Prediction Maps & Validation using Logistic Regression & ROC
In the this course, i have shared complete process (A to Z ) based on my published articles, about how to evaluate and compare the results of applying the multivariate logistic regression method in Hazard prediction mapping using GIS and R environment.
Since last decade, geographic information system (GIS) has been facilitated the development of new machine learning, data-driven, and empirical methods that reduce generalization errors. Moreover, it gives new dimensions for the integrated research field.
STAY FOCUSED: Logistic regression (binary classification, whether dependent factor will occur (Y) in a particular places, or not) used for fitting a regression curve, and it is a special case of linear regression when the output variable is categorical, where we are using a log of odds as the dependent variable.
Why logistic regression is special? It takes a linear combination of features and applies a nonlinear function (sigmoid) to it, so it’s a tiny instance of the neural network!
In the current course, I used experimental data that consist of : Independent factor Y (Landslide training data locations) 75 observations; Dependent factors X (Elevation, slope, NDVI, Curvature, and landcover)
I will explain the spatial correlation between; prediction factors, and the dependent factor. Also, how to find the autocorrelations between; the prediction factors, by considering their prediction importance or contribution. Finally, I will Produce susceptibility map using; R studio and ESRI ArcGIS only. Model prediction validation will be measured by most common statistical method of Area under (AUC) the ROC curve.
At the the end of this course, you will be efficiently able to process, predict and validate any sort of data related to natural sciences hazard research, using advanced Logistic regression analysis capability.
Keywords: R studio, GIS, Logistic regression, Mapping, Prediction