Lowest Absolute Error Research Articles

Efficient image segmentation performance of gray-level image using normalized graph cut based neutrosophic membership function

The study of image segmentation is a crucial impression in image processing and computer vision appliances, which segments an input image into distinct non-overlapping homogenous divisions and helps to depict the image more conveniently. To improve the performance of the segmentation of gray-level images by introducing a normalized graph cut measure as a thresholding principle to separate an object from the background based on the neutrosophic membership function. We proposed an innovative neutrosophic approach combining a set of features of some efficient algorithms. The implementation of the proposed algorithm is known as the neutrosophic normalized graph cut (NNGC) method. In the gray image segmentation, the problem of wrongly segmentation and segmentation with low accuracy can identify by using this approach. This proposed algorithm is compared with the fuzzy entropy method, neutrosophic graph cut (NGC), classical graphcut, Otsu, and Kittler method. Moreover, we examine that in most cases, our algorithm gives the lowest absolute error, misclassification error, and higher values of signal to noise ratio values that improve the segmentation process of gray images. Finally, we analyze the change of different parameter values in the NNGC and the effect of the substitutes. Also, we discuss the computational complexity of neutrosophic weight matrix results with a weight matrix (classical) results.

Using deep learning neural networks to predict the knowledge economy index for developing and emerging economies

Missing values and the inconsistency of the measures of the knowledge economy remain vexing problems that hamper policy-making and future research in developing and emerging economies. This paper contributes to the new and evolving literature that seeks to advance better understanding of the importance of the knowledge economy for policy and further research in developing and emerging economies. In this paper we use a supervised machine deep learning neural network (DLNN) approach to predict the knowledge economy index of 71 developing and emerging economies during the 1995–2017 period. Applied in combination with a data imputation procedure based on the K-closest neighbor algorithm, DLNN is capable of handling missing data problems better than alternative methods. A 10-fold validation of the DLNN yielded low quadratic and absolute error (0,382 +- 0,065). The results are robust and efficient, and the model’s predictive power is high. There is a difference in the predictive power when we disaggregate countries in all emerging economies versus emerging Central European countries. We explain this result and leave the rest to future endeavors. Overall, this research has filled in gaps due to missing data thereby allowing for effective policy strategies. At the aggregate level development agencies, including the World Bank that originated the KEI, would benefit from our approach until substitutes come along.

Particle swarm optimization technique-based prediction of peak ground acceleration of Iraq’s tectonic regions

Peak ground acceleration (PGA) is one of the critical factors that affect the determination of earthquake intensity. PGA is generally utilized to describe ground motion in a particular zone and is able to efficiently predict the parameters of site ground motion for the design of engineering structures. Therefore, novel models are developed to forecast PGA in the case of the Iraqi database, which utilizes the particle swarm optimization (PSO) approach. A data set of 187 historical ground-motion recordings in Iraq’s tectonic regions was used to build the explicit proposed models. The proposed PGA models relate to different seismic parameters, including the magnitude of the earthquake (Mw), average shear-wave velocity (VS30), focal depth (FD), and nearest epicenter distance (REPi) to a seismic station. The derived PGA models are remarkably simple and straightforward and can be used reliably for pre-design purposes. The proposed PGA models (i.e., models I and II) obtained via the explicit formula produced using the PSO method are highly correlated to the actual PGA records owing to low coefficients of variation (CoV) of approximately 2.12% and 2.06%, and mean values (i.e., close to 1.0) of approximately 1.005 and 1.004. Lastly, high-frequency, low absolute relative error (ARE), which is below 5%, is recorded for the proposed models, thereby showing an acceptable error distribution.

