Error Curve Research Articles

Is it sufficient? Assessment of two sampling methods for urban plant species richness investigations

Determining adequate sample size (ASS) and minimum plot size (MPS) are two fundamental issues in urban vegetation investigations. To estimate the ASS and MPS of urban plants, we introduced sampling completeness to the MPS calculation based on 54 samples in the urban area of Chongqing, China. Then, we examined the performance of the species-area curve method and CVJack1 (coefficient of variation (CV) for the first-order jackknife estimator (Jack1)) curve extrapolation method. We also tested the effectiveness and error distribution of extrapolating the ASS from CVJack1 curves using 180 samples from the urban area of Chongqing, China, and 222 samples from the urban area of Xiamen, China. The results of the urban plant analyses showed that (1) the constructed species-area accumulation curves fit a logistic function (R2 > 0.990); (2) the MPS increased with sampling completeness, whereas unrecorded species exhibited the opposite trend; (3) the CVJack1 curves fit the allometric1 function (R2 > 0.960) but might fail to reach a CV value of 0.05 for the calculation of ASS if the sample size was too small; and (4) the relative error values were below 5% when the error curves of extrapolation (by CVJack1 curve) approached a horizontal asymptote. The species-area accumulation curve with sampling completeness and the CVJack1 curve extrapolation method were promising approaches for determining the MPS and ASS of urban plants. The results and methods in this study provide a reference for determining the appropriate plot size and sample size for urban plant species richness investigations.

Tooth profile design of a novel helical gear mechanism with improved geometry for a parallel shaft transmission

Abstract. Based on the space curve meshing equation, a novel helical gear mechanism with improved geometry is presented. First, equations of the theoretical contact curves were derived based on the space curve meshing theory. Then, tooth surfaces with a concave–convex meshing form were constructed, depending on the contact curves. The tooth profiles were improved as the theoretical contact curves were corrected by predestining the designed transmission errors. The effect of the center distance error on the transmission errors was studied, and the effects of gear modifications on transmission errors and maximum contact stresses were also investigated. The results show that the transmission error curves of the improved gear drive become much smoother. Maximum contact stresses of the improved gear drive are decreased synchronously.

Ground Subsidence Monitoring in a Mining Area Based on Mountainous Time Function and EnKF Methods Using GPS Data

Ground subsidence is an important geomorphological phenomenon in mining areas. It is difficult to monitor and predict ground subsidence with high precision, especially in mountainous mining areas. Taking the mining workface of a mountainous coalfield in Taiyuan City, in the Shanxi Province of China as an example, this research selects five typical points from GPS observation data along the strike section. Based on the materials, the ground subsidence processes at these typical points are monitored and predicted using the mountainous time function method. Acquired from the mountains time function is a recurrence equation, which is regarded as the state equation, and the Ensemble Kalman (EnKF) method is conducted accordingly. Finally, the performance of the two methods is evaluated and compared using error curves and indexes. This research presents a recurrence equation based on the mountainous time function method and establishes the EnKF method for ground subsidence monitoring and prediction. Meanwhile, compared to the mountainous time function method, the values of the ME, MAE, RMSE and MAPE indexes are largely improved for the EnKF method. Hence, this research not only presents an effective method for ground subsidence monitoring in mountainous mining areas, but also provides theoretical support for safe coal mining and environmental protection.

