Artificial Neural Network Regression Models Research Articles

Machine learning based methods for ratemaking health care insurance

In insurance, proposing an accurate premium that is adjusted to the insured risk profile allows companies to better manage their portfolios and to be more competitive. Machine learning methods have recently been adopted for various improvements in insurance ratemaking, especially in the automobile industry. These models are specifically used to mine potential data information and to build a predictive model for a variable of interest using explanatory variables. In this paper, we aim to provide a pricing method for ratemaking individual healthcare insurance contracts using machine learning algorithms that are applied to a Tunisian healthcare insurance portfolio. We start with a simple Classification and Regression Tree, and we work toward more advanced methods that are Random Forest, Extreme gradient boosting, Support Vector Regression, and Artificial Neural network regression model. The predictive performance of these non-parametric methods is compared with the standard generalized linear model. Our results showed the applicability of machine learning in the healthcare insurance market and that the XGBoost algorithm outperforms the predictive capacity of the classical generalized linear model.

Harnessing Artificial Neural Networks for Spinal Cord Injury Prognosis.

Background: Prediction of neurorehabilitation outcomes after a Spinal Cord Injury (SCI) is crucial for healthcare resource management and improving prognosis and rehabilitation strategies. Artificial neural networks (ANNs) have emerged as a promising alternative to conventional statistical approaches for identifying complex prognostic factors in SCI patients. Materials: a database of 1256 SCI patients admitted for rehabilitation was analyzed. Clinical and demographic data and SCI characteristics were used to predict functional outcomes using both ANN and linear regression models. The former was structured with input, hidden, and output layers, while the linear regression identified significant variables affecting outcomes. Both approaches aimed to evaluate and compare their accuracy for rehabilitation outcomes measured by the Spinal Cord Independence Measure (SCIM) score. Results: Both ANN and linear regression models identified key predictors of functional outcomes, such as age, injury level, and initial SCIM scores (correlation with actual outcome: R = 0.75 and 0.73, respectively). When also alimented with parameters recorded during hospitalization, the ANN highlighted the importance of these additional factors, like motor completeness and complications during hospitalization, showing an improvement in its accuracy (R = 0.87). Conclusions: ANN seemed to be not widely superior to classical statistics in general, but, taking into account complex and non-linear relationships among variables, emphasized the impact of complications during the hospitalization on recovery, particularly respiratory issues, deep vein thrombosis, and urological complications. These results suggested that the management of complications is crucial for improving functional recovery in SCI patients.

Ann Prediction of Mechanical Properties of GGBFS and Alccofine Based High Strenth Self-Compacting Concrete

Abstract In this study, we use Artificial Neural Networks (ANN) and Multiple Regression Analysis to evaluate the prediction of two crucial self-compacting concrete properties: compressive strength and split tensile strength. It was possible to create four different datasets, each of which had different concrete mix proportions along with their respective ages in days, compressive strengths (MPa), and split tensile strengths (MPa). Separate ANN models and Regression models were trained and tested using these datasets. As a gauge of prediction accuracy, Mean Squared Error (MSE) was used to assess the performance of the models. This study offers insightful information on the application of multiple regression analysis and artificial neural networks to forecast the characteristics of self-compacting concrete using GGBS and Alccofine. Here Alccofine functions as an additive and GGBS acts as a partial substitute for cement at 0 to 60% with a fluctuation of 10%. The outcomes highlight the potential of neural networks as a tool for concrete mix design optimization and quality control since they can capture complex correlations between input variables and concrete strength.

Machine learning-based model to predict severe acute kidney injury after total aortic arch replacement for acute type A aortic dissection

