Smaller Prediction Errors Research Articles

Cloud White: Detecting and Estimating QoS Degradation of Latency-Critical Workloads in the Public Cloud

The increasing popularity of cloud computing has forced cloud providers to build economies of scale to meet the growing demand. Nowadays, data-centers include thousands of physical machines, each hosting many virtual machines (VMs), which share the main system resources, causing interference that can significantly impact on performance. Frequently, these data-centers run latency-critical workloads, whose performance is determined by tail latency, which is very sensitive to the interference of co-running workloads. To prevent QoS violations, cloud providers adopt overprovisioning strategies but they reduce the server utilization and increase the costs. A mechanism that accurately estimates performance degradation dynamically in a production system would allow cloud providers to improve the servers’ utilization. In this work we propose Cloud White, an approach that is able to detect the inter-VM interference in scenarios with multiple co-located latency-critical VMs and estimate the performance degradation using multi-variable regression models. Unlike previous proposals, Cloud White is built taking into account the limitations of a public cloud production system. Experimental results show that Cloud White is able to estimate performance degradation with a small overall prediction error of 5%.

Short-Term Load Forecasting Algorithm Based on LST-TCN in Power Distribution Network

In this paper, a neural network model called Long Short-Term Temporal Convolutional Network (LST-TCN) model is proposed for short-term load forecasting. This model refers to the 1-D fully convolution network, causal convolution, and void convolution structure. In the convolution layer, a residual connection layer is added. Additionally, the model makes use of two networks to extract features from long-term data and periodic short-term data, respectively, and fuses the two features to calculate the final predicted value. Long Short-Term Memory (LSTM) and Temporal Convolutional Network (TCN) are used as comparison algorithms to train and forecast 3 h, 6 h, 12 h, 24 h, and 48 h ahead of daily electricity load together with LST-TCN. Three different performance metrics, including pinball loss, root mean squared error (RMSE), and mean absolute error (RASE), were used to evaluate the performance of the proposed algorithms. The results of the test set proved that LST-TCN has better generalization effects and smaller prediction errors. The algorithm has a pinball loss of 1.2453 for 3 h ahead forecast and a pinball loss of 1.4885 for 48 h ahead forecast. Generally speaking, LST-TCN has better performance than LSTM, TCN, and other algorithms.

Tripability Analysis of Casing Strings in Directional Wells Using the Continuous Beam-Column and Buckling Theory

Hindered casing strings are often encountered in unconventional oil and gas exploration during the casing running process. This not only increases the operating costs and time but can also lead to downhole accidents and even abandonment in serious cases. Due to various assumptions, the calculation results of the existing soft models and hard models are different, which causes confusion for field operators when taking friction reduction measures. Moreover, a lowering force is often applied to assist hindered casing string running in a drilling field. However, its application is mainly based on work experience and lacks mechanistic analysis and theoretical guidance. Thus, in this study, a simulation model for the analysis of casing string tripability in a directional well was established and the model was combined with the continuous beam-column theory and buckling theory. The model was used to study how various factors including the friction coefficient, drilling fluid density, and casing diameter could affect the lowering force required when a casing string was hindered by buckling. The results showed that the maximum lowering force and the maximum effective lowering force decreased with the increase in the friction coefficient and the performance of the drilling fluid could be adjusted rapidly, which would be beneficial for ensuring that the casing string could be tripped smoothly by applying a lowering force. The increase in the drilling fluid density caused the maximum lowering force and the maximum effective lowering force to decrease, which was not conducive to hindered casing string running. The larger the casing diameter was, the greater the maximum lowering force and the maximum effective lowering force were. It was more convenient to apply a lowering force for a casing with a large diameter. In addition, the improved model could identify whether the casing string was in contact with the upper or lower borehole walls. Through finite element method verification, the prediction was in line with the actual casing running operation and the improved model has the smallest prediction error, i.e., 6.58%, compared with the existing models. Therefore, the improved model might provide necessary theoretical guidance for casing running operations in directional wells.