Image-based size estimation of broccoli heads under varying degrees of occlusion

The growth and the harvestability of a broccoli crop is monitored by the size of the broccoli head. This size estimation is currently done by humans, and this is inconsistent and expensive. The goal of our work was to develop a software algorithm that can estimate the size of field-grown broccoli heads based on RGB-Depth (RGB-D) images. For the algorithm to be successful, the problem of occlusion must be solved, which is the partial visibility of the broccoli head due to overlapping leaves. This partial visibility causes sizing errors. In this research, we studied the use of deep-learning algorithms to deal with occlusions. We specifically applied the Occlusion Region-based Convolutional Neural Network (ORCNN) that segmented both the visible and the amodal region of the broccoli head (which is the visible and the occluded region combined). We hypothesised that ORCNN, with its amodal segmentation, can improve the size estimation of occluded broccoli heads. The ORCNN sizing method was compared with a Mask R–CNN sizing method that only used the visible broccoli region to estimate the size. The sizing performance of both methods was evaluated on a test set of 487 broccoli images with systematic levels of leaf occlusion. With a mean sizing error of 6.4 mm, ORCNN outperformed Mask R–CNN, which had a mean sizing error of 10.7 mm. Furthermore, ORCNN had a significantly lower absolute sizing error on 161 heavily occluded broccoli heads with an occlusion rate between 50% and 90%. Our software and data set are available on https://git.wur.nl/blok012/sizecnn.

Application of Levenberg-Marguardt Algorithm for Prime Radio Propagation Wave Attenuation Modelling in Typical Urban, Suburban and Rural Terrains

The desire to achieve an adaptive prognostics regression learning processes of physical and empirical phenomenon is a complex task and open problem in radio frequency telecommunication engineering. One key method to solving such complex task or problems is by means of numerical based optimisation algorithms. The Levenberg– Marquardt algorithm (LMA) is an efficient nonlinear parametric machine learning based modelling algorithm with optimal, fast, and accurate convergence speed. This paper proposes and demonstrates the real-time application of the LMA in developing a log-distance like propagation loss model based on received radio strength measurements conducted over deployed long term evolution (LTE) eNodeBs antennas in three different propagation areas. The LTE eNodeB signal propagation areas were selected to reflect typical urban, suburban and rural terrains which represent urban, suburban and rural terrains. The heights of the three eNodeBs are 30, 28 and 32m respectively and each operate at 2.6GHz carrier frequency with 10MHz channel bandwidths. The resultant outcome of the proposed propagation loss modelling using LMA indicates a high approximation efficacy over the popular Gauss-Newton algorithm (GNA) modelling method, which has been used to benchmark the process. Precisely, the developed propagation loss model using LMA method attained lower maximum absolute error (MABE) of 7.73, 14.57and 10.53 for urban, suburban and rural terrains compared to the ones developed by GNA which yielded 15.19, 16.59 and 13.05 MABE values. The improved approximation performance of the LMA over the GNA can be ascribed to its capacity handle multiple free parameters and attain optimum solution irrespective of the selected values of initial guess parameters.

Estimation of higher heating value (HHV) of bio-oils from thermochemical liquefaction by linear correlation

With increasing interest in renewable energy sources, particularly on biomass, and on the technologies to transform them into liquids fuels comes the need to assess its properties. One of the processes that can convert biomass into liquid fuel, bio-oil, is thermochemical liquefaction. However, the experimental determination of many properties can be time and budget consuming. Thus, the numerical models’ development to compute the higher heating value (HHV), with high accuracy, is needed. This work developed a model based on the linear correlation of a dataset composed of 54 samples. The carbon, hydrogen and oxygen content was correlated with HHV to develop a model. The model was compared with other from literature and finally validated with nine samples. Overall, the model exhibited excellent prediction accuracy. The absolute bias error was relatively low, 0.01%. Besides, the model’s good accuracy could be verified by the low absolute average error, 1.48%. The model’s good precision and performance was also demonstrated by the mean absolute error’s low value.

