SMOOTH ESTIMATION OF ROC CURVE IN THE PRESENCE OF AUXILIARY INFORMATION

Receiver operating characteristic （ROC） curve is often used to study and compare two- sample problems in medicine. When more information may be available on one treatment than the other, one can improve estimator of ROC curve if the auxiliary population information is taken into account. The authors show that the empirical likelihood method can be naturally adapted to make efficient use of the auxiliary information to such problems. The authors propose a smoothed empirical likelihood estimator for ROC curve with some auxiliary information in medical studies. The proposed estimates are more efficient than those ROC estimators without any auxiliary information, in the sense of comparing asymptotic variances and mean squared error （MSE）. Some asymptotic properties for the empirical likelihood estimation of ROC curve are established. A simulation study is presented to demonstrate the performance of the proposed estimators.

Kernel Estimators of the ROC Curve with Censored Data

Receiver operating characteristic （ROC） curves are often used to study the two sample problem in medical studies. However, most data in medical studies are censored. Usually a natural estimator is based on the Kaplan-Meier estimator. In this paper we propose a smoothed estimator based on kernel techniques for the ROC curve with censored data. The large sample properties of the smoothed estimator are established. Moreover, deficiency is considered in order to compare the proposed smoothed estimator of the ROC curve with the empirical one based on Kaplan-Meier estimator. It is shown that the smoothed estimator outperforms the direct empirical estimator based on the Kaplan-Meier estimator under the criterion of deficiency. A simulation study is also conducted and a real data is analyzed.

Support vector machine and ROC curves for modeling of aircraft fuel consumption

To estimate the fuel consumption of a civil aircraft,we propose to use the receiver operating characteristic(ROC)curve to optimize a support vector machine(SVM)model.The new method and procedure has been developed to build,train,validate,and apply an SVM model.A conceptual support vector network is proposed to model fuel consumption,and the flight data collected from routes are used as the inputs to train an SVM model.During the training phase,an ROC curve is defined to evaluate the performance of the model.To validate the applicability of the trained model,a case study is developed to compare the data from an aircraft performance manual and from the implemented simulation model.The investigated aircraft in the case study is a Boeing 737-800 powered by CFM-56 engines.The comparison has shown that the trained SVM model from the proposed procedure is capable of representing a complex fuel consumption function accurately for all phases during the flight.The proposed methodology is generic,and can be extended to reliably model the fuel consumption of other types of aircraft,such as piston engine aircraft or turboprop engine aircraft.

Enhancing XRF sensor-based sorting of porphyritic copper ore using particle swarm optimization-support vector machine(PSO-SVM)algorithm

X-ray fluorescence(XRF)sensor-based ore sorting enables efficient beneficiation of heterogeneous ores,while intraparticle heterogeneity can cause significant grade detection errors,leading to misclassifications and hindering widespread technology adoption.Accurate classification models are crucial to determine if actual grade exceeds the sorting threshold using localized XRF signals.Previous studies mainly used linear regression(LR)algorithms including simple linear regression(SLR),multivariable linear regression(MLR),and multivariable linear regression with interaction(MLRI)but often fell short attaining satisfactory results.This study employed the particle swarm optimization support vector machine(PSO-SVM)algorithm for sorting porphyritic copper ore pebble.Lab-scale results showed PSO-SVM out-performed LR and raw data(RD)models and the significant interaction effects among input features was observed.Despite poor input data quality,PSO-SVM demonstrated exceptional capabilities.Lab-scale sorting achieved 93.0%accuracy,0.24%grade increase,84.94%recovery rate,57.02%discard rate,and a remarkable 39.62 yuan/t net smelter return(NSR)increase compared to no sorting.These improvements were achieved by the PSO-SVM model with optimized input combinations and highest data quality(T=10,T is XRF testing times).The unsuitability of LR methods for XRF sensor-based sorting of investigated sample is illustrated.Input element selection and mineral association analysis elucidate element importance and influence mechanisms.