A dual learning-based recommendation approach

Data sparsity and cold start are two critical issues which need to be addressed in recommender systems (RSs). Currently, most methods address these issues by applying user history files or some side information to improve the user model and complete the rating matrix. However, such methods cannot perform well when labeled data is scarce or unavailable. In this paper, we propose a dual learning-based recommendation approach (DLRA). DLRA can trigger initial recommendation and improve the quality of recommendations by using the duality characteristics of RSs, even when the available labeled information is scarce. Specifically, DLRA regards the recommendation task as two independent subtasks — primal task and dual task, and these two tasks show strong duality in DLRA. The primal task is item-centered which aims to find users who can rate high for items, while the dual task is user-centered that aims to recommend the most favorite items to users. These two tasks have strong dualities in terms of the recommendation space, selection probability and recommendation basis. Based on these dualities, we design three dual learning strategies to couple the whole recommendation process and realize the self-tuning and self-improvement of each task model, and finally optimize the whole recommendation model. Based on the dataset of Movielens and BookCrossing, we simulate data sparsity and cold start recommendation scenarios, the experimental results show that DLRA achieves substantial improvement when the labeled data is scare, and it outperforms other hybrid recommendation approaches and deep learning strategies with a smaller predictive error as well as better recommendation accuracy.

Interference Signal Feature Extraction and Pattern Classification Algorithm Based on Deep Learning

Aiming at the scarcity of Low Earth Orbit (LEO) satellite spectrum resources, this paper proposes an algorithm of interference signal feature extraction and pattern classification based on deep learning to further improve the stability of satellite–ground communication links. The algorithm can successfully predict the interference signal pattern, start–stop time, frequency change range and other parameters, and has the advantages of excellent interference detection performance, high detection accuracy and small parameter prediction error, etc. It can be applied in the field of channel monitoring of communication satellite-to-ground communication links, and realize the repeated and efficient utilization of spectrum resources. Experiments show that the precision and recall of the algorithm for detecting five kinds of interference signals are all close to 100%, the prediction error of starting and ending time is less than 4 ms, and the prediction error of starting and ending frequency is less than 6 KHz.

Splitting tensile strength prediction of sustainable high-performance concrete using machine learning techniques.

Splitting tensile strength is one of the commonly used mechanical properties in the design of high-performance concrete (HPC) structures. To achieve accurate prediction of splitting tensile strength of HPC, two optimized machine learning models GA-ANN and GS-SVR were employed to predict the target tensile strength on 714 sets of sample data with 12 features of HPC as input variables. The appropriate initial weights and thresholds of the ANN model and hyper-parameters of the SVR model were first obtained by genetic algorithm and grid search, respectively. Then, the optimized models were used to train and test the dataset, and the performance of the models was evaluated and compared using evaluation metrics. The results showed that the prediction accuracy of the optimized GA-ANN model was higher than that of the pre-optimized ANN model, but still inferior to that of the GS-SVR model. Compared with the known machine learning models in the literature, the GS-SVR model proposed in this paper has smaller prediction error, higher prediction accuracy, and better performance. The strong generalization ability of the GS-SVR model to unseen test data indicates its great potential in predicting the tensile strength and is recommended as an alternative method for splitting tensile strength prediction of HPC. Additionally, a parametric analysis based on the Shapley additive explanations approach (SHAP) was proposed to investigate the importance and contribution of these input variables on the output splitting tensile strength.

Optimization of Network Home Management System Based on Big Data

Network home has become a research hotspot in today’s society, and it can improve the comfort, safety, and convenience of people’s lives. The traditional network home model only makes certain actions to the home system according to people’s instructions, and it is difficult to realize the intelligence of network home. This also limits the security and convenience of an online home. This study makes full use of the advantages of big data technology in processing nonlinear data, and applies the convolutional neural network (CNN) method and long and short-term memory (LSTM) neural network method to the network home system. CNN can be used to extract people’s behavior information, and LSTM can be used to extract people’s speech features. CNN method can establish the relationship between people’s behavior information, speech information, and network home management system. At the same time, this research mainly analyzes the lighting system, home appliance system, security system, and floor heating system in the network home system. The results show that the CNN-LSTM method has high accuracy in predicting the four systems of network home. The largest prediction error is only 2.78%, and this part of the error comes from the prediction of the home appliance system. The smallest prediction error is only 0.98%.

