*6.1. Limitations*

Although the tick-by-tick dataset used in this study is large, which strengthens the results of the experiments, the number of cryptocurrencies should ideally be higher so that we can draw firmer conclusions regarding the cryptocurrency market (e.g., whether the results generalise for large-cap and small-cap crypto assets). This will be the subject of future work. Secondly, while 30 min granularity has been found to suffice for our calculations, it would be better if we could use a lower level, say 15 min or even finer. Unfortunately, some cryptocurrencies are not traded regularly causing a lot of missing values at these timescales. This will also be the subject of future work.

There is also a concern with respect to the use of Pearson correlation for clustering problems. In particular, although this correlation metric has been applied widely in the existing literature and proposed various findings in the financial markets [2,21,22], it is sensitive to outliers [58] and cannot capture non-linear relationships that might cause misleading results [25]. Consequently, this adversely affects the clustering results. Indeed, these issues are also observed in other correlation metrics such as Spearman and Kendall [58]. Furthermore, we noticed that the results of clustering vary significantly by using different correlation measuring methods. Thus, it is necessary to deeply investigate different methods for a specific research task and analyze the results from each of these methods. Indeed, the creation of new approaches for calculating correlation coefficients that overcome the current limitations needs to receive more attention.