PCN198 Statistical Methods for Pantumor Analysis: Models to Account for Tumor-Level Heterogeneity

Recent tumor-agnostic regulatory approvals indicate growing interest in estimating effects of drugs and biomarkers on outcomes in specific tumor types (TTs) and across all TTs (pantumor effects), which can be challenging due to tumor-level differences in prognosis and data availability. We compared the performance of six models for estimating simulated tumor-specific and pantumor effects of a biomarker on overall survival (OS). Patient survival times were simulated based on patients’ biomarker status, tumor type, and the pantumor and tumor-specific effects of the biomarker on OS. We varied the number of TTs in each dataset (5-50), and the number of patients per TT (5-2000). Each permutation was run 100 times. To each dataset, we fit separate Cox models for each TT, a fixed-effect Cox (FE) model, a random-effects (RE) Cox model, a stratified Cox model, and the sample-size (SS) and minimum-risk (MR) alternatives to the stratified Cox (Merhotra et al., 2012). We report the absolute error (AE) comparing the estimated and true effects (log hazard). In estimating tumor-specific effects, the RE model had the lowest AE overall (median: 0.11, IQR: 0.04-0.20) compared to the FE (median: 0.19, IQR: 0.07-0.47) and separate models (median: 0.2, IQR: 0.08-0.51). In TTs with 5 patients, the RE model had the lowest AE in 86.2% of simulations. In TTs with 2000 patients, RE and separate models performed similarly. In low variability scenarios (tumor-specific effect within 0.04 of pantumor effect), the RE model had the lowest AE in 79.8% of simulations. In high variability scenarios (tumor-specific effect >0.2 from pantumor effect), all models performed similarly. In estimating pantumor effects, all models had similar AE. RE models performed favorably for estimating tumor-specific effects, and all models performed similarly for estimating pantumor effects.

Real-Time Prediction of Acoustic Velocities While Drilling Vertical Complex Lithology Using AI Technique

Mechanical rock properties are often determined using sonic log data—compressional velocity (VP) and shear velocity (VS). However, a sonic well log is not always acquired due to deteriorated hole condition (i.e., hole washout), sonic tool failures, especially in high-pressure, high-temperature (HPHT) wells, and relatively high cost. This paper introduces two data-driven models, namely artificial neural network (ANN) and random forest (RF), to estimate VP and VS across different formations that are characterized by deep burial depth and strong heterogeneity. Two types of actual field data were used to develop the models: (i) drilling surface parameters, which include flow rate, standpipe pressure, rotary speed, and surface torque, and (ii) acoustic velocities VP and VS, which were acquired by a conventional sonic log. Well-1 and Well-2 with data points of 6,846 were used to develop the models, while Well-3 with 1,016 data points was used to evaluate the capability of the developed models to generalize on an unseen data set with different statistical behavior. Furthermore, Well-3 was used to compare the accuracy of the developed models with the earliest published correlations in estimating the VS. The results showed that the RF outperformed the optimized ANN in estimating VP and VS in Well-3. The RF predicted the VP with a low average absolute percentage error (AAPE) of 0.9% and correlation of coefficient (R) of 0.87, while the AAPE and R were 6.7 % and 0.45 in the case of ANN. Similarly, the RF estimated the VS with an AAPE of 1.1% and R of 0.85, whereas the ANN predicted the VS with an AAPE of 9.5% and R of 0.40. Furthermore, the RF was the most accurate in determining VS in Well-3 compared to the earliest published correlations.

Do Analysts Distribute Negative Opinions Earlier?

This paper examines analysts’ forecast timing when issuing negative opinions. We are motivated by two findings in the literature. One finding is that management tends to withhold bad news, which would result in good news being more abundant but relatively uninformative and bad news with the opposite features. The other finding is that analysts tend to issue accurate forecasts earlier. Taken together, the two findings suggest that analysts treat observed bad news as having higher precision and respond to it by issuing forecasts more quickly. Our theoretical predictions and empirical evidence support this. We find that relative to good news forecasts 1) bad news forecasts are released earlier, and 2) holding timing fixed, bad news forecasts have lower absolute forecast errors. These results are robust to alternative sample selections and variable measurements.

Effects of Plyometric and Balance Training on Neuromuscular Control of Recreational Athletes with Functional Ankle Instability: A Randomized Controlled Laboratory Study.