Activity optimization method in SPECT:A comparison with ROC analysis

A discriminant method for optimizing activity in nuclear medicine studies is validated by comparison with ROC （received operating characteristic）-curves. The method is tested in 21 single photon emission computerized tomography （SPECT）, performed with a cardiac phantom. Three different lesions （Lb L2 and L3） were placed in the myocardium-wall by pairs fbr each SPECT. Three activities （84, 37 or 18.5 MBq） of 99mTc were used as background. Linear discriminant analysis was used to select the parameters that characterize image quality among the measured variables in the images [（Background-to-Lesion （B/Li） and Signal-to-Noise （S/N） ratios）]. Two clusters with different image quality （P=0.021 ） were obtained. The ratios B/Lj, B/L2 and B/L3 are the parameters used to construct the function with 100% of cases correctly classified into the clusters. The value of 37 MBq was the lowest tested activity for which good results for the B/Li ratios were obtained. The result coincides with the applied ROC-analysis （r=0.89）.

ROC Analysis in Radiotherapy:A TCP Model-Based Test

Several mathematical models have been proposed to describe the dynamics of irradiated cancer cells and to evaluate the tumour control probability (TCP). In this article, we propose a TCP model-based statistical test for predicting the outcome of a radiation treatment. We determine the foresight capability of prostate tumour erradication (cure) from Monte Carlo simulations of the Dawson-Hillen TCP model. We construct the receiver operating characteristic (ROC) curves of the test from the probability distributions of the fraction of remaining tumour cells for simulated experiments that evolve either to cure or non-cure. Simulations show that a similar procedure may be applicable to clinical data. Results suggest that the evaluation of tumour sizes after the treatment has started may be used for short-term prognosis.

Prediction of Abdominal Visceral Obesity From Body Mass Index,Waist Circumference and Waist-hip Ratio in Chinese Adults:Receiver Operating Characteristic Curves Analysis

Objective To evaluate the sensitivity and specificity of body mass index (BMI), waist circumference (WC) and waist-to-hip ratio (WHR) measurements in diagnosing abdominal visceral obesity. Methods BMI, WC, and WHR were assessed in 690 Chinese adults (305 men and 385 women) and compared with magnetic resonance imaging (MRI) measurements of abdominal visceral adipose tissue (VA). Receiver operating characteristic (ROC) curves were generated and used to determine the threshold point for each anthropometric parameter. Results 1) MRI showed that 61.7% of overweight/obese individuals (BMI≥25 kg/m2) and 14.2% of normal weight (BMI<25 kg/m2) individuals had abdominal visceral obesity (VA≥100 cm2). 2) VA was positively correlated with each anthropometric variable, of which WC showed the highest correlation (r=0.73-0.77, P<0.001). 3) The best cut-off points for assessing abdominal visceral obesity were as followed: BMI of 26 kg/m2, WC of 90 cm, and WHR of 0.93, with WC being the most sensitive and specific factor. 4) Among subjects with BMI≥28 kg/m2 or WC≥95 cm, 95% of men and 90% of women appeared to have abdominal visceral obesity. Conclusion Measurements of BMI, WC, and WHR can be used in the prediction of abdominal visceral obesity, of which WC was the one with better accuracy.

A method of line spectrum extraction based on target radiated spectrum feature and its post-processing

To improve the ability of detecting underwater targets under strong wideband interference environment,an efficient method of line spectrum extraction is proposed,which fully utilizes the feature of the target spectrum that the high intense and stable line spectrum is superimposed on the wide continuous spectrum.This method modifies the traditional beam forming algorithm by calculating and fusing the beam forming results at multi-frequency band and multi-azimuth interval,showing an excellent way to extract the line spectrum when the interference and the target are not in the same azimuth interval simultaneously.Statistical efficiency of the estimated azimuth variance and corresponding power of the line spectrum band depends on the line spectra ratio(LSR)of the line spectrum.The change laws of the output signal to noise ratio(SNR)with the LSR,the input SNR,the integration time and the filtering bandwidth of different algorithms bring the selection principle of the critical LSR.As the basis,the detection gain of wideband energy integration and the narrowband line spectrum algorithm are theoretically analyzed.The simulation detection gain demonstrates a good match with the theoretical model.The application conditions of all methods are verified by the receiver operating characteristic(ROC)curve and experimental data from Qiandao Lake.In fact,combining the two methods for target detection reduces the missed detection rate.The proposed post-processing method in2-dimension with the Kalman filter in the time dimension and the background equalization algorithm in the azimuth dimension makes use of the strong correlation between adjacent frames,could further remove background fluctuation and improve the display effect.

