Predictive Error Compensating Wavelet Neural Network Model for Multivariable Time Series Prediction

Multivariable machine learning (ML) models are increasingly used for time series predictions. However, avoiding the overfitting and underfitting in ML-based time series prediction requires special consideration depending on the size and characteristics of the available training dataset. Predictive error compensating wavelet neural network (PEC-WNN) improves the time series prediction accuracy by enhancing the orthogonal features within a data fusion scheme. In this study, time series prediction performance of the PEC-WNNs have been evaluated on two different problems in comparison to conventional machine learning methods including the long short-term memory (LSTM) network. The results have shown that PECNET provides significantly more accurate predictions. RMSPE error is reduced by more than 60% with respect to other compared ML methods for Lorenz Attractor and wind speed prediction problems.

Predictive Refractive Error Post Congenital and Developmental Cataract Surgery

Cataract is one of the leading treatable causes of visual impairment in children. Visual rehabilitation is crucial for the development of good visual function after cataract surgery in children. The research aimd to describe post-operative Predictive Refractive Error (PRE) in congenital and developmental cataracts in Cicendo National Eye Hospital from January 2017 to December 2018. This is a retrospective analytic observational study from medical records. We found 107 eyes of 62 children with congenital and developmental cataracts had had cataract surgery and primary implantation of Intraocular Lens (IOL) in Pediatric Ophthalmology and Strabismus Unit, Cicendo National Eye Hospital. The patients were divided into two groups, with axial length (AXL) of ≤ 24 mm and > 24 mm. The paired t-test was used to compare Predictive Error (PE) in SRK/T, SRK II, and Showa SRK formula. Mean age at surgery was 6.7 ± 4.0 years. Ninety-five eyes had AXL ≤ 24 mm, and 12 eyes had AXL > 24 mm. Prediction Error from patients with AXL ≤ 24 mm was 0.29 D, and from patients with AXL > 24 mm was 2.40 D in SRK/T formula (P < 0.05). There was no significant difference between PE and Absolute Predictive Error (APE) in SRK/T, SRK II, and Showa SRK in patients with AXL > 24 mm (P > 0.05). SRK/T is the most predictable formula in patients with AXL ≤ 24 mm. There is no significant difference in patients with AXL > 24 mm in all formulas. Keywords: congenital and developmental cataract, axial length, Prediction Error, intraocular lens

Accurate Prediction of Children's ADHD Severity Using Family Burden Information: A Neural Lasso Approach

The deep lasso algorithm (dlasso) is introduced as a neural version of the statistical linear lasso algorithm that holds benefits from both methodologies: feature selection and automatic optimization of the parameters (including the regularization parameter). This last property makes dlasso particularly attractive for feature selection on small samples. In the two first conducted experiments, it was observed that dlasso is capable of obtaining better performance than its non-neuronal version (traditional lasso), in terms of predictive error and correct variable selection. Once that dlasso performance has been assessed, it is used to determine whether it is possible to predict the severity of symptoms in children with ADHD from four scales that measure family burden, family functioning, parental satisfaction, and parental mental health. Results show that dlasso is able to predict parents' assessment of the severity of their children's inattention from only seven items from the previous scales. These items are related to parents' satisfaction and degree of parental burden.

Predicting shear failure in reinforced concrete members using a three-dimensional peridynamic framework

The assumptions made in design codes can result in unconservative predictions of shear strength for reinforced concrete members. The limitations of empirical methods have prompted the development and use of numerical techniques. A three-dimensional bond-based peridynamic framework is developed for predicting shear failure in reinforced concrete members. The predictive accuracy and generality of the framework is assessed against existing experimental results. Nine reinforced concrete beams that exhibit a wide range of failure modes are modelled. The shear-span-to-depth ratio is systematically varied from 1 to 8 to facilitate a study of different load-transfer mechanisms and failure modes. A comprehensive validation study such as this has until now been missing in the peridynamic literature. A bilinear constitutive law is employed, and the sensitivity of the model is tested using two levels of mesh refinement. The predictive error between the experimental and numerical failure loads ranges from +3% to -57%, highlighting the importance of validation against a series of problems. The results demonstrate that the model captures many of the factors that contribute to shear and bending resistance. New insights into the capabilities and deficiencies of the peridynamic model are gained by comparing the expected load-transfer mechanisms with the predictive error.

Basic Statistical Estimation Outperforms Machine Learning in Monthly Prediction of Seasonal Climatic Parameters

Machine learning (ML) has been utilized to predict climatic parameters, and many successes have been reported in the literature. In this paper, we scrutinize the effectiveness of five widely used ML algorithms in the monthly prediction of seasonal climatic parameters using monthly image data. Specifically, we quantify the predictive performance of these algorithms applied to five climatic parameters using various combinations of features. We compare the predictive accuracy of the resulting trained ML models to that of basic statistical estimators that are computed directly from the training data. Our results show that ML never significantly outperforms the statistical baseline, and underperforms for most feature sets. Unlike previous similar studies, we provide error bars for the relative performance of different predictors based on jackknife estimates applied to differences in predictive error magnitudes. We also show that the practice of shuffling data sequences which was employed in some previous references leads to data leakage, resulting in over-estimated performance. Ultimately, the paper demonstrates the importance of using well-grounded statistical techniques when producing and analyzing the results of ML predictive models.

Model Predictive Current Control of Multiphase Motor at Low Carrier Ratio

Multiphase motors have multiple control planes, and harmonics are decoupled in different planes. Multiphase motors can improve magnetic field distribution, power density and core utilization by injecting certain harmonic currents into the harmonic planes. In the harmonic plane control process, due to the switching frequency of the inverter being limited, the ratio of the switching frequency to the current frequency (the carrier ratio) of the harmonic plane is low, the digital control delay increases, and the inverter output current contains more harmonics, which makes it difficult for the proportional-integral (PI) current controller to effectively control the d-axis and q-axis currents of the harmonic plane and thus unable to track the given values stably. Moreover, the PI current controller is relatively dependent on the motor parameters. For these reasons, a model predictive current control method with predictive error compensation is proposed. Taking a nine-phase induction motor as an example, the control voltage is calculated by the cost function and corrected by the current predictive error, which realizes the current control method at a low carrier ratio. Additionally, the robustness of the control method is analyzed after the parameters of the multiphase motor have large errors. The experimental results show that the proposed method can control the current of the harmonic plane at low carrier ratio, accurately track the harmonic current commands and attain strong robustness for the motor parameters.

North American historical monthly spatial climate dataset, 1901–2016

We present historical monthly spatial models of temperature and precipitation generated from the North American dataset version “j” from the National Oceanic and Atmospheric Administration’s (NOAA’s) National Centres for Environmental Information (NCEI). Monthly values of minimum/maximum temperature and precipitation for 1901–2016 were modelled for continental United States and Canada. Compared to similar spatial models published in 2006 by Natural Resources Canada (NRCAN), the current models show less error. The Root Generalized Cross Validation (RTGCV), a measure of the predictive error of the surfaces akin to a spatially averaged standard predictive error estimate, averaged 0.94 °C for maximum temperature models, 1.3 °C for minimum temperature and 25.2% for total precipitation. Mean prediction errors for the temperature variables were less than 0.01 °C, using all stations. In comparison, precipitation models showed a dry bias (compared to recorded values) of 0.5 mm or 0.7% of the surface mean. Mean absolute predictive errors for all stations were 0.7 °C for maximum temperature, 1.02 °C for minimum temperature, and 13.3 mm (19.3% of the surface mean) for monthly precipitation.