Universal approach for unsupervised classification of univariate data

Unsupervised machine learning, and in particular data clustering, is a powerful approach for the analysis of datasets and identification of characteristic features occurring throughout a dataset. It is gaining popularity across scientific disciplines and is particularly useful for applications without a priori knowledge of the data structure. Here, we introduce a universal approach for unsupervised data classification relevant for any dataset consisting of a series of univariate measurements. It is therefore ideally suited for a wide range of measurement types. Here, we apply it to the field of nanoelectronics and spectroscopy to identify meaningful structures in data sets, providing physically relevant information about the system under study. An important step in our approach is the guidelines for the estimation of the optimum number of clusters. In addition, we have performed an extensive benchmark of ML approaches found in literature for the classification of molecular break junction traces. We find that several feature space construction methods we have introduced and clustering algorithms yield accuracies up to 20% higher than methods reported so far, increasing the Fowlkes-Mallows index from 0.77 up to 0.91