Regression Neural Network Models Research Articles

Study on Life Prediction Method of Ball Screw Base on Constructed Degradation Feature and IGWO-BiLSTM

Regarding the problem of reduced remaining useful life (RUL) due to wear of the ball screw in the feed system of CNC (computer numerical control) machine tools, a prediction method based on constructing the degradation feature vector of the signal data and the improved gray-wolf optimization with bidirectional long short-term memory (IGWO-BiLSTM) neural network regression model is proposed. Firstly, a time-domain analysis and the complete ensemble empirical mode decomposition with adaptive noise analysis (CEEMDAN) were carried out based on the collected life cycle signal data of a ball screw. The time-domain feature vector and the energy feature vector of each IMF (intrinsic mode function) component after CEEMDAN decomposition were constructed. The Pearson correlation coefficient was used to filter feature vectors and construct the multivariate feature vector. Secondly, this paper improves the traditional gray wolf optimization algorithm, adds a search strategy based on dimension learning, and combines the improved algorithm with the BiLSTM model, based on the IGWO-BiLSTM theory. A regression model between feature vectors and the remaining life of a ball-screw system was established. Finally, the prediction model was established according to the proposed method and compared with the other five neural network models: LSTM, BiLSTM, BO-LSTM (Bayesian optimization of LSTM), BO-BiLSTM, and IGWO-LSTM. The results indicate that this method has high accuracy and good generalization ability for predicting the remaining life of a ball-screw system.

A neural network approach for classification of fault-slip data in geoscience

In geoscience, paleostress studies are a vital tool for understanding the tectonic evolution of the region. The collected hundreds or even thousands of heterogeneous fault-slip data need to be divided into homogeneous (i.e., belonging to similar tectonic environments) subgroups by geologists. Computer-based paleostress inversion programs are able to run homogenous data sets. Here, we are aiming for the classification of heterogeneous fault-slip data into homogeneous sub-data which is performed by using several techniques in Machine Learning (ML) algorithms, which are Artificial Neural Network (ANN), Naïve Bayes (NB) and Logistic Regression (LR) models. When these models are executed on the Anaconda Navigator interface with python language, the accuracies are obtained as 87.17 % for ANN, 79.71 % LR, and 62.1 % NB.

A machine learning model to predict treatment initiation among new patients in a community oncology network.

e13539 Background: In The US Oncology Network (The Network), about one-third of new patients with a cancer diagnosis started intravenous (IV) treatment after their first visit. The rest of the patients either came in for a consult only or might have received other treatments such as radiation, surgery, or oral therapy. We developed a machine learning model to predict IV treatment initiation among new patients and discovered features associated with the patient’s decision. This model could suggest interventions to improve patient’s access to care. Methods: A retrospective cohort was formed by identifying new patients with cancer from 27 practices in The Network between July 1, 2021 and June 30, 2022. Structured data were extracted and processed from the electronic health records, claims, physician referrals, and the American Community Survey. Patient characteristics included demographics, clinical information, payor types, and socioeconomic status. The referral pattern and the geographic region of practices, and the provider workload were considered as well. Gradient-boosted decision trees, random forest, neural network, and logistic regression models were developed to predict the probability of starting IV treatment within 90 days of the first visit. Model performance was evaluated based on the area under the receiver operating characteristic (AUROC) curve using cross-valuation and 4:1 training/validation random split. Shapley Additive Explanations (SHAP) values were applied to the model to explain feature importance. Results: A total of 117,340 new patients with a cancer diagnosis were included in the study, of whom 35% initiated IV treatment within 90 days of the first visit. A gradient-boosted decision tree algorithm with control of the imbalanced label was chosen as the final model because of the performance and the ability to handle missing values. The model achieved an AUROC of 0.80 on the validation dataset with both cross-valuation and 4:1 training/validation random split. Based on the SHAP values (log odds), we found that clinical information including diagnosis and stage is the most important feature to predict the initiation of IV treatment (mean absolute SHAP = 0.31 and 1.03, respectively). Medicaid contributes least to treatment initiation among all insurance types (mean absolute SHAP = 0.01). In addition, younger age and male patients have a higher chance to start IV treatment (Pearson correlation = -0.41, p-value < 0.01 for age versus SHAP values; p-value < 0.01, two-sided T-test for SHAP values by gender). Conclusions: This study reports a machine learning model to predict IV treatment initiation among new patients with cancer. Clinical features impact the treatment decision more than others. This model could guide patient service and direct personalized care navigation. Further, the model sheds light on future interventions that could enhance patient access to treatment promptly.