Modeling of temperature- and stress state-dependent yield and fracture behaviors for Mg-Gd-Y alloy

The research aims at characterizing and modeling the influence of temperature and stress state on the yield and fracture behaviors of an Mg-Gd-Y alloy. The mechanical experiments at 25∼300 ℃ were carried out by various designed specimens, including tension, compression and shear. The experimental result indicates that the strength of the Mg-Gd-Y alloy presents a monotonous and non-uniform downward trend as the temperature increases. The mechanical behavior at 100 ℃ and 150 ℃ are very similar, and the strength rapidly declines when the temperature is higher than 250 ℃. The tension-compression asymmetry presents the obvious nonlinear variation with the temperature and strain, which is weakened at elevated temperature. The fracture behavior belongs to the ductile fracture with tension and shear mechanisms, showing a coupling effect of temperature and stress state. A modified Zerilli-Armstrong model is proposed to characterize the plastic flow behavior with the coefficient of determination about 0.99, which is prior to the Johnson-Cook and Zerilli-Armstrong models. The Cazacu-Barlat2004 yield function is established to describe the temperature- and strain-related non-uniform evolution characteristic, and capture the fracture-related variables of all fracture tests with the acceptable prediction accuracy. The temperature-dependent parameter is introduced into the DF2016 fracture criterion to predict the onset of fracture under various temperatures and stress states with a small prediction error. The numerical simulation is conducted to validate the reliability and practicability of the established model.

Forecasting stock market with nanophotonic reservoir computing system based on silicon optomechanical oscillators.

The essence of stock market forecasting is to reveal the intrinsic operation rules of stock market, however it is a terribly arduous challenge for investors. The application of nanophotonic technology in the intelligence field provides a new approach for stock market forecasting with its unique advantages. In this work, a novel nanophotonic reservoir computing (RC) system based on silicon optomechanical oscillators (OMO) with photonic crystal (PhC) cavities for stock market forecasting is implemented. The long-term closing prices of four representative stock indexes are accurately forecast with small prediction errors, and the forecasting results with distinct characteristics are exhibited in the mature stock market and emerging stock market separately. Our work offers solutions and suggestions for surmounting the concept drift problem in stock market environment. The comprehensive influence of RC parameters on forecasting performance are displayed via the mapping diagrams, while some intriguing results indicate that the mature stock markets are more sensitive to the variation of RC parameters than the emerging stock markets. Furthermore, the direction trend forecasting results illustrate that our system has certain direction forecasting ability. Additionally, the stock forecasting problem with short listing time and few data in the stock market is solved through transfer learning (TL) in stock sector. The generalization ability (GA) of our nanophotonic reservoir computing system is also verified via four stocks in the same region and industry. Therefore, our work contributes to a novel RC model for stock market forecasting in the nanophotonic field, and provides a new prototype system for more applications in the intelligent information processing field.

Assessment of Machine Learning Models for the Prediction of Rate-Dependent Compressive Strength of Rocks

The prediction of rate-dependent compressive strength of rocks in dynamic compression experiments is still a notable challenge. Four machine learning models were introduced and employed on a dataset of 164 experiments to achieve an accurate prediction of the rate-dependent compressive strength of rocks. Then, the relative importance of the seven input features was analyzed. The results showed that compared with the extreme learning machine (ELM), random forest (RF), and the original support vector regression (SVR) models, the correlation coefficient R2 of prediction results with the hybrid model that combines the particle swarm optimization (PSO) algorithm and SVR was highest in both the training set and the test set, both exceeding 0.98. The PSO-SVR model obtained a higher prediction accuracy and a smaller prediction error than the other three models in terms of evaluation metrics, which showed the possibility of the model as a rate-dependent compressive strength prediction tool. Additionally, besides the static compressive strength, the stress rate is the most important influence factor on the rate-dependent compressive strength of the rock among the listed input parameters. Moreover, the strain rate has a positive effect on the rock strength.

