Dataset Selections: How to rapidly assess differences in datasets and identify areas of interest for decision makers

Keywords: Statistics, Unsupervised Machine Learning, Human Population Modeling, Feature Selection, Data Selection
Abstract Type: Paper Abstract

Abstract

The broad goal for human population modeling is to increase the knowledge of where human populations are across any given geography. High resolution human population presence datasets are invaluable datasets to various communities including disaster relief, economic planning, and climate adaptation planning, to name a few. The creation of high-resolution human population models or for users of these modeled data, the need to compare alternative datasets are crucial in decision making. Mechanisms to compare and make final dataset selections involve various forms of qualitative and quantitative reviews. Given the defined purpose of the review, these efforts can be time intensive. One mechanism for comparing datasets involves situational awareness of the differences between datasets. These differences can be described through a series of statistical descriptors. In addition, outlier detection methods can be used to identify areas of significant differences for decision makers to focus on. Here the author shows an automated approach for the comparison of multiple raster datasets. The proposed methods address many of the constraints that decision makers are faced with, in qualitative and quantitative reviews of datasets, by reducing the amount of time to conduct reviews and identify areas of interest.