Prediction of molten pool temperature in laser solid forming based on PSO-LSTM

The temperature of the molten pool in Laser Solid Forming has a direct effect on the dimensional accuracy and mechanical properties of the parts. Accurate prediction of the melt pool temperature is important to ensure the stability of the melt pool temperature and to improve the forming accuracy and quality of the part. In order to accurately predict the melt pool temperature, this study proposes a melt pool temperature prediction method based on particle swarm optimization (PSO) optimised long short-term memory neural network (LSTM). Using IR camera to obtain melt pool temperature data and establish long short-term memory neural network melt pool temperature prediction model based on experimental data. Optimization of the initial learning rate and the number of hidden layer units of the long short-term memory neural network model using the particle swarm optimization algorithm to build a PSO-LSTM model for prediction of melt pool temperature. The results show that the PSO-LSTM prediction model outperforms the long short-term memory neural network and Ridge Regression models in all evaluation indicators and can achieve accurate prediction of melt pool temperature.

Quantification of alloying elements in steel targets: The LIBS 2022 regression contest

We present the results of the regression contest organized for the LIBS 2022 conference. While the motivation and design of the contest are briefly presented, the work focuses on the methodologies of the three best-performing teams. The employed spectral preprocessing strategies, choice of regression models and its optimization are detailed for each team separately. The aim of the contest reflects the long-term challenges faced by quantitative laser-induced breakdown spectroscopy (LIBS) analysis. Thus, the contest was designed with the purpose of providing a transparent platform for comparing and evaluating the large range of data processing tools available in the LIBS literature. Namely, the contest consisted of the quantification of two major (Cr, Ni) and two minor (Mn, Mo) elements in 15 steel targets. For constructing an appropriate regression model, spectra of 42 targets were provided. The spectra were collected using a commercially available laboratory-based LIBS system and made publicly available. The contest lasted 53 days during which the teams did not receive feedback. In total, 21 teams participated out of which the three best-performing methodologies are presented here. A single linear partial least squares model and two artificial neural network regression models are presented. The corresponding feature selection strategies included emission line selection, spectral range selection, and automatized wavelength selection. Various spectral normalization strategies and data augmentation strategies are also presented.

Assessing the protection mechanisms on Enterobacter aerogenes ATCC 13048 by potentially probiotic strain Lacticaseibacillus casei XN18: An experimental and modeling study

In this study, Lacticaseibacillus casei XN18 had a remarkable resistant to simulated gastrointestinal conditions, hydrophobicity (38.60%), auto-aggregation (29.80%), co-aggregation (21.10%), adhesion (9.50%), anti-adhesion (24.40–36.90%), antioxidant activity (46.47%), cholesterol assimilation (41.10%), and antimicrobial effect on some pathogenic microorganisms. The modified double layer method, and Enterobacter aerogenes (inhibition zone (IZ) = 9.10 mm) and Listeria monocytogenes (IZ = 14.60 mm) were the most sensitive and resistant pathogens to the probiotic strain. The Lb. casei was sensitive to ciprofloxacin (IZ = 23 mm) and nitrofurantoin (IZ = 25.10 mm), semi-sensitive to imipenem (IZ = 18.80 mm), erythromycin (IZ = 16.90 mm), and chloramphenicol (IZ = 17.90 mm), and resistant to ampicillin (IZ = 9.60 mm) and nalidixic acid (IZ = 9.90 mm). The Lb. casei showed no haemolytic and DNase properties, and it could therefore be used for health-promoting purposes. In the next section, multilayer perceptron (MLP) neural network (NN) and gaussian process regression (GPR) models with k-fold cross validation method were used for predicting the rate of probiotic viability based on three levels of pH and time. The results showed that GPR has the lowest error. The mean absolute percentage error (MAPE), root mean absolute error (RMSE) and coefficient of determination (R2) for GPR and MLP models were 1.49 ± 0.40, 0.21 ± 0.03, 0.98 ± 0.05 and 6.66 ± 0.98, 0.83 ± 0.23 0.82 ± 0.09, respectively. So, the GPR model can be reliably used as a useful method to predict the probiotic viability in similar cases.