Land cover reconstruction in Northwest China since 6 ka BP: Preliminary application of a new strategy

The mid-Holocene is generally regarded as a relatively warm period, and therefore reconstructing land cover change from the mid-Holocene onwards may help predict future changes under warming scenarios. In this study, we used a new strategy which combines the REVEALS model with modern analogue technique (MAT) to achieve a land cover reconstruction since 6 ka BP for Northwest (NW) China. The REVEALS results are more consistent with our general understanding of pollen-vegetation relationships in NW China than the pollen percentage data. However, although the model has potential for future applications, further quantitative testing is needed. Modern calibration sets can provide modern “analogues” for most of the fossil samples, but the numbers of modern samples from areas with a low vegetation cover are limited, which may lead to over- or underestimation of the results, despite the overall small prediction error. Our land cover results show that, since ∼6 ka BP, the vegetation cover in mountains areas was significantly greater than at lower elevations, with the threshold cover value of 45%. At lower elevations, the vegetation cover gradually decreased from east to west. Elevation differences and the distance from the sea may be the main reasons for the differences. The mountains were dominated by steppe, meadow/meadow steppe, and forest, while the lower elevations were dominated by desert and desert steppe, with generally only small changes in vegetation type, reflecting the long-term arid climatic background of the region. However, since ∼6 ka BP there were several differences in the pattern of dry-wet climate shifts across the study region. The vegetation in most areas indicates a trend towards wetter conditions, or no significant trend, except for sites G1 and G3 which show an aridification trend.

A framework to model global, regional, and national estimates of intimate partner violence

BackgroundAccurate and reliable estimates of violence against women form the backbone of global and regional monitoring efforts to eliminate this human right violation and public health problem. Estimating the prevalence of intimate partner violence (IPV) is challenging due to variations in case definition and recall period, surveyed populations, partner definition, level of age disaggregation, and survey representativeness, among others. In this paper, we aim to develop a sound and flexible statistical modeling framework for global, regional, and national IPV statistics.MethodsWe modeled IPV within a Bayesian multilevel modeling framework, accounting for heterogeneity of age groups using age-standardization, and age patterns and time trends using splines functions. Survey comparability is achieved using adjustment factors which are estimated using exact matching and their uncertainty accounted for. Both in-sample and out-of-sample comparisons are used for model validation, including posterior predictive checks. Post-processing of models’ outputs is performed to aggregate estimates at different geographic levels and age groups.ResultsA total of 307 unique studies conducted between 2000–2018, from 154 countries/areas, and totaling nearly 1.8 million unique women responses informed lifetime IPV. Past year IPV had a similar number of studies (n = 332), countries/areas represented (n = 159), and individual responses (n = 1.8 million). Roughly half of IPV observations required some adjustments. Posterior predictive checks suggest good model fit to data and out-of-sample comparisons provided reassuring results with small median prediction errors and appropriate coverage of predictions’ intervals.ConclusionsThe proposed modeling framework can pool both national and sub-national surveys, account for heterogeneous age groups and age trends, accommodate different surveyed populations, adjust for differences in survey instruments, and efficiently propagate uncertainty to model outputs. Describing this model to reproducible levels of detail enables the accurate interpretation and responsible use of estimates to inform effective violence against women prevention policy and programs, and global monitoring of elimination efforts as part of the Sustainable Development Goals.

Nonlinear Mixed-Effects Height to Crown Base Model for Moso Bamboo (Phyllostachys heterocycla (Carr.) Mitford cv. Pubescens) in Eastern China

Height to crown base (HCB) is an important variable used as a predictor of forest growth and yield. This study developed a nonlinear, mixed-effects HCB model through inclusion of plot-level random effects using data from 29 sample plots distributed across a state-owned Yixing forest farm in Jiangsu province, eastern China. Among several predictor variables evaluated in the analyses, bamboo height, canopy density, and total basal area of bamboo with a diameter larger than that of the subject bamboo individual contributed significantly to the HCB variations. The inclusion of random effects improved the prediction accuracy of the model significantly, indicating that the HCB variations within and across the sample plots were substantial. The model was localized using four sampling strategies, and the study identified that using two medium-sized bamboos by diameter at breast height per sample plot resulted in the smallest prediction error. This strategy, which would balance both measurement cost and potential error, may be applied to estimate the random effects and localization of the nonlinear mixed-effects HCB model for moso bamboo in eastern China.

Atypical action updating in a dynamic environment associated with adolescent obsessive-compulsive disorder.

