Decision tree (DT) algorithms have been applied for classification and change detection in various geospatial studies and more recently, for urban expansion and land use/land cover (LULC) change modeling. However, these studies have not elaborated on specification of DT algorithms regarding data sampling, predictor variables, model configuration, and model evaluation. The focus of this study is to explore several balanced and unbalanced sampling methods, various predictor variables, different configurations of stopping rules, and reliable evaluation metrics to enhance the performance of classification and regression tree (CART), one of the most efficacious DT algorithms, for urban expansion modeling. The implementation of the model in the Triangle Region, North Carolina (NC) State, over the period of 2001 to 2011 demonstrates a striking performance with the training accuracy of 97%, the testing accuracy of 94%, and the Kappa value of 0.80. This performance was achieved using a training dataset containing all changed land cells and three times of that randomly selected from unchanged land cells and regulating the minimum number of records in a leaf node equal to 1, the minimum number of records in a parent node equal to 2, and the value of 10,000 for the maximum number of splits. The CART DT algorithm indicates that proximity to built areas, proximity to highways, current LULC type, elevation, and distance to water bodies are the most significant predictor variables for the urban expansion prediction in the study area.
The optimization of the total annual cost in heat exchanger networks has been one of the overarching goals when synthesizing these networks. Several methodologies and techniques have been developed to achieve optimal costs in mixed material heat exchanger networks. This paper demonstrates the application of two decomposition methodologies (total decomposition and partial decomposition) for typical cost rules. The objective function was defined as the optimization and minimization of the total annual cost in mixed materials heat exchanger network. Three optimization algorithms, hybrid genetic‐particle swarm optimization (GA‐PSO), shuffled frog leaping algorithm (SFLA) techniques, and ant colony optimization (ACO), were used to further optimize the total cost in mixed materials heat exchanger network. The results indicate that the total annual cost in partial decomposition method was smaller than that in full integration method and total decomposition method. The reduction of the total annual cost was about 27% for GA‐PSO algorithm, 24% for SFLA and 10% for ACO relative to the results reported in this work. In partial decomposition method, at least one mixed material of heat exchanger was used to reduce the hot and cold utility for decreasing the total annual cost. Partial decomposition method resulted in the highest reduction of the total annual cost compared with other methods. Percentage of difference of the total annual cost were 0.36%, 1.92%, and 5.05% for full integration, total decomposition, and partial decomposition methods, respectively, in comparison with the previous studies. Results have been compared with the results of other studies to demonstrate the accuracy of the applied algorithms. 相似文献
A rapid, sensitive, and reproducible high-performance liquid chromatographic procedure for the determination of nine biogenic amines in non-alcoholic beers was developed by an optimized benzoylation procedure. A Plackett–Burman factorial design was used in order to screen the statistically significant variables. The significant factors of biogenic amine benzoylation, reagent volume and pH, were optimized by a complete factorial response surface design, and optimal reaction conditions were generated. The optimized method showed good linearity (correlation coefficients > 0.997) and good recoveries (from 88.6 to 104.7 %). The repeatability and reproducibility of method were >3.9 and >4.6 %, respectively. Moreover, the detection limits of biogenic amines were calculated between 0.05 and 0.15 μg/ml in wine samples. The optimized method has been applied to the determination of biogenic amine contents of non-alcoholic beers consumed in Iran. Their values ranged from 0 to 2.56 mg/l, no significant differences (p?>?0.05) were observed between the analyzed samples, and none of these samples surpass the toxic levels reported in the literature. 相似文献
Nanocrystalline ZnO and Mn (1 wt.%)-doped ZnO particles have been synthesized via reverse micelle method. The structural, particulate, and optical properties of the synthesized nanoparticles have been studied by XRD, TEM, UV-Vis, and PL spectroscopy. The obtained data indicate the synthesis of the pure nanoparticles structure with wurtzite structure, average particle size of 18-21 nm, and high optical quality. Gas sensing properties of the nanocrystalline ZnO and Mn-doped ZnO particles toward gasoline and ethanol vapors have been investigated at different temperatures and concentrations. The results show that the optimum working temperature of the gas sensors based on ZnO and Mn-doped ZnO particles are about 633 and 620 K toward ethanol vapor and about 560 and 608 K toward gasoline vapor, respectively. Based on the results, although Mn impurities reduce the sensitivity of the ZnO gas sensor, they cause sensor to saturate at much higher gas concentration. 相似文献
The crystallization kinetics of Sn40Se60 thin films has been successfully investigated using sheet resistance versus temperature measurements. Thermal evaporation was used to deposit the films on ordinary glass substrates. The crystallization temperature for Sn40Se60 thin film was found to be 156.6 ± 0.3 ℃. In the as-deposited state, the sheet resistance was found to be 195 MΩ, this value declined to 1560 Ω/口 upon annealing. The value of activation energy obtained from the Kissinger plot was 0.62 ± 0.07 eV. From the results obtained, Sn40Se60 is a promising alloy for PCM application because of its high electrical contrast, high crystallization temperature, and relatively high activation energy. 相似文献