Machine Learning Techniques for Modelling Short Term Land-Use Change

Abstract

Abstract:The representation of land use change (LUC) is often achieved by using data-driven methods that include machine learning (ML) techniques. The main objectives of this research study are to implement three ML techniques, Decision Trees (DT), Neural Networks (NN), and Support Vector Machines (SVM) for LUC modeling, in order to compare these three ML techniques and to find the appropriate data representation. The ML techniques are applied on the case study of LUC in three municipalities of the City of Belgrade, the Republic of Serbia, using historical geospatial data sets and considering nine land use classes. The ML models were built and assessed using two different time intervals. The information gain ranking technique and the recursive attribute elimination procedure were implemented to find the most informative attributes that were related to LUC in the study area.The results indicate that all three ML techniques can be used effectively for short-term forecasting of LUC, but the SVM achieved the highest agreement of predicted changes.

Assessing the performance of 38 machine learning models: the case of land consumption rates in Bavaria, Germany

“…In our case, artificial neural networks also had only an average accuracy. Confirming previous studies (Samardžić-Petrović et al 2017), we also found that support vector machines (i.e. svmRadial) perform better than decision trees and artificial neural networks, though our study indicates that ensemble and boosting models can achieve better results.…”

“…This hypothesis is also supported by the very good performance of gam, which also does not model interactions. As we are not aware of a study similar to ours using LCR as an outcome, we discuss our findings in the general context of land-use modeling (Tayyebi and Pijanowski 2014, Samardžić-Petrović et al 2017, Shafizadeh-Moghadam et al 2017, Du et al 2018). While existing model comparisons consider rather similar models (i.e.…”

“…However, frequently only a single approach is applied (Azari et al, 2016, Huang et al 2010, Hagenauer and Helbich 2012, Linard et al 2013, Samardžić-Petrović et al 2016, Omrani et al 2019, often either artificial neural networks (Haykin 2009), random forests (Breiman 2001), and support vector machines (Scholkopf and Smola 2001). To support an evidence-based algorithmic selection, a small number of land-use studies have compared multiple ML algorithms (Rogan et al 2008, Tayyebi and Pijanowski 2014, Samardžić-Petrović et al 2017, Shafizadeh-Moghadam et al 2017. While supportive, these studies assessed the performance of only a small number of algorithms.…”

“…Multivariate adaptive regression splines and artificial neural networks were found to predict most accurately. Another study (Samardžić-Petrović et al 2017) concluded that support vector machines outperform their competitors, including neural networks and logistic regression (Huang et al 2010). While consensus is building that ML outperforms parametric models (e.g.…”

“…Previous studies formulate land-use change as a classification problem based on binary raster cells (Tayyebi and Pijanowski 2014, Samardžić-Petrović et al 2017, Shafizadeh-Moghadam et al 2017, however, research dealing with land-use change form a regressionbased perspective is virtually not existent. A special case of land-use change, which can analytically be explored on an area level, is the proportion of land consumption per municipality.…”