Computational research had determined that adults with obsessive-compulsive disorder (OCD) display heightened action updating in response to noise in the environment and neglect metacognitive information (such as confidence) when making decisions. These features are proposed to underlie patients' compulsions despite the knowledge they are irrational. Nonetheless, it is unclear whether this extends to adolescents with OCD as research in this population is lacking. Thus, this study aimed to investigate the interplay between action and confidence in adolescents with OCD. Twenty-seven adolescents with OCD and 46 controls completed a predictive-inference task, designed to probe how subjects' actions and confidence ratings fluctuate in response to unexpected outcomes. We investigated how subjects update actions in response to prediction errors (indexing mismatches between expectations and outcomes) and used parameters from a Bayesian model to predict how confidence and action evolve over time. Confidence-action association strength was assessed using a regression model. We also investigated the effects of serotonergic medication. Adolescents with OCD showed significantly increased learning rates, particularly following small prediction errors. Results were driven primarily by unmedicated patients. Confidence ratings appeared equivalent between groups, although model-based analysis revealed that patients' confidence was less affected by prediction errors compared to controls. Patients and controls did not differ in the extent to which they updated actions and confidence in tandem. Adolescents with OCD showed enhanced action adjustments, especially in the face of small prediction errors, consistent with previous research establishing 'just-right' compulsions, enhanced error-related negativity, and greater decision uncertainty in paediatric-OCD. These tendencies were ameliorated in patients receiving serotonergic medication, emphasising the importance of early intervention in preventing disorder-related cognitive deficits. Confidence ratings were equivalent between young patients and controls, mirroring findings in adult OCD research.

A Prediction Network for Hydraulic Support Pressure Based on Multitimescale Feature Extraction

The roof water hazard has severely weakened the safety of coal mine production. As the most direct data on the state of the roof, the monitoring big data of the hydraulic support is of great importance to coal mine safety. This paper proposes a neural network model based on multitimescale feature extraction for predicting the pressure-bearing data of hydraulic supports in coal mine working faces, drawing on the idea of multiscale time feature extraction in the empirical modal decomposition algorithm. The proposed model can excel in capturing temporal features at multiple scales in the time series, thus improving the prediction accuracy of the neural network model. We collected 80 days of monitoring data from 17 hydraulic supports in the working face and carried out comparative experiments with several existing models. The experiments show that the neural network model proposed in this paper is stable and reliable, with small prediction errors, and can provide important help for the prevention and control of the roof water hazard.

A new lead-acid battery state-of-health evaluation method using electrochemical impedance spectroscopy for second life in rural electrification systems

Providing electricity at an affordable price wherever one lives is still a concern. The use of batteries to store electrical energy is very common practice but can be a weak link in a system due to failures. Re-using old lead-acid car batteries discarded from thermal engine vehicles could be a cheap electricity storage solution for remote areas, in line with the principles of frugal innovation. Evaluating the State of Health (SoH) of these batteries is of the utmost importance for the second life of car lead-acid batteries. Some methods exist, such as the two-pulse method or the internal resistance method, but their results are not always reliable. Electrochemical Impedance Spectroscopy is another classical technique; however, a reliable formula is still required to link the SoH and EIS with unknown battery history. A new and original way to determine the SoH from EIS is detailed in this work. Old and new batteries are used for experimental prediction at different states of charge. A small average prediction error is found for 56 different configurations. The newly proposed method is named “CdS based method”. Results from this method, are compared to results from existing methods and improvement is clear.

Incorporating spatial autocorrelation into house sale price prediction using random forest model

AbstractThe random forest model has been frequently used for house sale price prediction because researchers have shown that it can create better results with smaller prediction errors than conventional statistical models. However, house sale prices tend to be spatially autocorrelated due to shared neighborhood environments and similar physical characteristics, which have been rarely considered and accounted for in the random forest model. A failure to address spatial autocorrelation may result in over‐ or underpredicted house sale prices, especially when these sale prices are highly spatially dependent. This research proposes an extended random forest approach, introducing additional proxy variables that furnish geographic proximity measures of house locations in order to capture spatial autocorrelation that is unexplained by covariates. An application of the extended random forest approach for house sale prices in Fairfax County, Virginia, confirms the effectiveness of the proxy variables with an improvement of house sale price prediction accuracy. The introduction of the proxy variables can correct over‐ and under‐predicted house sale prices that show an underlying spatial structure.

