data mining algorithms

Mass movement, according to their nature, variety, hazards for human lives, and properties, have always been a matter of interest to various scholars. Considering that the occurrence of this phenomenon has a complex mechanism and complex factors and variables can affect it, extensive studies to identify the effective factors, classification, zoning, and modeling of this process have been conducted. In this study, landslides of three watersheds in the southeast of Neishabour city were investigated and the hazard zonation map was prepared, using bivariate statistical methods of the information value and area density. There are few studies regarding the application of different data mining methods to determine the effective variables in the occurrence of landslides and most studies are based on other statistical methods. Data mining is called as knowledge discovery in databases and is a way to discover new and beneficial information through a lot of data. Some of the most important data mining algorithms include the decision tree, random forest, boosting aggregate demand, support vector machine, logistic regression, and neural network algorithm. The data mining extracts useful information from large volumes of data and has shown a good performance. Therefore, the aim of the present study was to prioritize

The present study aimed to investigate the factors affecting the occurrence of a landslide and its zoning in three watersheds including Kharv, Harimabad and Grineh watersheds in the Razavi Khorasn province. First, 99 landslides were identified in the area and the landslide distribution map was prepared. Then, all effective factors on watershed landslides, in 15 information layers including the altitude, slope, aspect, climate, land use, pedology, vegetation cover, geology, evaporation, temperature, rainfall, land type, distance from road, distance from fault, and distance from river were digitized in the ArcGIS environment. Then, using data-mining algorithms in R software, the preferable algorithm and effective factors on landslide occurrence, were introduced. Finally, the landslide hazard zonation in the GIS software was done using bivariate statistical models.

The results showed that the random forest algorithm with an accuracy of 92% is the best one and the variables of geology, climate, aspect, distance from road, altitude, pedology and land type are the most important variables in algorithms modeling. The most probability of occurrence of watershed landslides placed in areas with west and northwest directions, slopes higher than 30 degrees, dominant type of the environmental factors affecting the occurrence of a landslide including the altitude, slope, aspect, climate, land use, pedology, vegetation cover, geology, evaporation, temperature, rainfall, land type, distance from road, distance from fault, and distance from river using data mining algorithms, zoning its sensitivity, and bivariate statistical models of information value and area density in three watersheds including Kharv, Harimabad, Grineh watersheds in Razavi Khorasan province.   mountains, the semi-humid climate, 1500 to 2000 mm evaporation class, entisols, dense vegetation, the gardens, bushes and shrubs land uses, being close to the roads and faults and being far from the rivers, and the altitudes of 2000 to 2500 m with the phyllite, boulders and sandstone formations. The results of the zoning map evaluation using the information value and density area methods showed that 45.45% and 55.55 % of landslides were respectively located at the high and very high risk zones and the rest were in very low, low, and moderate risk zones. As a result, in both methods, most of landslides were in the high and very high risk zones that indicated the suitable accuracy of the model.

According to the results of this research, variables including the geology, climate, aspect, distance from road, altitude, soil science, and land type were considered as the most important factors in the occurrence of a landslide. In addition, factors such as slope, land use, vegetation cover, distance from fault and distance from river were identified as the most important factors influencing the development of landslide and classified as natural factors, which could be influenced by human factors. The comparison of two mentioned methods showed that the area density method was more appropriate than the information value method for the study area.