МЕТОД ОПТИМИЗАЦИИ КЛАССИФИКАТОРА ЖЕСТОВ

The work is devoted to the study of the possibility of optimizing the process of synthesis ofgesture classifiers by selecting the most significant channels of electromyographic (EMG) activityof the muscles of the forearm. The first part of the study is devoted to the development and analysisof the performance of gesture classifiers with a different number of EMG channels, ranked bysignificance based on the Pearson criterion. The solution of the problem of classification of gesturesby EMG signals was first implemented on the basis of ensembles of decision trees trained bythe gradient boosting method. For this, software was developed that allows automatic synthesisand training of gesture classifiers. Next, a series of studies was carried out to find the optimalnumber of EMG channels based on three criteria: the classifier learning rate, the performance ofthe trained model, and the area under the ROC AUC error curve. To do this, a cycle of trainingand testing of the classifier was carried out for data sets recorded at different positions of the electrodeson the forearm. Then, range diagrams of the studied criteria were constructed for variousnumbers of EMG channels involved in the work from 1 to 8, ranked by significance in each of thesamples. It was found that the optimal number of EMG channels involved under the experimentalconditions was 3-6, since a further increase did not lead to a decrease in the classification error,while significantly degrading the performance. The proposed method allows you to automaticallyselect the channels, the electrodes of which are located above the most informative areas of theforearm in case of an accidental change in the position of the sensors. The second part of the workcontains the results of a full-scale experiment to demonstrate the possibility of controlling awheeled robot through EMG analysis.

Industrial Robot Contouring Control Based on Non-Uniform Rational B-Spline Curve

This paper presents a novel algorithm about the industrial robot contouring control based on the NURBS (non-uniform rational B-spline) curve. First, aiming at the error between the industrial robot’s actual trajectory and the desired trajectory, the contour error is proposed as the trajectory evaluation index, and the estimation algorithm of contour error based on the tangent approximation is proposed. Based on the tangent approximation algorithm, the estimation algorithm of contour error in the local task coordinate frame is proposed to realize the transformation from the Cartesian coordinate frame to the local task coordinate frame. Second, according to the configuration of the industrial robot, a modified cross-coupling control scheme based on the local task coordinate frame is designed. Finally, the Bernoulli’s lemniscate curves are constructed by NURBS curve and five-order polynomial curve, respectively, and they are symmetrical. The contrast experiment is designed using the two types of constructed Bernoulli’s lemniscate curves as the incentive trajectory. Through the analysis and comparison between the obtained uniaxial tracking error and the contour error curve of the two incentive trajectories, it is concluded that the incentive trajectory constructed by the NURBS curve has better contour control performance than that constructed by the five-order polynomial curve. The results drawn from this paper lay a certain foundation for the future high-precision contouring control of industrial robots.

Neuro-computing-based Levenberg Marquardt algorithm for entropy optimized Darcy-Forchheimer nanofluid with variable viscosity

AI-based machine learning and intelligent computing solvers are famous among scientists and researchers for their resilience, robustness, stability, and fast convergence. Artificial neural network (ANN) is one of the important and burgeoning fields in artificial intelligence. The work in hand exploits the strength of artificial back propagated neural networks with the Levenberg Marquardt algorithm (ABNN-LMA) to examine the entropy generation of the magnetohydrodynamic Darcy-Forchheimer nanofluid flow model (MHD-DFNM) over a stretched surface. Two types of nanoparticles, i.e Silicon dioxide (SiO2) and Molybdenum disulfide (MoS2), and continuous phase liquid, i.e propylene glycol, are considered. The PDEs of MHD-DFNM are transformed into coupled ODEs by appropriate (similarity) transformations. The reference dataset is obtained by the variation of influential parameters of MHD-DFNM from the Homotopy Analysis Method (HAM). The reference HAM data are executed in training/testing/validation sets to find and analyze the approximated solution of the designed ABNN-LMA and its comparison with the reference data solution. The better performance is consistently certified with mean squared error (MSE) curves, regression index, and error histogram study. The results reveal that an increase in values of porosity parameter declines the velocity profile for both SiO2 and MoS2 nanoparticles suspended in propylene glycol nanofluids. The thermal profile improves for both SiO2 − propylene glycol and MoS2 − propylene glycol nanofluids when the heat generation parameter increases.

Reconstruction error model of distributed shape sensing based on the reentered frame in OFDR.