SUITOR: Selecting the number of mutational signatures through cross-validation.

For de novo mutational signature analysis, the critical first step is to decide how many signatures should be expected in a cancer genomics study. An incorrect number could mislead downstream analyses. Here we present SUITOR (Selecting the nUmber of mutatIonal signaTures thrOugh cRoss-validation), an unsupervised cross-validation method that requires little assumptions and no numerical approximations to select the optimal number of signatures without overfitting the data. In vitro studies and in silico simulations demonstrated that SUITOR can correctly identify signatures, some of which were missed by other widely used methods. Applied to 2,540 whole-genome sequenced tumors across 22 cancer types, SUITOR selected signatures with the smallest prediction errors and almost all signatures of breast cancer selected by SUITOR were validated in an independent breast cancer study. SUITOR is a powerful tool to select the optimal number of mutational signatures, facilitating downstream analyses with etiological or therapeutic importance.

Prediction of Self-Healing of Engineered Cementitious Composite Using Machine Learning Approaches

Engineered cementitious composite (ECC) is a unique material, which can significantly contribute to self-healing based on ongoing hydration. However, it is difficult to model and predict the self-healing performance of ECC. Although different machine learning (ML) algorithms have been utilized to predict several properties of concrete, the application of ML on self-healing prediction is considerably rare. This paper aims to provide a comparative analysis on the performance of various machine learning models in predicting the self-healing capability of ECC. These models include four individual methods, linear regression (LR), back-propagation neural network (BPNN), classification and regression tree (CART), and support vector regression (SVR). To improve prediction accuracy, three ensemble methods, namely bagging, AdaBoost, and stacking, were also studied. A series of experimental works on the self-healing performance of ECC samples was conducted, and the results were used to develop and compare the accuracy among the ML models. The comparison results showed that the Stack_LR model had the best predictive performance, showing the highest coefficient of determination (R2), the lowest root-mean-squared error (RMSE), and the smallest prediction error (MAE). Among all individual models studies, the BPNN model performed the best in terms of the RMSE and R2, while SVR performed the best in terms of the MAE. Furthermore, SVR had the smallest prediction error (MAE) for crack widths less than 60 μm or greater than 100 μm, while CART had the smallest prediction error (MAE) for crack widths between 60 μm and 100 μm. The study concluded that the individual and ensemble methods can be used to predict the self-healing of ECC. Ensemble models were able to improve the accuracy of prediction compared to the individual model used as their base learner, i.e., a 2.3% to 4.9% reduction in MAE. However, selecting an appropriate individual and ensemble method is critical. To improve the performance accuracy, researchers should employ different ensemble methods to compare their effectiveness with different ML models.

Harmful algal bloom warning based on machine learning in maritime site monitoring

Forecasting harmful algal blooms (HABs) is an important part of marine environmental monitoring. Algae bloom observation channels includes satellite remote sensing and maritime station monitoring (MSM). Compared with satellite remote sensing image, MSM can collect more accurate data, which includes various seawater inorganic salt content related to the HABs. However, the measured data of MSM is easily affected by regional sediment content and seawater dynamic field, which leads to lack the accurate forecasting models. Therefore, this paper proposed a local spatio-temporal HABs forecasted model (STHFM) based on few impact factors in MSM. The advantage of this model is to first use principal component analysis to select main environment factors (MEFs) related to HABs. Then, this model distinguishes multiple warning levels of HABs in spatio-temporal according to the algae growth rate. Finally, this paper generates continuous MEFs time series information based on the Autoregressive Integrated Moving Average (ARIMA) model in the high-level warning area. And an improved LSTM network with MEFs time series as input is established to forecast HABs in future. The proposed model is tested on the NOAA website public dataset, which contains historical harmful algae Alexandrium data on the East Coast of US. The experimental shows that our model has good HABs monitoring performance. Under the public NOAA Alexandrium dataset, the proposed model can achieve the highest prediction accuracy of 82.1%, and has a small prediction error.