3-030-91608-4_25