Plyometric exercise has been suggested for knee injury prevention in sports participation, but studies on ankle plyometric training are limited. This study aims to investigate the change of joint position sense and neuromuscular activity of the unstable ankle after six-week integrated balance/plyometric training and six-week plyometric training. Thirty recreational athletes with functional ankle instability were allocated into three groups: plyometric group (P) vs. plyometric integrated with balance training group (BP) vs. control group (C). Ankle joint position sense, integrated electromyography (EMG), and balance adjusting time during medial single-leg drop-landing tasks were measured before and after the training period. Following the six-week period, both training groups exhibited a lower absolute error in plantar flexion (P group: pre: 3.79° ± 1.98°, post: 2.20° ± 1.31°, p = 0.016; BP group: pre: 4.10° ± 1.87°, post: 2.94° ± 1.01°, p = 0.045), and the integrated group showed a lower absolute error in inversion angles (pre 2.24° ± 1.44° and post 1.48° ± 0.93°, p = 0.022), and an increased integrated EMG of ankle plantar flexors before landing. The plyometric group exhibited a higher integrated EMG of the tibialis anterior before and after landing (pre: 102.88 ± 20.93, post: 119.29 ± 38.33, p = 0.009 in post-landing) and a shorter adjusting time of the plantar flexor following landing as compared to the pre-training condition (pre: 2.85 ± 1.15 s, post: 1.87 ± 0.97 s, p = 0.006). In conclusion, both programs improved ankle joint position sense and muscle activation of the ankle plantar flexors during single-leg drop landing. The plyometric group showed a reduced adjusting time of the ankle plantar flexor following the impact from drop landing.

Modeling and simulation of dynamic recrystallization in super austenitic stainless steel employing combined cellular automaton, artificial neural network and finite element method

A cellular automaton (CA) model for dynamic recrystallization (DRX) is established by employing Moore’s neighboring rule to predict flow stress, DRX grain size (DDRX) and DRX fraction (XDRX). The CA model has been optimized for super austenitic stainless steel at different strain rates (0.001–10 s−1) and temperatures (1173–1423 K) under isothermal deformation conditions. The output of the CA simulation has been used for establishing ANN-based constitutive models. The trained ANN-based constitutive models have been further implemented in FEM software (ABAQUS 6.14) to evaluate flow behavior and microstructure response of the alloy under various non-isothermal deformation conditions. The conventional CA (CAC) model has failed to provide a good depiction of the microstructure evolution, as it revealed a very low correlation coefficient (R) for XDRX (R ~ 0.75) and DDRX (R ~ 0.8). This inaccuracy of the model could be related to its inherent inability to consider the effect of solute drag on grain growth and DRX kinetics. Therefore, a modified cellular automata (CAM) model has been developed by introducing a new temperature-strain rate-dependent mobility parameter for numerically considering the solute drag effect. Employing non-isothermal simulations, the CAM model has revealed a higher correlation coefficient than the CAC model for predicting XDRX (R ~ 0.95) and DDRX (R ~ 0.98). Moreover, the developed CAM model has also predicted the flow behavior of the alloy in the entire domain investigated, revealing a higher correlation coefficient (R ~ 0.987) and a low average absolute relative error (8.6%).

A novel method to correct repolarization time estimation from unipolar electrograms distorted by standard filtering.

Reliable patient-specific ventricular repolarization times (RTs) can identify regions of functional block or afterdepolarizations, indicating arrhythmogenic cardiac tissue and the risk of sudden cardiac death. Unipolar electrograms (UEs) record electric potentials, and the Wyatt method has been shown to be accurate for estimating RT from a UE. High-pass filtering is an important step in processing UEs, however, it is known to distort the T-wave phase of the UE, which may compromise the accuracy of the Wyatt method. The aim of this study was to examine the effects of high-pass filtering, and improve RT estimates derived from filtered UEs. We first generated a comprehensive set of UEs, corresponding to early and late activation and repolarization, that were then high-pass filtered with settings that mimicked the CARTO filter. We trained a deep neural network (DNN) to output a probabilistic estimation of RT and a measure of confidence, using the filtered synthetic UEs and their true RTs. Unfiltered ex-vivo human UEs were also filtered and the trained DNN used to estimate RT. Even a modest 2Hz high-pass filter imposes a significant error on RT estimation using the Wyatt method. The DNN outperformed the Wyatt method in 62.75% of cases, and produced a significantly lower absolute error (p=8.99E-13), with a median of 16.91ms, on 102 ex-vivo UEs. We also applied the DNN to patient UEs from CARTO, from which an RT map was computed. In conclusion, DNNs trained on synthetic UEs improve the RT estimation from filtered UEs, which leads to more reliable repolarization maps that help to identify patient-specific repolarization abnormalities.