At present, the reconstruction error of optical fiber shape sensing is commonly represented by Euclidean distance error. However, the Euclidian error of shape reconstruction will be dependent on the shape complexity, which depends on length, curvature and torsion. In this paper, we establish a reconstruction error model of distributed shape sensing in optical frequency domain reflectometry (OFDR) based on the Frenet-Serret frame and the error delivering theory, which illustrates the relationship between the reconstruction error and parameters such as curvature, torsion, fiber length and strain measurement error. We experimentally verify the feasibility and applicability of the proposed reconstruction error model by distributed optical fiber shape sensing system based on OFDR. The proposed reconstruction error model can provide a prediction of the maximal reconstruction error when the estimated range of curvature, torsion, fiber length of a shape needs to be reconstructed and strain measurement errors of OFDR system are known. It is very useful to judge whether the shape reconstruction error meets the requirement according to the shape to be reconstructed.

Challenges in Petroleum Characterization—A Review

252 literature sources and about 5000 crude oil assays were reviewed in this work. The review has shown that the petroleum characterization can be classified in three categories: crude oil assay; SARA characterization; and molecular characterization. It was found that the range of petroleum property variation is so wide that the same crude oil property cannot be measured by the use of a single standard method. To the best of our knowledge for the first time the application of the additive rule to predict crude oil asphaltene content from that of the vacuum residue multiplied by the vacuum residue TBP yield was examined. It was also discovered that a strong linear relation between the contents of C5-, and C7-asphaltenes in crude oil and derived thereof vacuum residue fraction exists. The six parameter Weibull extreme function showed to best fit the TBP data of all crude oil types, allowing construction of a correct TBP curve and detection of measurement errors. A new SARA reconstitution approach is proposed to overcome the poor SARA analysis mass balance when crude oils with lower density are analyzed. The use of a chemometric approach with combination of spectroscopic data was found very helpful in extracting information about the composition of complex petroleum matrices consisting of a large number of components.

Deep learning based short-range forecasting of Indian summer monsoon rainfall using earth observation and ground station datasets

We develop a deep learning model (DL) for Indian Summer Monsoon (ISM) short-range precipitation forecasting using a ConvLSTM network. The model is built using daily precipitation records from both ground-based observations and remote sensing. Precipitation datasets from the Tropical Rainfall Measuring Mission and the India Meteorological Department are used for training, testing, forecasting, and comparison. For lead days 1 and 2, the correlation coefficient (CC), which was determined using predicted data from the previous five years and corresponding observational records (from both in-situ and remote sensing products), yielded values of 0.67 and 0.42, respectively. Interestingly, the CCs are even higher over the Western Ghats and Monsoon trough region. The model performance evaluated based on skill scores, Normalized Root Mean Squared Error (NRMSE), Mean absolute percentage error (MAPE) and ROC curves show a reasonable skill in short-range precipitation forecasting. Incorporating multivariable-based DL has the potential to match or even better the forecasts made by the state-of-the-art numerical weather prediction models.

Hydrodynamic modelling of large arrays of modularized floating structures with independent oscillations

It is a challenging issue to fully consider the radiation interaction among floating bodies in a large array configuration. It requires large computational efforts to resolve the interaction matrix of floaters oscillating independently with 6 degrees of freedom (DoF). Obviously, when the distance between 2 floaters is large enough, their interaction will gradually vanish. It inspires the present study to investigate a cut-off radius, outside which the hydrodynamic interaction can be ignored. It should be noted that the computational efficiency and accuracy is a pair of contradictions: a large cut-off radius is always accompanied by a high accuracy, but requires more computational efforts, and vice versa. The objective of the present study is to quantify the interaction effect and to find an optimal cut-off radius which could reduce the computational time while ensuring a satisfactory accuracy in engineering practice. Based on the potential flow theory, we calculated the hydrodynamic interaction among multiple rectangular boxes and eventually quantified the interaction effects determined by the oscillating frequency and separating distance. Some critical curves of various truncation errors (Et) were obtained, showing whether the hydrodynamic interaction effects can be neglected, were depicted. The results from two case studies showed that the present cut-off scheme could provide a very reliable prediction of the hydrodynamic responses of multiple floating bodies in an array, while the computational time was significantly saved.