Stability Prediction of Soil Slopes Based on Digital Twinning and Deep Learning

This paper proposes a slope stability prediction model based on deep learning and digital twinning methods. To establish a reliable slope database, 30 actual slopes were collected, and 100 digital twin (DT) models were generated for each actual slope by fine-tuning the slope profiles. The safety factors of all slope samples were calculated using the Limit Equilibrium Methods (LEMs). A convolutional neural network (CNN) regression model was established, and the root mean square error (RMSE) was used as the evaluation indicator. In order to find an excellent CNN model, the K-fold (K = 10) cross-validation was used to compare the predictive effect of 1D CNN and 2D CNN on the slope safety factor. On this basis, CNN models with different network depths were compared. The results showed that the 2D CNN model with six convolutional layers had the best network prediction effect for the slope dataset. To validate the generalization ability of the model, an actual slope was input into the CNN model; its prediction result was 1.0229, and the absolute error with its real safety factor (1.0197) was 0.0032. With the slope stability prediction model proposed in this paper, the safety factor of slopes can be obtained from their geological and physical data, which greatly simplifies the calculation of the safety factor and has great engineering significance.

Using econometric and machine learning models to forecast crude oil prices: Insights from economic history

The volatility of the crude oil market and its effects on the global economy increased the concerns of individual investors, states/governments, and corporations. Forecasting the price of crude oil is difficult owing to its complicated, nonlinear, and chaotic nature in economic history. Multiple variables influence crude oil prices, such as the economic history, economic cycle, international relations, and geopolitics. Predicting the price of crude oil is a complex but valuable endeavor. Crude oil price forecasting is done using historical data (time series method) or dependent variables/factors (regression method) using traditional econometric or machine learning models. In this study, we use both methods (regression and time series) to examine the prediction performance of both models (econometric and machine learning models) for daily WTI crude oil prices covering the period December 18, 2011, through December 31, 2018. We present a performance analysis of conventional econometric models (ARIMA, GARCH, and OLS), Artificial Neural Network (ANN) regression models, and ANN Time Series models to compare their results to find out the best-performing method (time series or regression) and the best model (econometric or machine learning model). Based on our study results, we propose a novel Artificial Neural Network model to improve the prediction performance of existing models by adjusting the bias and weights of ANN hidden layers. We used historical prices of 14 different variables, including gold, silver, S&P500, USD Index price, and US-EU conversion rates for regression models, whereas historical time series data of WTI crude oil for time series models. Analysis of the results reveals that the performance of our proposed model remained better than all tested models. The comparative results of existing models show that the overall performance of Neural Networks remained better than econometric models. Our results have substantial implications for governments, businesses, and investors, and for the sustainable growth of economies that rely on energy.

Multimodal fraudulent website identification method based on heterogeneous model ensemble

The feature analysis of fraudulent websites is of great significance to the combat, prevention and control of telecom fraud crimes. Aiming to address the shortcomings of existing analytical approaches, i.e. single dimension and venerability to anti-reconnaissance, this paper adopts the Stacking, the ensemble learning algorithm, combines multiple modalities such as text, image and URL, and proposes a multimodal fraudulent website identification method by ensembling heterogeneous models. Cross-validation is first used in the training of multiple largely different base classifiers that are strong in learning, such as BERT model, residual neural network (ResNet) and logistic regression model. Classification of the text, image and URL features are then performed respectively. The results of the base classifiers are taken as the input of the meta-classifier, and the output of which is eventually used as the final identification. The study indicates that the fusion method is more effective in identifying fraudulent websites than the single-modal method, and the recall is increased by at least 1%. In addition, the deployment of the algorithm to the real Internet environment shows the improvement of the identification accuracy by at least 1.9% compared with other fusion methods.

Non-destructive determination of soluble solids content in intact apples using a self-made portable NIR diffuse reflectance instrument