Application of the Strain Compensation Model and Processing Maps for Description of Hot Deformation Behavior of Metastable β Titanium Alloy.

The flow behavior of metastable β titanium alloy was investigated basing on isothermal hot compression tests performed on Gleeble 3800 thermomechanical simulator at near and above β transus temperatures. The flow stress curves were obtained for deformation temperature range of 800–1100 °C and strain rate range of 0.01–100 s−1. The strain compensated constitutive model was developed using the Arrhenius-type equation. The high correlation coefficient (R) as well as low average absolute relative error (AARE) between the experimental and the calculated data confirmed a high accuracy of the developed model. The dynamic material modeling in combination with the Prasad stability criterion made it possible to generate processing maps for the investigated processing temperature, strain and strain rate ranges. The high material flow stability under investigated deformation conditions was revealed. The microstructural analysis provided additional information regarding the flow behavior and predominant deformation mechanism. It was found that dynamic recovery (DRV) was the main mechanism operating during the deformation of the investigated β titanium alloy.

Validity and Interinstrument Reliability of a Medical Grade Physical Activity Monitor in Older Adults

Wearable physical activity monitors are associated with an increase in user’s habitual physical activity levels. Most of the older adult population do not meet the national moderate- to vigorous-intensity physical activity (MVPA) recommendations and may benefit from being prescribed a physical activity monitor. The PiezoRx is a class one medical grade device that uses step rate thresholds to measure MVPA. The validity and reliability of the PiezoRx in measuring MVPA has yet to be determined in older persons. We assessed the validity and interinstrument reliability of the PiezoRx to measure steps and MVPA in older adults. Participants (n = 19; 68.8 ± 2.3 years) wore an Omron HJ-320 pedometer, ActiGraph GT3X accelerometer, and four PiezoRx monitors during a five-stage treadmill walking protocol. The PiezoRx devices were set at moderate physical activity and vigorous physical activity step rate thresholds (steps per minute) of 100/120, 110/130, adjusted for height and adjusted for height + fitness. The PiezoRx exhibited a stronger correlation (intraclass correlation coefficient = .82) with manually counted steps than the ActiGraph (intraclass correlation coefficient = .53) and Omron (intraclass correlation coefficient = .54) and had a low absolute percentage error (3 ± 6%). The PiezoRx with moderate physical activity/vigorous physical activity step thresholds adjusted to 110/130 was strongly correlated to indirect calorimetry (0.84, p < .001) and best distinguished each walking stage as MVPA or not (sensitivity: 88%; specificity: 95%). The PiezoRx monitor is a valid and reliable measure of step count and MVPA among older adults. The device’s ability to measure MVPA in absolute terms was improved when step rate thresholds for moderate physical activity/vigorous physical activity were increased to 110/130 steps per minute in this population.

Aboveground tree biomass of Araucaria araucana in southern Chile: measurements and multi-objective optimization of biomass models

Abstract: Estimating carbon stocks in wooded systems is crucial to quantify national greenhouse gas balance estimates. However, inaccurate estimates are likely due to the divergent architecture of many species. The monkey puzzle tree Araucaria araucana, with its umbrella-like architecture is a vivid example. This species, often found in monodominant stands at high elevations, is the greatest carbon reservoir in the landscape, hence estimating its carbon storage is crucial. To provide the necessary basis for these estimations, we documented the variation in basic density and moisture content along the stem profile, identified the most suitable biomass estimation models, and quantified biomass allocation for three age ranges. We measured, felled, weighed, and separated trees into three categories: stem wood, stem bark, and foliage (branches + scaly leaves). The log-linear form of the simple allometric equation Y = aXb, based on diameter at breast height as the explanatory variable, covered a large part of the variation and showed good cross-validation performance (>0.96). Models using more covariates achieved lower absolute errors, but the estimation of the additional model parameters was associated with greater uncertainty. A multi-objective model comparison revealed that the best additional covariate to further improve biomass estimation was total tree height. The mean absolute percentage error was 9.8% for the total aboveground biomass, 8% for stem wood, 12% for stem bark and 24% for foliage. Changes in biomass distribution among tree components were related to age. For older trees, there was a relative increase in stem wood, a decreased proportion of foliage, but no change in stem bark. The proportion of stem bark biomass is similar to that of Araucaria angustifolia, but higher than in other conifers and most trees in general. Our results provide key properties for A. araucana and general guidance for the selection of easily-measurable variables allowing for excellent predictive power for local biomass estimation.