FOCUS ON ARTIFICIAL INTELLIGENCE FOR PREDICTING THE OUTFLOW OF CLIENTS FROM ON-LINE EDUCATION SITES

The article examines the task of assessing the cost of housing in the cities of Ukraine. The purpose of this work is to simplify the determination of the value of apartments on the real estate market using machine learning technologies. To solve this problem, it is proposed to use a program module in Python using the Sequential direct distribution model of the keras library. A program was created that estimates the value of apartments according to their parameters using a neural network. The importance of forecasting in the field of real estate is shown, because the housing market is a systemic part of the regional economy. The results of the software application, which consists of two parts, are presented. The first program collects the necessary data for training a neural network about apartments from the OLX site ads, their structuring and recording in a csv file. The second program provides tools for preliminary analysis of the collected data, after which they are cleaned, divided into training and test samples and trained on their basis by a multilayer neural network of direct propagation using a machine learning algorithm. The layers API of the keras library was used to design the neural network, which allows the user to create arbitrary layers. For regularization, the keras.regularizers tool, which is also in the layers API, is used. To configure model metrics, the compile method was used. Three hidden layers were defined, for each of which 512 neurons were introduced and the Relu activation function was chosen. Calculations of the correlation of prediction indicators and error curves of machine learning are given. As a result of testing the trained neural network on a test set of 652 examples, an average absolute error of 3570.88 was obtained, and the accuracy of the model was approximately 85%. Thus, the neural network has reached an acceptable level of accuracy for estimating the cost of apartments in the city of Kharkiv. Ways to reduce test errors and learning errors using cross-validation are proposed. Concepts of learning hyper-parameters and their regularization are considered Keywords: neural networks, deep learning, machine learning, regression, prediction, estimation, data analysis.

Evaluation of College Students’ Classroom Learning Effect Based on the Neural Network Algorithm

With the advent of the digital society, the amount of information we face will increase exponentially, which will challenge our level of educational knowledge, so we begin to pay attention to the effect of education and teaching in the context of digitalization. The purpose of this paper is to study the evaluation of students’ classroom learning effect based on the neural network algorithm and scientific objectivity. Assessment and other principles are to create an assessment system for students’ learning outcomes in the classroom. The system includes three first-level indicators, including first-level indicators and their weights. By selecting a university to test the system, the results show that the system can quantitatively evaluate the learning outcomes, and the corresponding scores and grades can be obtained through the formula. After adjusting the parameters of the hidden layer nodes, the BP elastic gradient algorithm is used to complete the evaluation model generation. The training error results show that the target curve and the output curve almost coincide, and the error curve is also between −0.2 and 0.5. Therefore, the learning outcome score obtained by the BP neural network based on the principal component factor data is basically consistent with the given learning outcome score.

Percentile curves of refractive errors in a Spanish paediatric population

ObjectiveThe main objective of this study was to obtain percentile curves of refractive errors in a Spanish paediatric population aged between 3 and 12 years. Materials and methodA descriptive, observational and cross-sectional study was conducted, including children aged between 3 and 12 years who did not present with any known ocular and/or systemic diseases. The convenience sampling method was used to select the sample from three schools and one hospital in the Community of Madrid. The refractive error was obtained using a Retinomax K-plus 3 autorefractometer (RTX; Right Mfg. Co. Ltd., Tokyo, Japan). The 5th, 10th, 25th, 50th, 75th, 90th and 95th percentiles were calculated using the IBM SPSS Statistics v.24 statistical software (IBM Corporation, Armonk, NY, United States). ResultsA total of 688 children with a mean age of 7.68 ± 2.17 years were analysed. In the 50th percentile curve, spherical equivalent values started to become myopic at 3 years (SE < –0.50 D) and the 75th percentile curve also turned myopic at 4 years. As a result, it was observed that the spherical equivalent value became more negative with time, starting from the ages indicated above. Therefore, the 90th percentile curve was negative at 11 years. ConclusionPercentile curves of refractive errors in a Spanish paediatric population have been presented for the first time in order to help eyecare professionals detect children with refractive errors at an early age.

