BP neural networks and random forest models to detect damage by Dendrolimus punctatus Walker

The construction of a pest detection algorithm is an important step to couple"ground-space"characteristics,which is also the basis for rapid and accurate monitoring and detection of pest damage.In four experimental areas in Sanming City,Jiangle County,Sha County and Yanping District in Fujian Province,sample data on pest damage in 182 sets of Dendrolimus punctatus were collected.The data were randomly divided into a training set and testing set,and five duplicate tests and one eliminating-indicator test were done.Based on the characterization analysis of the host for D.punctatus damage,seven characteristic indicators of ground and remote sensing including leaf area index,standard error of leaf area index(SEL)of pine forest,normalized difference vegetation index(NDVI),wetness from tasseled cap transformation(WET),green band(B2),red band(B3),near-infrared band(B4)of remote sensing image are obtained to construct BP neural networks and random forest models of pest levels.The detection results of these two algorithms were comprehensively compared from the aspects of detection precision,kappa coefficient,receiver operating characteristic curve,and a paired t test.The results showed that the seven indicators all were responsive to pest damage,and NDVI was relatively weak;the average pest damage detection precision of six tests by BP neural networks was 77.29%,the kappa coefficient was 0.6869 and after the RF algorithm,the respective values were 79.30%and 0.7151,showing that the latter is more optimized,but there was no significant difference(p>0.05);the detection precision,kappa coefficient and AUC of the RF algorithm was higher than the BP neural networks for three pest levels(no damage,moderate damage and severe damage).The detection precision and AUC of BP neural networks were a little higher for mild damage,but the difference was not significant(p>0.05)except for the kappa coefficient for the no damage level(p<0.05).An"over-fitting"phenomenon tends to occur in BP neural networks,while RF method is more robust,providing a detection effect that is better than the BP neural networks.Thus,the application of the random forest algorithm for pest damage and multilevel dispersed variables is thus feasible and suggests that attention to the proportionality of sample data from various categories is needed when collecting data.

Comparison of electrochemiluminescence and ELISA methods in the detection of blood borne pathogens in Gabon

Objective: To assess the performances of Cobas 6000 e601 and EVOLIS Bio Rad in the detection of HIV, HBV and HCV in blood donors in Libreville(Gabon).Methods: A cross-sectional investigation was conducted in July 2017 in a total of 2 000 blood donors recruited at the National Blood transfusion Center, Libreville Gabon.Among them, 363 donors were selected to compare the performances of COBAS 6000 e601(electro-chemiluminescence) and EVOLIS Bio Rad in detecting HIV, HBV and HCV using Cohen's kappa coefficient.Results: Both methods yielded similar results for the detection of HIV and HBs Ag. A very good agreement of 93.39% and an excellent agreement of 98.90% were obtained for the detection of HIV and Hbs Ag, with kappa values of 0.80 and 0.98, respectively. The observed agreement of 91.86% was found for the detection of HCV, which gave a fair agreement between the two methods with kappa = 0.33.Conclusions: The two evaluation methods showed a similar performance in the detection of HIV, HBV. However, given the high rate of intra and inter-genotypes recombination known for HIV and HBV, more robust techniques of detection such as polymerase chain reaction should be used to prevent post-transfusion contaminations.

Assessing and mapping soil erosion risk zone in Ratlam District, central India

Evaluation of physical and quantitative data of soil erosion is crucial to the sustainable development of the environment. The extreme form of land degradation through different forms of erosion is one of the major problems in the sub-tropical monsoon-dominated region. In India, tackling soil erosion is one of the major geo-environmental issues for its environment. Thus, identifying soil erosion risk zones and taking preventative actions are vital for crop production management. Soil erosion is induced by climate change, topographic conditions, soil texture, agricultural systems, and land management. In this research, the soil erosion risk zones of Ratlam District was determined by employing the Geographic Information System(GIS), Revised Universal Soil Loss Equation(RUSLE), Analytic Hierarchy Process(AHP), and machine learning algorithms(Random Forest and Reduced Error Pruning(REP) tree). RUSLE measured the rainfall eosivity(R), soil erodibility(K), length of slope and steepness(LS), land cover and management(C), and support practices(P) factors. Kappa statistic was used to configure model reliability and it was found that Random Forest and AHP have higher reliability than other models. About 14.73%(715.94 km^(2)) of the study area has very low risk to soil erosion, with an average soil erosion rate of 0.00-7.00×10^(3)kg/(hm^(2)·a), while about 7.46%(362.52 km^(2)) of the study area has very high risk to soil erosion, with an average soil erosion rate of 30.00×10^(3)-48.00×10^(3)kg/(hm^(2)·a). Slope, elevation, stream density, Stream Power Index(SPI), rainfall, and land use and land cover(LULC) all affect soil erosion. The current study could help the government and non-government agencies to employ developmental projects and policies accordingly. However, the outcomes of the present research also could be used to prevent, monitor, and control soil erosion in the study area by employing restoration measures.

Probabilistic Methods in Multi-Class Brain-Computer Interface

Abstract-Two probabilistic methods are extended to research multi-class motor imagery of brain-computer interface （BCI）： support vector machine （SVM） with posteriori probability （PSVM） and Bayesian linear discriminant analysis with probabilistic output （PBLDA）. A comparative evaluation of these two methods is conducted. The results shows that： 1） probabilistie information can improve the performance of BCI for subjects with high kappa coefficient, and 2） PSVM usually results in a stable kappa coefficient whereas PBLDA is more efficient in estimating the model parameters.