The soluble solids content (SSC) is an important internal quality parameter of fruits that is monitored on the fruit market to meet different consumer demands. The aim of this study was to develop a self-made portable near-infrared (NIR) diffuse reflectance instrument to evaluate and monitor the SSC of intact apples. The visible and near-infrared (Vis-NIR) spectra of 118 ‘Fuji’ apples were collected using a Vis-NIR diffuse reflectance spectroscopic measurement system in the spectral range of 450–1100 nm. Savitzky-Golay convolution smoothing and first derivative methods were subsequently applied to eliminate noise interference and baseline shift. Successive projections algorithm (SPA) was performed to extract feature wavelengths from the pretreatment spectra, and the back-propagation artificial neural network (BP-ANN) and multivariate nonlinear regression (MNLR) models were employed to evaluate the predictive ability of the extracted and selected feature wavelengths for the SSC of apples. Seven wavelengths (881, 890, 901, 926, 941, 951, 978 nm) were selected as optimal feature wavelengths from the extracted feature wavelengths (18 wavelengths). The MNLR model (R2 = 0.953, RMSE = 0.391 %) exhibited a high prediction accuracy compared to the BP-ANN model (R2 = 0.865, RMSE = 0.754 %). Based on the results, NIR combined with the MNLR model could be applied in the self-made portable instrument to innovatively monitor the SSC of apples. Finally, a self-made portable NIR diffuse reflectance instrument was designed and developed, and the spectral information of 298 ‘Fuji’ apples was collected using the instrument. According to the results, the MNLR model could well predict the SSC of apples (R2 = 0.871, RMSE = 0.687 %). The overall results also revealed that the developed portable NIR instrument is a promising device for rapid non-destructive monitoring of fruit SSC, thereby meeting the practical application requirement of postharvest commercial processing systems. This instrument could also be beneficial to customers, growers, and producers.

Predicting injury-severity for cyclist crashes using natural language processing and neural network modelling

The use of machine learning techniques in safety research has increased as has the interest in using new data sources. This study’s unique contribution is the application of text mining—focusing on perceived cyclist safety and crash occurrence in an urban environment. We analysed crash data collected by the emergency forces in the Capital Region of Denmark from 2013 to 2017 and self-reported textual data provided by cyclists from 2018 to 2019. The analysis included natural language processing and topic modelling to identify Latent Dirichlet Allocation (LDA) topics from self-reports, representing environment characteristics that cyclists’ perceive as unsafe. A multi-output neural network regression model is applied to predict the injury-severity distribution of cyclists involved in crashes (measured by emergency response level [ERL]) based on the obtained topic distributions together with additional variables like cycle flow. We identified six LDA topics which address buses and cycle paths, conflicts with parked cars, roundabouts and inadequate maintenance, fast-moving cars and lack of cycle path, school zones and heavy traffic, and intersections and interactions with vehicles. Cycle flow was found to be the highest impacter on ERL prediction. However, other factors also impacted ERLs, especially school zones and heavy traffic. The results bring new insights into safety perception and actual safety for cyclists. The results contribute to a novel procedure for the joint correlation analysis using machine learning techniques on self-reported textual data thereby providing a better tool for infrastructure planning. The findings show the importance of including perceived safety in crash modelling and that authorities should focus on safety around schools and in intersections in order to improve safety for cyclists in a urban environment.

Neural age screening on question answering communities

For online social networks, demographic analysis is absolutely essential for improving their services in many ways. It is instrumental in understanding their different audiences, members and competitors. As well as that, it is pivotal in designing effective personalization and contextualization strategies, especially for displaying and creating better content. There is, for this reason, a great bulk of research into how demographic variables are characterized and how they impact online platforms such as Facebook and Twitter. But surprisingly, only a handful of works delve into their characterization and effect on community Question-Answering (cQA) websites. In this particular context, the subject of age demographics remains largely unexplored.This paper takes the lead on interpreting automatic age recognition on CQAs (a.k.a. age screening) as a regression task. To this effect, it compares state-of-the-art graph-based neural network regression and embedding models on a massive activity-graph encompassing ca. 16 and 837 million nodes (members) and edges, respectively. For this study, a large-scale subset of ca. 657,000 community fellows was automatically associated with their age via aligning their profile texts with a limited number of linguistic patterns.In short, our results show that Node2vec significantly outperforms other embeddings regardless of the regression model used for casting predictions. When this embedding is combined with Artificial Neural Network Regressions, we obtained our best configuration scoring a Root Mean Square Error (RMSE) of 8.39. An interesting qualitative feature of this embedding space is that age-based centroid vectors tend to form a trail ordered by age. Lastly, our outcomes also signal that activity graph based models can rival its counterparts based on image and textual inputs, paving the way for constructing effective multi-modal approaches.

Crop Yield Prediction Using Gradient Boosting Neural Network Regression Model

The finest utility sector is agriculture, especially in emerging nations like India. Utilizing historical data in agriculture can change the context of decision-making and increase farmer productivity. Approximately a part of India's population is employed in agriculture, however this sector contributes just 14% of the country's GDP. This can be explained in part by farmers not making sufficient decisions on yield forecast. By examining numerous climatic elements, such as rainfall, and land characteristics, such as soil type and ground water salinity, as well as historical records of crops cultivated, the suggested machine learning technique tries to estimate the agricultural yield for a certain location. Finally, we anticipate that our proposed Machine Learning Gradient Boosting Neural Network Regression (Grow Net) model was predicting the accurate yield. Finally our system is expected to predict the yield based on dataset we have taken. We were compared our proposed algorithm with various Machine Learning algorithms such as Random Forest, Support Vector Machine, KNN, Multi-layer Perceptron Regressor, Gradient Boosting Regressor and results shows that proposed was given best RMSE ,MAE and R2 value.