Stochastic numerical computing for entropy optimized of Darcy-Forchheimer nanofluid flow: Levenberg Marquardt Algorithm

The work in hand examines the entropy generation of magnetohydrodynamic Darcy-Forchheimer nanofluid flow model (MHD-DFNM) over a stretched surface via Artificial back propagated neural networks with Levenberg Marquardt Algorithm (ABNN-LMA). Two types of nanoparticles i.e Silicon dioxide (SiO2) and Molybdenum disulfide (MoS2) and continuous phase liquid i.e propylene glycol is considered. The PDEs of MHD-DFNM are transformed into coupled ODEs via appropriate (similarity) transformations. The reference dataset is obtained by the variation of 1st and 2nd order slip parameters (γ1 and γ2), porosity parameter (β), viscosity parameter (λ), radiation parameter(R), volume friction coefficient(ϕ), heat generation/absorption parameter(Q), Biot number (B1) and Eckert number (Ec) from Homotopy Analysis Method (HAM). The reference data is executed in training/testing/validation sets to find and analyze the approximated solution of designed ABNN-LMA as well as it’s comparison with reference data solution. The better performance is consistently certified with Mean Squared Error (MSE) curves, regression index and error histogram study. Results reveal that, increase in values of porosity parameter declines the velocity profile for both SiO2 and MoS2 nanoparticles suspended in propylene glycol nanofluids. The thermal profile improves for both SiO2 − propylene glycol and MoS2 − propylene glycol nanofluids when heat generation parameter increases.

Closed-loop experimental optimization of tunable lenses.

Tunable lenses (TLs) are optical devices that can change their optical power in response to an electrical signal. In many applications, they are often pushed to or beyond their temporal limits. Fast periodic and/or abrupt variations of the optical power induce undesired distortions in their transient response and produce a decrease in their performance. A low-cost focimetry system, along with a custom closed-loop iterative optimization algorithm, was developed to (1) characterize a TL's response at high speed and (2) optimize their performance in realistic TL working conditions. A significant lens performance improvement was found in about 23 iterations with a decrease in the area under the error curve and an improved effective time. Applying the closed-loop optimization algorithm in a depth scanning experiment enhanced the image quality. Quantitatively, the image quality was evaluated using the structural similarity index metric that improves in individual frames, on average, from 0.345 to 0.895.

High efficiency and precision approach to milling stability prediction based on predictor-corrector linear multi-step method

Regenerative chatter is the most important factor affecting the stability of the milling process. It is core for suppressing chatter and improving production efficiency to accurately and efficiently identify the stable region of milling chatter. Therefore, according to the theory of predictor-corrector, three predictor–corrector methods (PCM) are, respectively, proposed for the milling stability region by applying the fourth-order Adams-Bashforth-Moulton formula, Simpson formula, and Hamming formula. Firstly, the regenerative chatter milling process is described as a second-order time-delay differential equation (DDE) with periodic coefficients. Thus, the forced vibration time can uniformly be discretized as a time node set. Secondly, the fourth-order Adams-Bashforth formula is used to predict the displacement at every time node, whereas the fourth-order Adams-Moulton formula can be employed to correct this predicted value. In addition, the fourth-order Simpson formula and Hamming formula can also correct the predicted value. Thus, a higher precision discrete prediction-correction expansion is constructed for the transformation of DDE into the state transition express. The Floquet theory can be depended on to present the judgment criterion of milling stability. Moreover, finally, under the same milling process parameters, comparisons of both the stability lobe curve and the local discrete error curve show that the PCM has a faster convergence rate than the 1st-SDM (first-order semi-discretization method) and 2nd-FDM (second-order full-discretization method). This shows that the PCM can obtain better computational accuracy under the same discrete number, whereas the PCM is significantly higher computational efficiency over 1st-SDM and 2nd-FDM. Meanwhile, considering the actual machining environment, helix angle effect and multiple modes effect of the tool are analyzed; experimental verification considering multiple modes with helix angle further indicates the applicability of the PCM.