ROC analysis of CT hemodynamic in the diagnosis of breast cancer

Objective:The aim of this study was evaluate the diagnostic value of computed tomography(CT) perfusion in breast cancer by the method of receiver operator characteristic curve(ROC) analysis.Methods:Eighty-one cases with breast masses found by health examination or mammography were scanned by multi-slice spiral CT(MSCT) perfusion and hemodynamic parameters of blood flow(BF), mean transit time(MTT) and blood volume(BV) were calculated by deconvolution arithmetic.According to the pathologic results, two groups, benign and malignant were classified and statistical analysis were performed between them.The ROC characteristics of BF, MTT, BV were compared for each and the diagnostic value of the hemodynamic parameters were confirmed.Results:In the malignant group, BF was(0.735 ± 0.440) mL/min/mL, MTT was(22.771 ± 7.647) s and BV was 0.234 ± 0.082.In the benign group, BF was(0.466 ± 0.527) mL/min/mL, MTT was(26.712 ± 12.934) s and BV was 0.179 ± 0.117.There was a significant difference for BF and BV between the benign and malignant groups.When the hemodynamic parameters were used to discriminate the breast lesions, the area under the ROC curve(AUCROC) of BF was 0.832 ± 0.086, the maximum, while AUCROC of BV was 0.695 ± 0.092.There was no significant statistical difference between BF and BV.AUCROC of MTT was 0.473 ± 0.102, which was minimal.Since the threshold of BF was 0.381 mL/min/mL, the sensitivity was 82.3%, the specificity was 73.2%, the positive likelihood ratio(LR) was 3.071 and the negative LR was 0.242.The threshold of BV was 0.190 with sensitivity 73.3%, specificity 56.5%, positive likelihood ratio 1.685 and negative LR 0.473.Conclusion:BF and BV among CT hemodynamic parameters have certain diagnostic value in breast cancer, but BF or BV can not yet be single index to confirm or deny the diagnosis.

Evaluation of the diagnostic value of tumor markers with ROC and logistic regression in gastrocolic tumors

Objective: The aim of the study was to evaluate the clinical significance of multiple tumor markers (CEA, AFP, CA72-4 and CA19-9) in patients with gastrocolic tumors by receiver operating characteristic (ROC) curve and stepwise logistic regression (LR) analysis. Methods: The serum concentrations of CEA, AFP, CA72-4 and CA19-9 were measured with electrochemiluminescence immunoassay in 126 patients with gastrocolic tumors, 137 patients with benign gastrocolic disorders and 109 healthy controls. The area under the ROC curve (AUC) of CEA, AFP, CA72-4 and CA19-9 and stepwise LR results were compared by sensitivity, specificity, Youden's index and positive likelihood ratio/negative likelihood ratio. Results: The levels of four tested tumor markers in patients with gastrocolic tumors were significantly higher than those in benign gastrocolic group and normal controls. In the benign gastrocolic group, the AUC from stepwise logistic regression was larger than the AUC of four tumor markers respectively. Sensitivity, Youden's index and positive likelihood ratio/negative likelihood ratio were the highest in the combination assay of CA72-4, CEA, and CA19-9, as compared with one of the tumor markers alone. Conclusion: The use of ROC established by LR analysis model improved the diagnostic accuracy of gastrocolic tumors. For the screening of gastrocolic tumors, the AUC value of the combination probability index (sensitivity and specificity) was significantly higher than the values of the different tumour markers.

