174 Landslide Susceptibility Mapping with Data Mining Methods—a Case Study from Maily-Say, Kyrgyzstan Anika Braun, Tomas Fernandez-Steeger, Hans-Balder Havenith, and Almaz Torgoev Abstract Multiple factors, such as geology, high mountain topography, seismic activity, climatic conditions and mining activities cause significant landslide hazard in the region around Maily- Say, Kyrgyzstan. To assess the landslide susceptibility a database containing landslide information and geological, morphological and hydrological parameters associated with landslide occurrence was established and analyzed with different data mining algorithms. The most promising results were achieved with an Artificial Neural Network, a Bayesian Network and a Support Vector Machine. All three methods developed the ability to predict landslide occurrence and produced spatially reasonable results. Other models, such as CHAID Decision Tree and Logistic Regression developed only poor ability for landslide prediction and the results were from a geologically point of view not plausible. The C5.0 Decision Tree almost perfectly predicted landslide occurrence, however it is most likely overfit to the data and would only have a poor ability to generalize the prediction on new datasets. In general the method proved to be useful for the analysis of landslide susceptibility in remote regions where landslide occurrence is related to multiple factors; it also allowed us to extract a maximum of information from a relatively simple dataset. Keywords Landslides Á Susceptibility Á Data mining 174.1 Introduction A landslide susceptibility analysis was carried out for Maily- Say, a former uranium mining town in Kyrgyzstan, using a data mining approach. Several large scaled landslides have already claimed fatalities, caused severe damage to housing and infrastructure or induced floods along the main river in the Maily-Say valley. Landslides also threaten radioactive tailing ponds, which are distributed around the town. Destabilization of those tailings by landslides or flooding bears the potential of a major environmental catastrophe. 174.1.1 Setting Maily-Say is located in the North of the Ferghana Valley within the foothills of the Tien Shan high mountain belt. The region is highly prone to landslides due to steep slopes, high seismo-tectonic activity, the presence of soft sedimentary rocks and climatic conditions causing high run off associated with snow melt and intense precipitations in spring. Another important factor causing slope failures was the extensive mining activity between 1946 and 1968 resulting in collapse of underground galleries, rock weakening, groundwater rise after the mining activities and an intensified land use due to the machines and growing population. A. Braun (&) Á T. Fernandez-Steeger Department of Engineering Geology and Hydrogeology, RWTH Aachen University, Lochnerstraße 4–20, 52064, Aachen, Germany e-mail: braun@lih.rwth-aachen.de H.-B. Havenith Á A. Torgoev Georisks and Environment, Department of Geology, University of Liège, B20 Sart Tilmann, 4000, Liège, Belgium G. Lollino et al. (eds.), Engineering Geology for Society and Territory – Volume 2, DOI: 10.1007/978-3-319-09057-3_174, © Springer International Publishing Switzerland 2015 995