Machine learning-guided covariate selection for time-to-event models developed from a small sample of real-world patients receiving bevacizumab treatment.

Therapeutic outcomes in patients with metastatic colorectal cancer (mCRC) receiving bevacizumab treatment are highly variable, and a reliable predictive factor is not available. Progression‐free survival (PFS) and overall survival (OS) were recorded from an observational, prospective study after 5 years of follow‐up, including 46 patients with mCRC receiving bevacizumab treatment. Three vascular endothelial growth factor (VEGF)‐A and two intercellular adhesion molecule‐1 genes polymorphisms, age, gender, weight, dosing scheme, and co‐treatments were collected. Given the relatively small number of events (37 [80%] for the PFS and 26 [57%] for the OS), to study the effect of these covariates on PFS and OS, a covariate analysis was performed using statistical and supervised machine learning techniques, including Cox regression, penalized Cox regression techniques (least absolute shrinkage and selection operator [LASSO], ridge regression, and elastic net), survival trees, and survival forest. The predictive performance of each method was evaluated in bootstrapped samples, using prediction error curves and the area under the curve of the receiver operating characteristic. The LASSO penalized Cox‐regression model showed the best overall performance. Nonlinear mixed effects (NLME) models were developed, and a conventional stepwise covariate search was performed. Then, covariates identified as important by the LASSO model were included in the base NLME models developed for PFS and OS, resulting in improved models as compared to those obtained with the stepwise covariate search. It was shown that having gene polymorphisms in VEGFA (rs699947 and rs1570360) and ICAM1 (rs1799969) are associated with a favorable clinical outcome in patients with mCRC receiving bevacizumab treatment.

Distributed active disturbance rejection formation containment control for multiple autonomous underwater vehicles with prescribed performance

In this paper, a distributed formation containment control (FCC) issue is studied for multiple autonomous underwater vehicles (MAUVs) system with prescribed performance based on the active disturbance rejection control (ADRC) method in three-dimensional space. The unknown nonlinearities and external disturbances of the MAUVs system are approximated in real time by means of extended state observers (ESOs). Furthermore, tracking differentiators (TDs) are introduced for the sake of refraining from the explosion of complexity induced by the derivatives of the virtual control variables of autonomous underwater vehicles. Applying funnel variables, our designed control strategy ensures the formation tracking errors and containment errors to keep within the specified ranges, hence enhancing the control performance of the MAUVs system. Then, a FCC-ADRC scheme for MAUVs system is proposed. Compared with the traditional FCC scheme, the proposed FCC-ADRC scheme not only does not depend on the specific system configuration, but also can deal with any bounded disturbances, thus obtaining a more general, effective and robust disturbance rejection FCC strategies. Additionally, the stability of MAUVs system is verified through Lyapunov stability analysis. At last, the error curves and time-varying formation containment trajectories obtained by simulation confirm the feasibility and effectiveness of our method.

Interval assessment using task- and procedure-based simulations: an attempt to supplement neurosurgical residency curriculum.