BOOTSTRAP TECHNIQUE FOR ROC ANALYSIS: A STABLE EVALUATION OF FISHER CLASSIFIER PERFORMANCE

This paper presents a novel bootstrap based method for Receiver Operating Characteristic (ROC) analysis of Fisher classifier. By defining Fisher classifier’s output as a statistic, the bootstrap technique is used to obtain the sampling distributions of the outputs for the positive class and the negative class respectively. As a result, the ROC curve is a plot of all the (False Positive Rate (FPR), True Positive Rate (TPR)) pairs by varying the decision threshold over the whole range of the boot- strap sampling distributions. The advantage of this method is, the bootstrap based ROC curves are much stable than those of the holdout or cross-validation, indicating a more stable ROC analysis of Fisher classifier. Experiments on five data sets publicly available demonstrate the effectiveness of the proposed method.

Multi-Features Disease Analysis Based Smart Diagnosis for COVID-19

Coronavirus 2019(COVID-19)is the current global buzzword,putting the world at risk.The pandemic’s exponential expansion of infected COVID-19 patients has challenged the medical field’s resources,which are already few.Even established nations would not be in a perfect position to manage this epidemic correctly,leaving emerging countries and countries that have not yet begun to grow to address the problem.These problems can be solved by using machine learning models in a realistic way,such as by using computer-aided images during medical examinations.These models help predict the effects of the disease outbreak and help detect the effects in the coming days.In this paper,Multi-Features Decease Analysis(MFDA)is used with different ensemble classifiers to diagnose the disease’s impact with the help of Computed Tomography(CT)scan images.There are various features associated with chest CT images,which help know the possibility of an individual being affected and how COVID-19 will affect the persons suffering from pneumonia.The current study attempts to increase the precision of the diagnosis model by evaluating various feature sets and choosing the best combination for better results.The model’s performance is assessed using Receiver Operating Characteristic(ROC)curve,the Root Mean Square Error(RMSE),and the Confusion Matrix.It is observed from the resultant outcome that the performance of the proposed model has exhibited better efficient.

Prediction of Flash Flood Susceptibility of Hilly Terrain Using Deep Neural Network:A Case Study of Vietnam

Flash floods are one of the most dangerous natural disasters,especially in hilly terrain,causing loss of life,property,and infrastructures and sudden disruption of traffic.These types of floods are mostly associated with landslides and erosion of roads within a short time.Most of Vietnamis hilly and mountainous;thus,the problem due to flash flood is severe and requires systematic studies to correctly identify flood susceptible areas for proper landuse planning and traffic management.In this study,three Machine Learning(ML)methods namely Deep Learning Neural Network(DL),Correlation-based FeatureWeighted Naive Bayes(CFWNB),and Adaboost(AB-CFWNB)were used for the development of flash flood susceptibility maps for hilly road section(115 km length)of National Highway(NH)-6 inHoa Binh province,Vietnam.In the proposedmodels,88 past flash flood events were used together with 14 flash floods affecting topographical and geo-environmental factors.The performance of themodels was evaluated using standard statisticalmeasures including Receiver Operating Characteristic(ROC)Curve,Area Under Curve(AUC)and Root Mean Square Error(RMSE).The results revealed that all the models performed well(AUC>0.80)in predicting flash flood susceptibility zones,but the performance of the DL model is the best(AUC:0.972,RMSE:0.352).Therefore,the DL model can be applied to develop an accurate flash flood susceptibility map of hilly terrain which can be used for proper planning and designing of the highways and other infrastructure facilities besides landuse management of the area.

Application of Support Vector Machine to Predict 5-year Survival Status of Patients with Nasopharyngeal Carcinoma after Treatment

Objective： Support Vector Machine （SVM） is a machine-learning method, based on the principle of structural risk minimization, which performs well when applied to data outside the training set. In this paper, SVM was applied to predict 5-year survival status of patients with nasopharyngeal carcinoma （NPC） after treatment, we expect to find a new way for prognosis studies in cancer so as to assist right clinical decision for individual patient. Methods： Two modelling methods were used in the study; SVM network and a standard parametric logistic regression were used to model 5-year survival status. And the two methods were compared on a prospective set of patients not used in model construction via receiver operating characteristic （ROC） curve analysis. Results： The SVM1, trained with the 25 original input variables without screening, yielded a ROC area of 0.868, at sensitivity to mortality of 79.2% and the specificity of 94.5%. Similarly, the SVM2, trained with 9 input variables which were obtained by optimal input variable selection from the 25 original variables by logistic regression screening, yielded a ROC area of 0.874, at a sensitivity to mortality of 79.2% and the specificity of 95.6%, while the logistic regression yielded a ROC area of 0.751 at a sensitivity to mortality of 66.7% and gave a specificity of 83.5%. Conclusion： SVM found a strong pattern in the database predictive of 5-year survival status. The logistic regression produces somewhat similar, but better, results. These results show that the SVM models have the potential to predict individual patient＇s 5-year survival status after treatment, and to assist the clinicians for making a good clinical decision.

