Statistical Depth for Text Data: An Application to the Classification of Healthcare Data

This manuscript introduces a new concept of statistical depth function: the compositional D-depth. It is first data developed exclusively for text data, in particular, those vectorized according to frequency-based criterion, such as tf-idf (term frequency–inverse document frequency) statistic, which results most vector entries taking value zero. The proposed consists considering inverse discrete Fourier transform fragments and then applying functional D. intended address problem sparsity numerical features resulting from transformation qualitative into quantitative common procedure natural language processing frameworks. Indeed, this hinders use traditional depths machine learning techniques classification purposes. In order demonstrate potential proposal, it applied real-world case study involves mapping Consolidated Framework Implementation Research (CFIR) constructs healthcare data. shown that DDG-classifier yields competitive outperforms all studied (logistic regression with LASSO regularization, artificial neural networks, decision trees, support machines) when used combination newly defined