BackgroundSevere acute kidney injury (AKI) after total aortic arch replacement (TAAR) is related to adverse outcomes in patients with acute type A aortic dissection (ATAAD). However, the early prediction of severe AKI remains a challenge. This study aimed to develop a novel model to predict severe AKI after TAAR in ATAAD patients using machine learning algorithms. MethodsA total of 572 ATAAD patients undergoing TAAR were enrolled in this retrospective study, and randomly divided into a training set (70 %) and a validation set (30 %). Lasso regression, support vector machine-recursive feature elimination and random forest algorithms were used to screen indicators for severe AKI (defined as AKI stage III) in the training set, respectively. Then the intersection indicators were selected to construct models through artificial neural network (ANN) and logistic regression. The AUC-ROC curve was employed to ascertain the prediction efficacy of the ANN and logistic regression models. ResultsThe incidence of severe AKI after TAAR was 22.9 % among ATAAD patients. The intersection predictors identified by different machine learning algorithms were baseline serum creatinine and ICU admission variables, including serum cystatin C, procalcitonin, aspartate transaminase, platelet, lactic dehydrogenase, urine N-acetyl-β-d-glucosidase and Acute Physiology and Chronic Health Evaluation II score. The ANN model showed a higher AUC-ROC than logistic regression (0.938 vs 0.908, p < 0.05). Furthermore, the ANN model could predict 89.1 % of severe AKI cases beforehand. In the validation set, the superior performance of the ANN model was further confirmed in terms of discrimination ability (AUC = 0.916), calibration curve analysis and decision curve analysis. ConclusionThis study developed a novel and reliable clinical prediction model for severe AKI after TAAR in ATAAD patients using machine learning algorithms. Importantly, the ANN model showed a higher predictive ability for severe AKI than logistic regression.

Real‐time monitoring and ageing detection algorithm design with application on SiC‐based automotive power drive system

AbstractThe article describes an innovative methodology for the design and experimental validation of monitoring and anomaly detection algorithms, with a particular focus on the aging phenomenon, linked to the anomalous modification of the , in devices switching in power electronic systems integrated into modern high‐performance electrified vehicles. The case study concerns an electric drive for fully electrified vehicles, in which a three‐phase axial flux synchronous motor integrated into a wheel motor (Elaphe) is used and in which a high‐efficiency three‐phase inverter, designed with SiC technology (silicon carbide). The article proposes the design and validation of the innovative aging monitoring and detection system, in four consecutive phases. The first phase involves the creation of a real‐time model of electric drive, validated through experimental data extrapolated directly during a WLTP (Worldwide Harmonized Light Vehicle Test Procedure) test. The second phase consists of the creation of a virtual dataset representative of the aging phenomenon, via an anomaly injection procedure, emulating this phenomenon with a scaling factor (depending on the value of the ) on the current phase of the motor, relating to the inverter branch whose SiC device is affected. The third phase concerns the design of an estimator of the , based on an ANN (Artificial Neural Network) regression model, and involves a data manipulation phase with features extraction and reduction techniques. The fourth and final phase, involves the experimental validation of the method, through PIL (Processor‐In‐the‐Loop) tests, integrating the monitoring algorithm (consisting of a real‐time model and AI‐based regression model) on the NXPs32k144 embedded platform (based on Cortex‐M4), making the algorithm interact with the electric drive model on which anomaly injection is applied.

Valence-isomer selective cycloaddition reaction of cycloheptatrienes-norcaradienes

The rapid and precise creation of complex molecules while controlling multiple selectivities is the principal objective in synthetic chemistry. Combining data science and organic synthesis to achieve this goal is an emerging trend, but few examples of successful reaction designs are reported. We develop an artificial neural network regression model using bond orbital data to predict chemical reactivities. Actual experimental verification confirms cycloheptatriene-selective [6 + 2]-cycloaddition utilizing nitroso compounds and norcaradiene-selective [4 + 2]-cycloaddition reactions employing benzynes. Additionally, a one-pot asymmetric synthesis is achieved by telescoping the enantioselective dearomatization of non-activated benzenes and cycloadditions. Computational studies provide a rational explanation for the seemingly anomalous occurrence of thermally prohibited suprafacial [6 + 2]-cycloaddition without photoirradiation.

Risk factor analysis and prediction model construction for surgical patients with venous thromboembolism: A prospective study

Abstract Objective: Patients undergoing surgery are at high risk of developing venous thromboembolism (VTE). This study aimed to determine the predictive value of risk factors for VTE in surgical patients and to develop a prediction model by integrating independent predictors. Methods: A total of 1111 patients who underwent surgery at clinical departments in a tertiary general hospital were recruited between May and July 2021. Clinical data, including patient-related, surgery-related, and laboratory parameters, were extracted from the hospital information system and electronic medical records. A VTE prediction model incorporating ten risk variables was constructed using artificial neural networks (ANNs). Results: Ten independent factors (X1: age, X2: alcohol consumption, X3: hypertension, X4: bleeding, X5: blood transfusions, X6: general anesthesia, X7: intrathecal anesthesia, X8: D-dimer, X9: C-reactive protein, and X10: lymphocyte percentage) were identified as associated with an increased risk of VTE. Ten-fold cross-validation results showed that the ANN model was capable of predicting VTE in surgical patients, with an area under the curve (AUC) of 0.89, a Brier score of 0.01, an accuracy of 0.96, and a F1 score of 0.92. The ANN model slightly outperformed the logistic regression model and the Caprini model, but a DeLong test showed that the statistical difference in the AUCs of the ANN and logistic regression models was insignificant (P&gt;0.05). Conclusion: Ten statistical indicators relevant to VTE risk prediction for surgical patients were identified, and ANN and logistic regression both showed promising results as decision-supporting tools for VTE prediction.