Support vector machine for prediction of siRNA silencing efficacy

In order to assist the design of short interfering ribonucleic acids （siRNA）, 573 non-redundant siRNAs were collected from published literatures and the relationship between siRNAs sequences and RNA interference （RNAi） effect is analyzed by a support vector machine （SVM） based algorithm relied on a basebase correlation （BBC） feature. The results show that the proposed algorithm has the highest area under curve （AUC） value （0. 73） of the receive operating characteristic （ROC） curve and the greatest r value （0. 43） of the Pearson＇s correlation coefficient. This indicates that the proposed algorithm is better than the published algorithms on the collected datasets and that more attention should be paid to the base-base correlation information in future siRNA design.

Combined detection of serum tumor markers for differential diagnosis of solid lesions located at the pancreatic head

BACKGROUND: The differential diagnosis of solid lesions located at the pancreatic head is very important for choosing therapies and setting up surgical tactics. This study was designed to evaluate the clinical significance of combined measurement of multiple serum tumor markers and the application of the receiver-operating characteristic (ROC) curves in the differential diagnosis of solid lesions located at the pancreatic head. METHODS: The serum levels of CA19-9, CA242, CA50 and carcinoembryonic antigen (CEA) in 112 patients with carcinoma of the pancreatic head and 38 patients with focal chronic pancreatitis in the pancreatic head were measured with ELISA. The sensitivity, specificity, positive likelihood ratio (PLR) and negative likelihood ratio (NLR) of the four serum tumor markers were calculated. The ROC curves for the four serum tumor markers were constructed and the area under the curve (AUC) was calculated. RESULTS: The AUCs of CA19-9, CA242, CA50 and CEA were 0.805, 0.749, 0.738 and 0.705; the PLRs were 1.91, 3.43, 5.09 and 5.46; and the NLRs were 0.41, 0.56, 0.59 and 0.71, respectively. Combined measurements increased the diagnostic specificity, and parallel combined testing increased the diagnostic sensitivity. CONCLUSIONS: Combined measurement of serum tumor markers CA19-9, CA242, CA50 and CEA is valuable in differential diagnosis of solid lesions located at the pancreatic head, and CA19-9 has the best diagnostic ability. Combined measurements can increase the specificity of diagnosis. Evaluation with the ROC curve is better than the sensitivity or specificity alone and the results are more integrated and objective.

Diagnostic value of maximal-outer-diameter and maximal-mural-thickness in use of ultrasound for acute appendicitis in children

MM： To evaluate the maximal-outer-diameter （MOD） and the maximal-mural-thickness （MMT） of the appendix in children with acute appendicitis and to determine their optimal cut-off values to diagnose acute appendicitis.METHODS： In total, 164 appendixes from 160 children between 1 and 17 years old （84 males, 76 females; mean age, 7.38 years） were examined by high-resolution abdominal ultrasound for acute abdominal pain and the suspicion of acute appendicitis. We measured the MOD and the MMT at the thickest point of the appendix. Patients were categorized into two groups according to their medical records： patients who had surgery （surgical appendix group） and patients who did not have surgery （non-surgical appendix group）. Data were analyzed by MedCalc v.9.3. The rank sum test （Mann-Whitney test） was used to evaluate the difference in the MOD and the MMT between the two groups. ROC curve analysis was used to determine the optimal cut-off value of the MOD and the MMT on diagnosis of acute appendicitis.RESULTS： There were 121 appendixes （73.8%） in the non-surgical appendix group and 43 appendixes （26.2%） in the surgical appendix group. The median MOD differed significantly between the two groups （0.37 cm vs 0.76 cm, P〈 0.0001）, and the median MMT also differed （0.15 cm vs 0.33 cm, P 〈 0.0001）. The optimal cut-off value of the MOD and the MMT for diagnosis of acute appendicitis in children was 〉 0.57 cm （sensitivity 95.4%, specificity 93.4%） and 〉 0.22 cm （sensitivity 90.7%, specificity 79.3%）, respectively.CONCLUSION： The MOD and the MMT are reliable criteria to diagnose acute appendicitis in children. An MOD 〉 0.57 cm and an MMT 〉 0.22 cm are the optimal criteria.