The longer learning curve and smaller margin of error make nontraditional, or "out of operating room" simulation training, essential in neurosurgery. In this study, the authors propose an evaluation system for residents combining both task-based and procedure-based exercises and also present the perception of residents regarding its utility. Residents were evaluated using a combination of task-based and virtual reality (VR)-based exercises. The results were analyzed in terms of the seniority of the residents as well as their laboratory credits. Questionnaire-based feedback was sought from the residents regarding the utility of this evaluation system incorporating the VR-based exercises. A total of 35 residents were included in this study and were divided into 3 groups according to seniority. There were 11 residents in groups 1 and 3 and 13 residents in group 2. On the overall assessment of microsuturing skills including both 4-0 and 10-0 microsuturing, the suturing skills of groups 2 and 3 were observed to be better than those of group 1 (p = 0.0014). Additionally, it was found that microsuturing scores improved significantly with the increasing laboratory credits (R2 = 0.72, p < 0.001), and this was found to be the most significant for group 1 residents (R2 = 0.85, p < 0.001). Group 3 residents performed significantly better than the other two groups in both straight (p = 0.02) and diagonal (p = 0.042) ring transfer tasks, but there was no significant difference between group 1 and group 2 residents (p = 0.35). Endoscopic evaluation points were also found to be positively correlated with previous laboratory training (p = 0.002); however, for the individual seniority groups, the correlation failed to reach statistical significance. The 3 seniority groups performed similarly in the cranial and spinal VR modules. Group 3 residents showed significant disagreement with the utility of the VR platform for improving surgical dexterity (p = 0.027) and improving the understanding of surgical procedures (p = 0.034). Similarly, there was greater disagreement for VR-based evaluation to identify target areas of improvement among the senior residents (groups 2 and 3), but it did not reach statistical significance (p = 0.194). The combination of task- and procedure-based assessment of trainees using physical and VR simulation models can supplement the existing neurosurgery curriculum. The currently available VR-based simulations are useful in the early years of training, but they need significant improvement to offer beneficial learning opportunities to senior trainees.

Integrated Forecasts Based on Public Health Surveillance and Meteorological Data Predict West Nile Virus in a High-Risk Region of North America

Background:West Nile virus (WNV), a global arbovirus, is the most prevalent mosquito-transmitted infection in the United States. Forecasts of WNV risk during the upcoming transmission season could provide the basis for targeted mosquito control and disease prevention efforts. We developed the Arbovirus Mapping and Prediction (ArboMAP) WNV forecasting system and used it in South Dakota from 2016 to 2019. This study reports a post hoc forecast validation and model comparison.Objectives:Our objective was to validate historical predictions of WNV cases with independent data that were not used for model calibration. We tested the hypothesis that predictive models based on mosquito surveillance data combined with meteorological variables were more accurate than models based on mosquito or meteorological data alone.Methods:The ArboMAP system incorporated models that predicted the weekly probability of observing one or more human WNV cases in each county. We compared alternative models with different predictors including a) a baseline model based only on historical WNV cases, b) mosquito models based on seasonal patterns of infection rates, c) environmental models based on lagged meteorological variables, including temperature and vapor pressure deficit, d) combined models with mosquito infection rates and lagged meteorological variables, and e) ensembles of two or more combined models. During the WNV season, models were calibrated using data from previous years and weekly predictions were made using data from the current year. Forecasts were compared with observed cases to calculate the area under the receiver operating characteristic curve (AUC) and other metrics of spatial and temporal prediction error.Results:Mosquito and environmental models outperformed the baseline model that included county-level averages and seasonal trends of WNV cases. Combined models were more accurate than models based only on meteorological or mosquito infection variables. The most accurate model was a simple ensemble mean of the two best combined models. Forecast accuracy increased rapidly from early June through early July and was stable thereafter, with a maximum AUC of 0.85. The model predictions captured the seasonal pattern of WNV as well as year-to-year variation in case numbers and the geographic pattern of cases.Discussion:The predictions reached maximum accuracy early enough in the WNV season to allow public health responses before the peak of human cases in August. This early warning is necessary because other indicators of WNV risk, including early reports of human cases and mosquito abundance, are poor predictors of case numbers later in the season. https://doi.org/10.1289/EHP10287