Soil-Surface-Image-Feature-Based Rapid Prediction of Soil Water Content and Bulk Density Using a Deep Neural Network

This study aimed to develop a deep neural network model for predicting the soil water content and bulk density of soil based on features extracted from in situ soil surface images. Soil surface images were acquired using a Canon EOS 100d camera. The camera was installed in the vertical direction above the soil surface layer. To maintain uniform illumination conditions, a dark room and LED lighting were utilized. Following the acquisition of soil surface images, soil samples were collected using a metal cylinder to obtain measurements of soil water content and bulk density. Various features were extracted from the images, including color, texture, and shape features, and used as inputs for both a multiple regression analysis and a deep neural network model. The results show that the deep neural network regression model can predict soil water content and bulk density with root mean squared error of 1.52% and 0.78 kN/m3. The deep neural network model outperformed the multiple regression analysis, achieving a high accuracy for predicting both soil water content and bulk density. These findings suggest that in situ soil surface images, combined with deep learning techniques, can provide a fast and reliable method for predicting important soil properties.

Integrated intelligent neuro computing technique for mixed convective flow and heat transfer in heterogeneous permeability media

This article examines the flow and heat transfer characteristics of variable permeability such as linear, exponential and quadratic permeability functions with Brinkmann–Forchheimer extended Darcy model by utilizing the intelligent computing paradigm via artificial Levenberg–Marquardt back propagated neural networks (ALM – BPNNs). The vertical porous channel walls are considered as electrically conducting and non-conducting. The variable permeability and induced magnetic field effects are significant on flow and heat transfer. To solve the obtained governing equations, finite element method is adopted. It is observed that for the three cases of permeability functions, increase in Hartmann number drops the velocity profile and is accompanied with an increase in the fluid thermal state level. The significance of the Brinkmann number on the fluid temperature in all the cases of permeability indicates a rise in temperature due to increased heat generation. But, the reverse effect is observed when the mixed convection parameter rises, and the boundary layer thickness reduces due to which temperature drops for distinct cases of permeability under consideration. The heat transfer characteristics are analyzed through an artificial neural network and multiple linear regression models. The present work finds its applications in studying packed bed heat exchangers and fixed-bed catalytic reactors.

Experimental study and machine learning algorithms for evaluating the performance of U-shaped ultra-high performance reinforced fiber concrete under static and impact loads

The inherent brittle nature of ultra-high performance concrete (UHPC) increased with strength and is considered its shortcoming in structural applications. This study investigated the impact strength of U-shaped UHPFRC mixtures involving different fiber volume fractions (Vf) under drop-weight impact loads. A U-shaped configuration was introduced in this research to modify the testing procedure of the ACI 544-2R committee for the drop-weight impact test aiming to minimize the scatter result of this testing method. The micro steel fiber (MSF) was incorporated into the UHPFRC mixes at 0–3% Vf (interval of 1%). Moreover, artificial intelligence (AI)-based techniques, including artificial neural network (ANN) and multilinear regression (MLR) model, were employed to train and test the experimental dataset for the prediction of energy absorption capacity (EI2) at the failure crack stage. The results indicated that micro steel has remarkably improved the absorbed energy absorption of the U-shaped UHPFRC mixtures. The U-shaped configuration controls the crack formation. The impact strength at the two cracking stages increased with micro steel fiber content. At the first crack stage, the impact strength of UHPFRC-2 and UHPFRC-3 is 100% and 408% higher than that of UHPFRC-1. The average ductility index value first increased and then decreased with high fiber content. The AI-based technique accurately predicts the energy absorption capacity of a U-shaped UHPFRC mixture at failure crack strength. The ANN and MLR model had an overall R2 value of 0.8231 and 0.9502, respectively.