Modelling the Yield and Estimating the Energy Properties of Miscanthus x Giganteus in Different Harvest Periods

This research aims to use artificial neural networks (ANNs) to estimate the yield and energy characteristics of Miscanthus x giganteus (MxG), considering factors such as year of cultivation, location, and harvest time. In the study, which was conducted over three years in two different geographical areas, ANN regression models were used to estimate the lower heating value (LHV) and yield of MxG. The models showed high predictive accuracy, achieving R2 values of 0.85 for LHV and 0.95 for yield, with corresponding RMSEs of 0.13 and 2.22. A significant correlation affecting yield was found between plant height and number of shoots. In addition, a sensitivity analysis of the ANN models showed the influence of both categorical and continuous input variables on the predictions. These results highlight the role of MxG as a sustainable biomass energy source and provide insights for optimizing biomass production, influencing energy policy, and contributing to advances in renewable energy and global energy sustainability efforts.

Machine learning assisted strip waveguide Bragg gratings design for refractive index-based biosensing applications

Waveguide Bragg grating (WBG) biosensors have attracted significant interest because of their high sensitivity, immunity to electromagnetic signals, and capability for lab-on-chip applications. Designing a WBG biosensor requires solving Maxwell’s equations for the optical waveguide geometries numerically. This process is resource-intensive and hinders quick optimization of the waveguides for sensing. In this work, we explore the design of a silicon strip WBG structure for biosensing using machine learning. We develop an artificial neural network (ANN) classification model for mode classification and a regression model for predicting the effective refractive index (neff) for a given strip waveguide. We utilize results from finite element method and Transfer Matrix method for strip WBG structures to train a multi-output ANN regression model for quickly predicting sensing parameters. The mode classification model provides an accuracy of over 99%, and the regression model predicts the neff with a mean absolute error (MAE) of 0.5%. The multi-output regression model predicts the sensitivity with 0.5% MAE and quality factor and maximum reflectivity with less than 3% MAE. ANN model training requires a few minutes of computational time and reduces the dependency on computationally expensive resources for optimization of the sensor design, thereby fast-tracking the biosensor design cycle.

Employing artificial neural networks for the assessment of the aqueous solubility of drug –like substances

This research has advanced Quantitative Structure-Property Relationship (QSPR) models for predicting the aqueous solubility of drug-like substances. By integrating multivariate regression and neural network techniques, the study utilized the backward algorithm to strategically select 2D and 3D molecular descriptors, resulting in the development of an optimal QSPRMLR model with k = 23. The artificial neural network regression model(QSPRANN), derived from selected descriptors of the multivariable linear regression model(QSPRMLR), demonstrated enhanced predictive capabilities for logS values in both validationand prediction groups, yielding SE values of 0.786 and 0.808, respectively. The QSPRANNsignificantly improved the overall predictability of the multivariate regression model. Statistical assessments of the QSPRANN model revealed SE = 0.699, R2train = 0.918, and Q2v = 0.878. The predicted logS values from the QSPRANN model align well with experimental data,confirming the reliability and accuracy of the developed model.

Prediction of Acceleration Amplification Ratio of Rocking Foundations Using Machine Learning and Deep Learning Models

Experimental results reveal that rocking shallow foundations reduce earthquake-induced force and flexural displacement demands transmitted to structures and can be used as an effective geotechnical seismic isolation mechanism. This paper presents data-driven predictive models for maximum acceleration transmitted to structures founded on rocking shallow foundations during earthquake loading. Results from base-shaking experiments on rocking foundations have been utilized for the development of artificial neural network regression (ANN), k-nearest neighbors regression, support vector regression, random forest regression, adaptive boosting regression, and gradient boosting regression models. Acceleration amplification ratio, defined as the maximum acceleration at the center of gravity of a structure divided by the peak ground acceleration of the earthquake, is considered as the prediction parameter. For five out of six models developed in this study, the overall mean absolute percentage error in predictions in repeated k-fold cross validation tests vary between 0.128 and 0.145, with the ANN model being the most accurate and most consistent. The cross validation mean absolute error in predictions of all six models vary between 0.08 and 0.1, indicating that the maximum acceleration of structures supported by rocking foundations can be predicted within an average error limit of 8% to 10% of the peak ground acceleration of the earthquake.

