Circulation Modeling Listed here is a set of notebooks detailing our OBOC circulation modeling efforts as of July 2017 — from raw data to modeling and visualization. All of this code is now available on our GitHub repository as well as the branch demographic data. The Chicago Library data we are not at liberty to share, so some of the files will require a password to run. Data management: Filtering and pre-processing of the demographic data. Filtering and pre-processing of the circulation data. Filtering and pre-processing the holdings data. Modeling Computing principal components from the demographic data. The components proved more accurate in circulation modeling than any individual dimension. Computing a multi-level linear model of branch circulation. We are modeling circulation at each branch as a function of the principal components and the branch holdings of the OBOC text, with a single-level of hierarchy — the book itself. Clustering Computing clusters of branches based on principal components. Various visualizations of the clusters. Visualization of circulation by branch by book with branches colored by cluster. Time series Visualizing the time course of the OBOC editions.