Innovative soft computing techniques including artificial neural network and nonlinear regression models to predict the compressive strength of environmentally friendly concrete incorporating waste glass powder

Innovative soft computing techniques including artificial neural network and nonlinear regression models to predict the compressive strength of environmentally friendly concrete incorporating waste glass powder

De-noising groundwater level modeling using data decomposition techniques in combination with artificial intelligence (case study Aspas aquifer)

Considering the recent significant drop in the groundwater level (GWL) in most of world regions, the importance of an accurate method to estimate GWL (in order to obtain a better insight into groundwater conditions) has been emphasized by researchers. In this study, artificial neural network (ANN) and support vector regression (SVR) models were initially employed to model the GWL of the Aspas aquifer. Secondly, in order to improve the accuracy of the models, two preprocessing tools, wavelet transform (WT) and complementary ensemble empirical mode decomposition (CEEMD), were combined with former methods which generated four hybrid models including W-ANN, W-SVR, CEEMD-ANN, and CEEMD-SVR. After these methods were implemented, models outcomes were obtained and analyzed. Finally, the results of each model were compared with the unit hydrograph of Aspas aquifer groundwater based on different statistical indexes to assess which modeling technique provides more accurate GWL estimation. The evaluation of the models results indicated that the ANN model outperformed the SVR model. Moreover, it was found that combining these two models with the preprocessing tools WT and CEEMD improved their performances. Coefficient of determination (R2) which indicates model accuracy was increased from 0.927 in the ANN model to 0.938 and 0.998 in the W-ANN and CEEMD-ANN models, respectively. It was also improved from 0.919 in the SVR model to 0.949 and 0.948 in the W-SVR and CEEMD-SVR models, respectively. According to these results, the hybrid CEEMD-ANN model is found to be the most accurate method to predict the GWL in aquifers, especially the Aspas aquifer.

A robust discretization method of factor screening for landslide susceptibility mapping using convolution neural network, random forest, and logistic regression models

ABSTRACT The selection of discretization criteria and interval numbers of landslide-related environmental factors generally fails to quantitatively determine or filter, resulting in uncertainties and limitations in the performance of machine learning (ML) methods for landslide susceptibility mapping (LSM). The aim of this study is to propose a robust discretization criterion (RDC) to quantify and explore the uncertainty and subjectivity of different discretization methods. The RDC consists of two steps: raw classification dataset generation and optimal dataset extraction. To evaluate the robustness of the proposed RDC method, Lushan County of Sichuan Province in China was chosen as the study area to generate the LSM based on three datasets (optimal dataset, original dataset with continuous values, and statistical dataset) using three popular ML methods, namely, convolution neural network, random forest, and logistic regression. The results show that the areas under the receiver operating characteristic curve (AUCs) of the optimal dataset for the abovementioned ML models are 0.963, 0.961, and 0.930 which are higher than those of the original dataset (0.938, 0.947, and 0.900) and statistical dataset (0.948, 0.954, and 0.897). In conclusion, the RDC method can extract the more representative features from environmental factors and outperform the other conventional discretization methods.

A comparison of artificial neural networks and regression modeling techniques for predicting dominant heights of Oriental spruce in a mixed stand

Aim of study: This paper introduces comparative evaluations of artificial neural network models and regression modeling techniques based on some fitting statistics and desirable characteristics for predicting dominant height.
 Area of study: The data of this study were obtained from Oriental spruce (Picea orientalis L.) felled trees in even-aged and mixed Oriental spruce and Scotch pine (Pinus sylvestris L.) stands in the northeast of Türkiye.
 Material and methods: A total of 873 height-age pairs were obtained from Oriental spruce trees in a mixed forest stand. Nonlinear mixed-effects models (NLMEs), autoregressive models (ARM), dummy variable method (DVM), and artificial neural networks (ANNs) were compared to predict dominant height growth.
 Main results: The best predictive model was NLME with a single random parameter (root mean square error, RMSE: 0.68 m). The results showed that NLMEs outperformed ARM (RMSE: 1.09 m), DVM in conjunction with ARM (RMSE: 1.09 m), and ANNs (RMSE: from 1.11 to 2.40 m) in the majority of the cases. Whereas considering variations among observations by random parameter(s) significantly improved predictions of dominant height, considering correlated error terms by autoregressive correlation parameter(s) enhanced slightly the predictions. ANNs generally underperformed compared to NLMEs, ARM, and DVM with ARM.
 Research highlights: All regression techniques fulfilled the desirable characteristics such as sigmoidal pattern, polymorphism, multiple asymptotes, base-age invariance, and inflection point. However, ANNs could not replicate most of these features, excluding the sigmoidal pattern. Accordingly, ANNs seem insufficient to assure biological growth assumptions regarding dominant height growth.