Exploring the predictive power of ANN and traditional regression models in real estate pricing: evidence from Prishtina

PurposeThe purpose of the study is to examine the efficiency of linear, nonlinear and artificial neural networks (ANNs), in predicting property prices.Design/methodology/approachThe present study uses a dataset of 1,468 real estate transactions from 2020 to 2022, obtained from the Department of Property Taxes of Republic of Kosovo. Beginning with a fundamental linear regression model, the study tackles the question of overlooked nonlinearity, employing a similar strategy like Peterson and Flanagan (2009) and McCluskey et al. (2012), whereby ANN's predictions are incorporated as an additional regressor within the ordinary least squares (OLS) model.FindingsThe research findings underscore the superior fit of semi-log and double-log models over the OLS model, while the ANN model shows moderate performance, contrary to the conventional conviction of ANN's superior predictive power. This is notably divergent from the prevailing belief about ANN's superior predictive power, shedding light on the potential overestimation of ANN's efficacy.Practical implicationsThe study accentuates the importance of embracing diverse models in property price prediction, debunking the notion of the ubiquitous applicability of ANN models. The research outcomes carry substantial ramifications for both scholars and professionals engaged in property valuation.Originality/valueDistinctively, this research pioneers the comparative analysis of diverse models, including ANN, in the setting of a developing country's capital, hence providing a fresh perspective to their effectiveness in property price prediction.

Applying Multivariate Analysis and Machine Learning Approaches to Evaluating Groundwater Quality on the Kairouan Plain, Tunisia

In the Zeroud basin, a diverse array of methodologies were employed to assess, simulate, and predict the quality of groundwater intended for irrigation. These methodologies included the irrigation water quality indices (IWQIs); intricate statistical analysis involving multiple variables, supported with GIS techniques; an artificial neural network (ANN) model; and an XGBoost regression model. Extensive physicochemical examinations were performed on groundwater samples to elucidate their compositional attributes. The results showed that the abundance order of ions was Na+ &gt; Ca2+ &gt; Mg2+ &gt; K+ and SO42− &gt; HCO3− &gt; Cl−. The groundwater facies reflected Ca-Mg-SO4, Na-Cl, and mixed Ca-Mg-Cl/SO4 water types. A cluster analysis (CA) and principal component analysis (PCA), along with ionic ratios, detected three different water characteristics. The mechanisms controlling water chemistry revealed water–rock interaction, dolomite dissolution, evaporation, and ion exchange. The assessment of groundwater quality for agriculture with respect IWQIs, such as the irrigation water quality index (IWQI), sodium adsorption ratio (SAR), sodium percentage (Na%), soluble sodium percentage (SSP), potential salinity (PS), and residual sodium carbonate (RSC), revealed that the domination of the water samples was valuable for agriculture. However, the IWQI and PS fell between high-to-severe restrictions and injurious-to-unsatisfactory. The ANN and XGBoost regression models showed robust results for predicting IWQIs. For example, ANN-HyC-9 emerged as the most precise forecasting framework according to its outcomes, as it showcased the most robust link between prime attributes and IWQI. The nine attributes of this model hold immense significance in IWQI prediction. The R2 values for its training and testing data stood at 0.999 (RMSE = 0.375) and 0.823 (RMSE = 3.168), respectively. These findings indicate that XGB-HyC-3 emerged as the most accurate forecasting model, displaying a stronger connection between IWQI and its exceptional characteristics. When predicting IWQI, approximately three of the model’s attributes played a pivotal role. Notably, the model yielded R2 values of 0.999 (RMSE = 0.001) and 0.913 (RMSE = 2.217) for the training and testing datasets, respectively. Overall, these results offer significant details for decision-makers in managing water quality and can support the long-term use of water resources.

