Weighting information layers using logistic regression for mine exploration applications
Hosseinall, Farhad, Reza Farajloo and Mohammed Ali Rajabi
Logistic regression can be used with spatial data as a predictor in a Geospatial Information System (GIS). Most of the other regression methods are based on the assumption that the independent variables are normally distributed which unfortunately is not always the case with the spatial data. However, this assumption is not needed in logistic regression. This paper describes an application of logistic regression for mapping the potential of existing copper in the Ali-Abad region (Taft, Yazd, Iran). The basic information layers used in this research include geological maps, geomagnetic, geoelectric and geochemical data. After preprocessing of these raw data useful information is extracted out and then integrated within a GIS using a logistic model. Borehole data is used to fit the model and find the coefficients as well as the weight of the spatial information layers. Then the model is used to determine the potential of existing copper in the other parts of the region for which the borehole information have not been used. The results show that as a data driven model logistic regression is comparable with some knowledge driven methods like Analytical Hierarchy Process (AHP) where experts decide on the importance and weight of the information. Layers weighted using both AHP and logistic regression methods are used with Index Overlay method to check the effect of the assigned weights. Predicting the minerals is the main goal of two approaches and both of them generate a mineral potential map. Logistic regression is successful in predicting the quality of minerals in most of the boreholes. While AHP strongly depends on the expert or experts and their ideas, logistic regression is relied only on the data. Being a data driven method where the relative importance of data are determined by data itself, and having no assumption on the normality of the independent variables are two main advantages of using logistic regression for prediction in a GIS environment.
Event: XXIII International FIG Congress : Shaping the change
Only personal, non-commercial use of this document is allowed.