Prediction of river flow is one of the main issues in the field of water resources management. Because of the complexity of the rainfall-runoff process, data-driven methods have gained increased importance. In the current study, two newly developed models called Least Square Support Vector Regression (LSSVR) and Regression Tree (RT) are used. The LSSVR model is based on the constrained optimization method and applies structural risk minimization in order to yield a general optimized result. Also, in the RT, data movement is based on laws discovered in the tree. Both models have been applied to the data in the Kashkan watershed. Variables include (a) recorded precipitation values in the Kashkan watershed stations, and (b) outlet discharge values of one and two previous days. Present discharge is considered as output of the two models. Following that, a sensitivity analysis has been carried out on the input features and less important features have been diminished, so that both models have provided better prediction on the data. The final results of both models have been compared. It was found that the LSSVR model has better performance. Finally, the results present these models as suitable models in river flow forecasting.

1. Introduction: From 1960s several attempts have been made to measuring the rock brittleness index BI. Schwartz (1964) using results of a series of triaxial tests on rock samples, stated that the rock’ s behavior from frangibility to ductility happens in 4. 3 ratios of principal stresses. Altindag (2002; 2003) introduced a new method for prediction of the BI by the division of the uniaxial compressive strength (UCS) of the rock to Brazilian tensile strength (BTS). In the late 1960s punch penetration test (PPT) introduced by Handewith (1971) to measure some physical properties of rock sample related to hardness and toughness of rock. Yagiz (2006) stated that the PPT’ s results for measuring the BI have a very high correlation with TBM penetration rate. Although the PPT has very delightful results, application of this test is very expensive and needs much time as well...

Cox proportional hazards models are the most common modelling framework to prediction and evaluation of co-variate e ects in time-to-event analyses. These models usually do not account the relationship among covari-ates which may have impacts on survival times. In this article, we introduce regression tree models for survival analyses by incorporating dependencies among covari-ates. Various properties of the proposed model are stud-ied in details. To assess the accuracy of the proposed model, a Monte{Carlo simulation study is conducted. A real data set from assay of serum free light chain is also analysed to illustrate advantages of the proposed method in medical investigations.

In this paper, regression tree ensembles (bagged and boosted) have been utilized in predicting atomic coordinate of Carbone nanotubes (CNTs). The aim of this study is to use ensembles classifiers to compute the atomic coordinates of Carbone nanotubes rather than other simulation tools. The dataset we used in this paper are provided by the UCI Repository of Machine Learning and it has a total of (10721) instances with (8) attributes (five as inputs and three as outputs) and it has no missing data. Various performance measures are also calculated to evaluate the classifiers we employed. The results show that there is a slight difference in performance between bagged and boosted trees, however, they are preferable classifiers for carbon atom coordinates prediction due to their high accuracy and short computation time. Using these predicted atomic coordinates as early coordinates for the simulation tool, the actual atomic coordinates can be retrieved in minutes or seconds instead of days by minimizing the iterations in the computation process.

Objectives: According to health surveys, population growth and total fertility rate (TFR) are decreasing in Iran. The economic and social factors in addition to the changing values and attitudes in the Iranian society have had a major impact on fertility decisions and the actions of families, especially women towards childbearing. This is an important issue for policymakers and many researchers in demography and public health thus the investigation of factors that affect low TFR is considered as a necessity. Materials and Methods: The classification and regression trees (CART) algorithm, as one of the most applicable classification trees, along with logistic regression was applied to model the tendency of 4898 women for childbearing in provinces with a TFR lower than the replacement level in Iran. The secondary data were then analysed by SPSS version 24. 0. Results: Based on these two approaches, it was concluded that despite the CART algorithm, logistic regression suffers from some shortcomings including the difficult interpretation of three levels of interactions while not containing a specific method for handling the outliers. In addition, CART results demonstrated that women’ s children ever born (CEB), age, and opinion had significant impacts on their desire to have a child. The groups encompassing “ 10-39-year-old women with CEB≤ 2” and “ 40-49-year-old women with positive attitudes towards childbearing” desired to have more children while “ women with CEB ≥ 3” showed no tendency for childbearing. Conclusions: In general, the results revealed that adopting policies for changing women’ s views on childbearing and creating the necessary resources for preventing the delays in marriage are regarded as important actions toward altering fertility rates. Another important conclusion is applying the CART algorithm as a convenient method for classifying demographical data.

Introduction: Early detection of osteoporosis is a key to preventing of it; but recognition, without the use of appropriate diagnostic methods, due to the complexity of risk factors and gradual bone loss process, is problem. The purpose of this study is to develop and efficiency evaluation a predictive model of osteoporosis using decision tree technique as a diagnostic method based on available risk factors; thereby to identify individuals at risk for preventive activities. Methods: In this study used data from 131 women aged 20 – 40 years. Response variable was amount of BMD (t-score) L1-L4 lumbar region that divided on two group, normal (t-score>=-1) and at risk of osteoporosis (t-score<-1). To determine risk factors of osteoporosis used from decision tree model with method of k-fold cross validation k=4 and logistic regression. To assess the accuracy prediction of two model, the area under receiver operative characteristic curves (AUROC) was used. Data analysis was performed by R software. Results: Three variables number of pregnancies, BMI and calcium levels as risk factors for osteoporosis were obtained from the decision tree model and Area under receiver operative characteristic decision tree and logistic regression, respectively 0. 665 and 0. 686 were obtained. Conclusion: Area under receiver operative characteristic curve showed advantage superiority of logistic regression that according to advantages of the decision tree applying simultaneously of two models is recommended.