Ballistic impact: predicting penetration depth in ultra-high-performance concrete targets

This paper presents the application of an artificial neural network (ANN) model for predicting the penetration depth under projectile impact in ultra-high-performance concrete (UHPC) targets containing steel fibres. Despite the availability of a large number of existing empirical models, the prediction of penetration depth remained inconclusive, partly due to the phenomenon's complexity and partly due to the limitation of statistical regression. From the results of this study, it is evident that the ANN model is capable of predicting the penetration depth of UHPC more accurately than other machine-learning (ML) models (linear regression, decision tree regression and random forest regression) and empirical formulae. The ANN model achieved a lower root mean square error (RMSE) of 11.68 compared to the other ML models (RMSE: 16.66–19.74) and empirical equations (RMSE: 25.17–53.42), when applied to the test data set. The velocity, impact energy, diameter of the projectile and thickness of the UHPC targets are the most significant parameters (P-value &lt;5%) for predicting the penetration depth using ANN and multiple linear regression models.

Identification of critical SARS-CoV-2 amino acids associated with COVID-19 hospitalization rate using machine learning and statistical modeling: An observational study in the United States.

The COVID-19 pandemic has put many medical systems on the verge of collapse in the last two years. Virus mutation was one of the important factors affecting the COVID-19 infection severity and hospitalizations. Although over ten thousand SARS-CoV-2 mutations being reported since the beginning of the COVID-19 pandemic, only a small percentage of mutations are likely to affect the virus phenotype and change its severity. Finding out which amino acids have the greatest impact on COVID-19 hospitalization rate is an important research question. This observational study used the COVID-19 case hospitalization ratio (CHR) to represent the virus severity related with hospitalization. The database is based on the daily state-level epidemiological and genomic sequential data in the United States from the Alpha wave to the first Omicron wave. The critical amino acids that mostly affected the CHR were determined by using four types of models including extreme gradient boosting decision trees (XGBoost), artificial neural networks (ANNs), logistic regression and Lasso regression models. The XGBoost, ANN, logistic regression, and Lasso regression models all produce excellent results (mean square error for all state-level models does not exceed 0.0008 using the testing dataset). Based on the rank of importance of all covariates, the critical amino acids most affecting the CHR were identified, including T19, L24, P25, P26, A27, A67, H69, V70, T95, G142, V143, Y145, E156, F157, N211, L212, V213, R214, D215, G339, R346, S373, L452, S477, T478, E484, N501, A570, P681, and T716. This study identified critical amino acids that are most likely to affect the hospitalization rate, allowing public health workers to monitor these highly risky amino acids and raise an alarm immediately when more severe mutations occur. Furthermore, the methodology and results may be extended to other regions.

Assessing the color status and daylight chromaticity of road signs through machine learning approaches

The color of road signs is a critical aspect of road safety, as it helps drivers quickly and accurately identify and respond to these signs. Properly colored road signs improve visibility during the day and make it easier for drivers to make informed decisions while driving. In order to ensure the safety and efficiency of road traffic, it is essential to maintain the appropriate color level of road signs.The objective of this study was to analyze the color status and daylight chromaticity of in-use road signs using supervised machine learning models, and to explore the correlation between road sign's age and daylight chromaticity. Three algorithms were employed: Random Forest (RF), Support Vector Machine (SVM), and Artificial Neural Network (ANN). The data used in this study was collected from road signs that were in-use on roads in Sweden.The study employed classification models to assess the color status (accepted or rejected) of the road signs based on minimum acceptable color levels according to standards, and regression models to predict the daylight chromaticity values. The correlation between road sign's age and daylight chromaticity was explored through regression analysis. Daylight chromaticity describes the color quality of road signs in daylight, that is expressed in terms of X and Y chromaticity coordinates.The study revealed a linear relationship between the road sign's age and daylight chromaticity for blue, green, red, and white sheeting, but not for yellow. The lifespan of red signs was estimated to be around 12 years, much shorter than the estimated lifespans of yellow, green, blue, and white sheeting, which are 35, 42, 45, and 75 years, respectively.The supervised machine learning models successfully assessed the color status of the road signs and predicted the daylight chromaticity values using the three algorithms. The results of this study showed that the ANN classification and ANN regression models achieved high accuracy of 81% and R2 of 97%, respectively. The RF and SVM models also performed well, with accuracy values of 74% and 79% and R2 ranging from 59% to 92%. The findings demonstrate the potential of machine learning to effectively predict the status and daylight chromaticity of road signs and their impact on road safety in the Swedish context.

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.