Estimation of soil water retention in conservation agriculture using published and new pedotransfer functions

Conservation agriculture has been developed as a means to improve the sustainability of agricultural systems and reduce drawbacks of conventional agricultural practices. Cropping practices can have a large influence on soil properties such as water retention. Proper tools are needed to assess and model effects of conservation agriculture on soil properties. As measuring soil water retention is expensive and time consuming, pedotransfer functions (PTFs) are now commonly used to predict them. The objectives of this study were to (i) present a new dataset of conservation agriculture data, (ii) assess performances of existing PTFs in predicting soil water retention of soils under conservation agriculture and (iii) develop new specific PTFs to predict water retention in conservation agriculture more accurately. We used data collected only in fields under conservation agriculture in France to evaluate several published PTFs with three evaluation criteria (RMSE, prediction bias (ME) and Nash-Sutcliffe Efficiency (EF)). We then developed new PTFs using three methods ― multiple linear regression, regression tree and random forest ― to predict soil water content at matric heads of -100 (θ100, field capacity for sandy soils), -330 (θ330, field capacity for other soils) and -15 000 cm (θ15 000, wilting point). Soil tillage, presence of a cover crop, rotation length and previous reduced/no tillage were used as predictors in addition to basic soil properties for regression trees and random forests. The quality of prediction (RMSE, ME and EF) was calculated for each new PTF using a cross-validation procedure. Generally, predictions of wilting point had lower absolute error than those of sandy-soil field capacity (RMSE = 0.044 and 0.066 cm3/cm3, respectively). EF was usually negative for all water contents. The cross-validation performance of the new PTFs was similar for multiple linear regression (RMSE: 0.028, ME: 0.000, EF: 0.34 for θ100) and random forest (RMSE: 0.027, ME: 0.000, EF: 0.36 for θ100), and generally worse for regression tree (especially EF). Multiple linear regression that did not consider cropping practices performed as well as random forest and thus did not identify any major influence of agricultural management on predicted water content. Future research on developing PTFs should focus on identifying more relevant predictors.

Analysis of Coupling Effect in Human-Commanded Stiffness During Bilateral Tele-Impedance

Tele-impedance augments classic teleoperation by enabling the human operator to actively command remote robot stiffness in real-time, which is an essential ability to successfully interact with the unstructured and unpredictable environment. However, the literature is missing a study on benefits and drawbacks of different types of stiffness command interfaces used in bilateral tele-impedance. In this article, we introduce a term called coupling effect, which pertains to the coupling between human-commanded stiffness going to the remote robot and force feedback coming from the remote robot. We hypothesize that, whenever the operator's commanded stiffness and force feedback are subject to coupling effect (e.g., muscle activity based stiffness command interfaces), force feedback can invoke involuntary changes in the commanded stiffness due to human reflexes. Although the coupling effect takes away some degree of the operator's control over the commanded stiffness, these involuntary changes can be either beneficial (e.g., during position tracking) or detrimental (e.g., during force tracking) to the task performance on the remote robot side. We examined the coupling effect in an experimental study with <b xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">16</b> participants, who performed position and force tracking tasks by using both coupled type (muscle activity based) and decoupled type (external device based) of interface. The results demonstrate a benefit of the coupling effect when the remote robot is operating in presence of unexpected force perturbations, where lower absolute error in position tracking task was observed. On the other hand, the decoupled type of interface is beneficial for force tracking tasks on the remote robot side, such as establishing or maintaining a stable contact with objects. However, the coupling effect negatively influences the commanding of reference stiffness to the remote robot in both position and force tracking tasks for the coupled type of interface, compared to the decoupled type of interface, which is not affected.