Due to the drought and land use changes in recent years, the phenomenon of storm dust in Iran is increasing as a dangerous environment. Dust influences climate change and human health, causing serious damage. The subject of this research is to identify and prepare a map of sensitivity of dust source area for controlling and determining the role of each of the factors affecting its occurrence using the regression tree data mining model (BRT) in eastern Iran. For this purpose, at first 147 dust source area were identified in the region and divided into two groups for modeling and evaluation. According to the studies, eight effective factors including land use, geology, slope degree, elevation, normalized vegetation index (NDVI), distance from the river, wind speed and rainfall were identified and the layers of these factors were prepared in GIS environment. To evaluate the results, the ROC curve was used. The results showed that the BRT model with the area under the curve (79. 6) had a good performance in producing a dust sensitivity map in the study area. Based on the results of the model, vegetation index, elevation and slope had the most impact on the occurrence of dust in the region.

A debutanizer column is an integral part of any petroleum refinery. Online composition monitoring of debutanizer column outlet streams is highly desirable in order to maximize the production of liquefied petroleum gas. In this article, data-driven models for debutanizer column are developed for real-time composition monitoring. The dataset used has seven process variables as inputs and the output is the butane concentration in the debutanizer column bottom product. The input– output dataset is divided equally into a training (calibration) set and a validation (testing) set. The training set data were used to develop fuzzy inference, adaptive neuro fuzzy (ANFIS) and regression tree models for the debutanizer column. The accuracy of the developed models were evaluated by simulation of the models with the validation dataset. It is observed that the ANFIS model has better estimation accuracy than other models developed in this work and many data-driven models proposed so far in the literature for the debutanizer column.

Introduction: Soil erosion is one of the most important and serious threats to food security and as a consequence of human life. In order to perform soil protection activities against soil erosion, knowledge about the amount of soil loss tolerable is very important. In fact, the soil loss tolerable is the potential for soil erosion, loss of productivity and lost production, and the final criterion for controlling soil erosion and degradation of land. Soil thickness methods, particularly Skidmore equation, based on their ability to estimate the tolerable amount of soil loss have been widely used. In the mathematical function developed by Skidmore based on soil thickness, the soil loss tolerable is calculated based on the soil's current depth, the lowest and maximum soil depth for sustained growth of crops, and the upper limit of tolerable erosion in accordance with the environment. Since the determination of soil loss tolerance by soil thickness method and the Skidmore equation requires time, cost and energy, the researchers have tried to estimate the soil tolerance is supported by regression methods using pedotransfer functions and easily available soil properties. Therefore, the present study was carried out with the aims of determining the tolerable tolerance of soil loss by thickness method and the development of regression pedotransfer functions for estimating this property in the upstream of the dam. Materials and Methods: The study is place on Kamfiruz Watershed with an area of 422 km2, an average annual precipitation of 443 mm and an average annual temperature of 14 ° C. It is closed to the Dorudzan Dam sub-basins and is considered as one of the five parts of Marvdasht plain in Fars province. For this work, 60 soil profiles were excavated by excavating machine. In addition to measuring the depth of soil, some physico-chemical soil properties were measured from the surface layer (0-30 cm) including; soil texture, organic matter, salinity, percentage calcium carbonate, mean weight diameter in the laboratory and filed. In order to develop regression models for estimating the tolerable soil loss, information from 60 soil profiles was divided into two data-sets. One set of the data with 42 samples (70% of whole samples) was used for developing the models and another set of the data with 18 soil samples (30% of whole samples) was used for validation. Multiple linear regression was used to develop the linear models. The same soil properties used in the multiple regression method were considered as inputs in the tree regression method to estimate the tolerable amount of loss. Results and Discussion: The results showed that the minimum and maximum Z1 parameters (the lowest soil depth for stable growth of crops in the study area) were considered as 0. 25 and 0. 51 m based on the current depth of soil. Organic matter of the soils with the highest standardized coefficient (Beta = 0. 44) and the highest correlation (-0. 77) with soil loss tolerance was the most important soil properties for estimating the soil loss tolerance. In the regression model, only the coefficients of four characteristics of permeability, soil aggregate stability, pH and organic matter appeared among the soil grazing characteristics and entered into the model. Based on the evaluation statistic, tree regression method with the highest determination coefficient in both calibration data sets (R2 = 0. 96) and validation (R2 = 0. 78) and the lowest error value in the validation data (RMSE= 0. 29 ton ha-1 year-1) and validation (RMSE = 0. 125 ton ha-1 year-1) were more efficient than the multiple regression method in estimating the tolerable soil loss. Conclusion: Soil loss tolerance was estimated using regression methods (multiple linear regression and regression tree) in Doroudzan Watershed, Fars province. The soil loss tolerable determined using Skidmore method, was 1. 04 tons per hectare per year ranging from 0. 29 to 2. 25 ton ha-1 year-1. The soils of this area are slightly deep and their depth varies from 0. 4 m in the marginal areas in the upstream parts of the catchment area of the dam and the slope of mountain up to 2 meters in the center of the plain with agricultural lands uses. In general, the tree regression method had a better performance than linear regression method for estimating the soil loss tolerance based on the statistical indices.