Effects of ‘social’ environmental stability and entrepreneurial parameters in assessing the relationship among entrepreneurship, green innovation, and strategic orientation: an entrepreneur development of documentary

Sustainable entrepreneurship is essential for solving the issue of economic and social stability. This paper is about the determinants of entrepreneurship among green innovation and strategic innovation. The research studies the communication and R&D variables that impact entrepreneurship development. The literature shows that green innovation is vital in developing strategic orientation. Entrepreneurship must work on R&D to gain economic and social stability. Entrepreneurship development is a dependent variable, and communication and R&D are independent variables. This paper evaluates the performance of Multiple Linear Regression (MLR) and the Artificial Neural Network Regression (ANN) approach for estimating entrepreneurship development (ED) in China from 2000 to 2021. For this work, nine independent variables related to entrepreneurial development were selected, which include CO2 emissions (CO2), GDP growth (GDP), Total greenhouse gas emissions (GHG), Computer, communications, and other services (CCS), Nitrous oxide emissions (NO2), Personal remittances, received (PRR), Research and development expenditure (R&D), New businesses registered (NBR), and patents in environment-related technologies (PERT). Data were gathered from the World Bank (WB) official data bank portal. The highest correlation (−0.85 and −0.71) with ED was observed with GHG, followed by a strong correlation with R & D. The MLR model generated 0.78 R2 (error = 0.26) whereas the ANN model produces 0.83 R2 (error = 0.01). The results of the ANN showed that the association among parameters is strong. Ten input variables and 15 hidden neurons were used to estimate the dependent variable. The process took 87 steps, and the optimum result had a Sum of Square Error of 0.001.The top 5 essential variables are NO, NBR, GHG, R&D, and CO2. The least essential variables are PPR and CCS, respectively, with the lowest variable importance (VI) score. This work is unique in the sense that it covers the research gap with new advanced techniques (ANN) applied to time series data. We believe the idea that we developed is a novel approach for social scientists, entrepreneurial development, social stability, and a sustainable environment. The findings of this research can benefit entrepreneurs in many different ways. By understanding the effects of social environmental stability on their success, entrepreneurs can better tailor their strategies to ensure they get the best possible results.

Delamination identification in sandwich composite structures using machine learning techniques

Recent advances in machine learning have enabled powerful strategies for autonomous data-driven damage detection and identification in structural systems. This work proposes a novel method for 3D delamination identification in sandwich composite structures. Sandwich structures are prone to damage that is difficult to detect, necessitating efficient and complex inspection techniques. In this context, automated structural health monitoring techniques are of great impact and importance in engineering. Such techniques use data obtained by sensors to identify damage to the structure. In this paper, a complete methodology for damage (delamination) identification in sandwich composite structures using machine learning is proposed. The damage was parameterized in two different ways: as parametrized two- and three-dimensional ellipses, and it was considered in three different groups: the core, interface, and skin. The modal data, obtained by the finite element method, was used to train several machine learning models in order to classify the location of the damage. In addition, modal dataset was also used to train artificial neural network regression models for damage localization and sizing. Both classification and regression strategies showed substantial results, and the models proved to be robust enough to identify a wide variety of damages. Results showed that classification models correctly identified the damage on the composite skin and incorrectly identified the interface core-skin damage. The regression model proved to be reliable in identifying an approximate location with an average accuracy of 85%. However, damage sizing is still a challenge to predict based only on modal datasets.

Weight Prediction of Landlly Pigs from Morphometric Traits in Different Age Classes Using ANN and Non-Linear Regression Models

The present study was undertaken to identify the best estimator(s) of body weight based on various linear morphometric measures in Landlly pigs using artificial neural network (ANN) and non-linear regression models at three life stages (4th, 6th and 8th week). Twenty-four different linear morphometric measurements were taken on 279 piglets individually at all the stages and their correlations with body weight were elucidated. The traits with high correlation (≥0.8) with body weight were selected at different stages. The selected traits were categorized into 31 different combinations (single, two, three, four and five) and subjected to ANN modelling for determining the best combination of body weight predictors at each stage. The model with highest R2 and lowest MSE was selected as best fit for a particular trait. Results revealed that the combination of heart girth (HG), body length (BL) and paunch girth (PG) was most efficient for predicting body weight of piglets at the 4th week (R2 = 0.8697, MSE = 0.4419). The combination of neck circumference (NCR), height at back (HB), BL and HG effectively predicted body weight at 6 (R2 = 0.8528, MSE = 0.8719) and 8 (R2 = 0.9139, MSE = 1.2713) weeks. The two-trait combination of BL and HG exhibited notably high correlation with body weight at all stages and hence was used to develop a separate ANN model which resulted into better body weight prediction ability (R2 = 0.9131, MSE = 1.004) as compared to age-dependent models. The results of ANN models were comparable to non-linear regression models at all the stages.