Knowledge of Time-bin Data Selection using Gini Index based Type Classification in GitHub
Ayako Masuda, Tohru Matsuodani · Procedia Computer Science · 2022
The number of references to knowledge is considered one of the indicators for evaluating the usefulness of knowledge in open science. However, the value of knowledge is diverse and cannot be evaluated only by the number of references. We seek to determine indices of how engineers learn and grow from knowledge on GitHub based on knowledge references and their impact as characteristics of usefulness. Dynamic analysis requires the conversion of a variety of recorded data into time-series data and the selection of an appropriate method of time-series analysis. When selecting a method of time-series analysis, data classification is necessary. In this study, we focused on the differences in the type of time series to capture characteristics of usefulness of knowledge. We attempted to classify data by focusing on the bias of bin values at the stage of converting time-stamped recorded data to time bins. The data classification was based on the Lorenz curve, which is a measure of distortion from a uniform distribution, and used the Gini index and concentration ratio defined by us. The concentration ratio is an index of the partial features of the Lorenz curve. The results of applying this method confirmed that there is a type in the usefulness of knowledge. In this study, we present classification methods of this type as knowledge. The value sought by engineers appears to depend on the respective type of usefulness of the knowledge. The results of this study should serve as a basis for future analysis.