Use of Very High Spatial Resolution Commercial Satellite Imagery and Deep Learning to Automatically Map Ice-Wedge Polygons across Tundra Vegetation Types.

We developed a high-throughput mapping workflow, which centers on deep learning (DL) convolutional neural network (CNN) algorithms on high-performance distributed computing resources, to automatically characterize ice-wedge polygons (IWPs) from sub-meter resolution commercial satellite imagery. We applied a region-based CNN object instance segmentation algorithm, namely the Mask R-CNN, to automatically detect and classify IWPs in North Slope of Alaska. The central goal of our study was to systematically expound the DLCNN model interoperability across varying tundra types (sedge, tussock sedge, and non-tussock sedge) and image scene complexities to refine the understanding of opportunities and challenges for regional-scale mapping applications. We corroborated quantitative error statistics along with detailed visual inspections to gauge the IWP detection accuracies. We found promising model performances (detection accuracies: 89% to 96% and classification accuracies: 94% to 97%) for all candidate image scenes with varying tundra types. The mapping workflow discerned the IWPs by exhibiting low absolute mean relative error (AMRE) values (0.17–0.23). Results further suggest the importance of increasing the variability of training samples when practicing transfer-learning strategy to map IWPs across heterogeneous tundra cover types. Overall, our findings demonstrate the robust performances of IWPs mapping workflow in multiple tundra landscapes.

Asymmetric Inventory Management and the Direction of Sales Changes*

ABSTRACTWe study manufacturing firms' asymmetric inventory investment in response to sales changes. Focusing on the costs of resource adjustment and stockout that likely differ in sales‐increasing and sales‐decreasing periods, we predict and find that inventory investment declines less during periods with sales decreases than it rises during periods with sales increases. We validate this claim by showing that managers' expectations of future demand and desire to avoid inventory stockouts are important determinants of this asymmetry. In addition, we find that asymmetric inventory investment provides useful information for predicting future sales growth, and that both managers' and analysts' sales forecasts are positively associated with the asymmetry. Lastly, we document that forecasts of future sales growth that incorporate asymmetric inventory investment are associated with lower absolute forecast errors than benchmark forecasts. Overall, we highlight the importance of inventory information in understanding managers' resource adjustment and utilization decisions that have implications for forecasting future demand. Our findings on asymmetric inventory management provide new insights to fundamental analysis based on inventory signals.

Fracture Pressure Prediction Using Surface Drilling Parameters by Artificial Intelligence Techniques

AbstractSeveral correlations are available to determine the fracture pressure, a vital property of a well, which is essential in the design of the drilling operations and preventing problems. Some of these correlations are based on the rock and formation characteristics, and others are based on log data. In this study, five artificial intelligence (AI) techniques predicting fracture pressure were developed and compared with the existing empirical correlations to select the optimal model. Real-time data of surface drilling parameters from one well were obtained using real-time drilling sensors. The five employed methods of AI are functional networks (FN), artificial neural networks (ANN), support vector machine (SVM), radial basis function (RBF), and fuzzy logic (FL). More than 3990 datasets were used to build the five AI models by dividing the data into training and testing sets. A comparison between the results of the five AI techniques and the empirical fracture correlations, such as the Eaton model, Matthews and Kelly model, and Pennebaker model, was also performed. The results reveal that AI techniques outperform the three fracture pressure correlations based on their high accuracy, represented by the low average absolute percentage error (AAPE) and a high coefficient of determination (R2). Compared with empirical models, the AI techniques have the advantage of requiring less data, only surface drilling parameters, which can be conveniently obtained from any well. Additionally, a new fracture pressure correlation was developed based on ANN, which predicts the fracture pressure with high precision (R2 = 0.99 and AAPE = 0.094%).