Study on the value of serum miR-106b for the early diagnosis of hepatocellular carcinoma

AIM To analyze the incidence of hepatocellular carcinoma (HCC) in a population that underwent health checkups and had high serum miR-106b levels. METHODS A total of 335 subjects who underwent checkups in the Digestive and Liver Disease Department of our hospital were randomly selected. RT-PCR was used to detect the level of miR-106b in serum samples. Laboratory and imaging examinations were carried out to confirm the HCC diagnosis in patients who had a > 2-fold change in miR-106b levels. Ultrasound-guided biopsy was also used for HCC diagnosis when necessary. On this basis, the clinical data of these subjects, including history of hepatitis virus infection, obesity, long-term history of alcohol use and stage of HCC, were collected. Then, the impact of these factors on the level of miR1-06b in serum was analyzed. Furthermore, receiver operating characteristic (ROC) curve was drawn to evaluate the diagnostic efficacy of miR-106b for HCC. RESULTS A total of 35 subjects had abnormal serum miR-106b levels, of which 20 subjects were diagnosed with HCC. t-test revealed that the difference in serum miR-106b level in terms of sex, age, history of hepatitis virus infection, obesity and long-term history of alcohol use was not statistically significant. However, serum miR-106b levels in patients with advanced HCC (stage. /.) was higher than in patients with early HCC (stage./.), and the difference was statistically significant (P = 0.000). Moreover, the ROC curve revealed that the area under the curve value for miR-106b was 0.885, which shows that serum miR-106b level has a certain clinical value for HCC diagnosis. CONCLUSION The random sampling survey shows that serum miR-106b level is a valuable diagnostic marker for HCC. However, the diagnostic threshold value needs to be further researched.

Percent free prostate-specific antigen for prostate cancer diagnosis in Chinese men with a PSA of 4.0-10.0 ng/mL:Results from the Chinese Prostate Cancer Consortium

Objective:To test the diagnostic performance of percent free prostate-specific antigen(%fPSA)in predicting any prostate cancer(PCa)and high-grade prostate cancer(HGPCa)in a retrospective multi-center biopsy cohort with a PSA level of 4.0e10.0 ng/mL in China.Methods:Consecutive patients with a PSA of 4.0-10.0 ng/mL who underwent transrectal ultrasound-guided biopsy were enrolled at 16 Chinese medical centers from January 1st,2010 to December 31st,2013.Total and free serum PSA determinations were performed using three types of electro-chemiluminescence immunoassays recalibrated to the World Health Organization(WHO)standard.The diagnostic accuracy of PSA,%fPSA,and %fPSA in combination with PSA(%fPSA t PSA)was determined using the area under the receiver operating characteristic(ROC)curve(AUC).Results:A total of 2310 consecutive men with PSA levels between 4.0 and 10.0 ng/mL were included,and the detection rate of PCa was 25.1%.The AUC of%fPSA and %fPSA t PSA in predicting any PCa was superior to PSA alone in men aged≥60 years(0.623 vs.0.534,p<0.0001)but not in men aged 40e59 years(0.517 vs.0.518,p=0.939).Similar result was yield in predicting HGPCa.Conclusion:In a clinical setting of Chinese men with 4.0e10.0 ng/mL PSA undergoing initial prostate biopsy,adding %fPSA to PSA can moderately improve the diagnostic accuracy for any PCa and HGPCa compared with PSA alone in patients≥60 but not in patients aged 40-59 years.

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.