Linear or non-linear multivariate calibration models? That is the question

Aim: This study set out to determine how well AI approaches like Artificial Neural Networks (ANNs) and Deep Learning Neural Networks (DLNNs) might be used to forecast rainfall (DNN). These methods of weather prediction were tested and ranked in terms of their efficiency. Substances and Techniques: Group 1 uses a Deep Learning Neural Network (DNN) for analysis, whereas Group 2 uses an Artificial Neural Network (ANN) with a pre-test power (G-power) of 80% and an alpha error rate (Err) of 0.05 for sample test analysis. In all, 20 samples were identified; 10 from each category. According to the outcomes of the MATLAB simulations, the accuracy of the Deep Learning Neural Network (DNN) is 92.59%, while that of the Artificial Neural Network (ANN) is 95.68%. By means of SPSS statistical program, we find that the achieved accuracy ratio is 0.034 (p 0.05). This study concludes that the unique Artificial Neural Network (ANN) algorithm outperforms the Deep Learning Neural Network (DNN) method in predicting rainfall in the context of cutting-edge weather forecasting.

<p>Fire is an important element of the ecosystems, nevertheless, high severity fires can have negative impacts on the ecosystems as consequence of the high temperatures reaches. High temperatures normally have detrimental impacts on soil properties. The objective of this work was to determine the relationship of spectral reflectance and soil pH, electrical conductivity (EC), carbonates (CaCO<sub>3</sub>) and total carbon (TC) content after a medium to high severity wildfire occurred in Croatia using linear and nonlinear calibration models.</p><p>Soils were sampled 2 days after a medium to high severity wildfire in Zadar County, Croatia. A total of 120 soil samples (0-5 cm) were collected from three different treatments (n= 40 per treatment): control (C), mean severity (MS), high severity (HS). Soil pH, EC, CaCO<sub>3</sub> and TC content were determined using standard laboratory methods.  Soil spectral measurements were carried out using a portable spectroradiometer (20 per treatment, 60 in total). Linear statistical model - partial least squares regression (PLSR) and non-linear - artificial neural network (ANN) were generated to estimate changes in soil pH, EC, CaCO<sub>3</sub> and TC content based on the original spectral reflectance and its first derivative in form of principal components (PC). One-way ANOVA revealed pH values were significantly different in all three treatments. EC, CaCO<sub>3</sub> and TC were significantly higher in HS plots compared with the other treatments.</p><p>Different wildfire severity indicated very collinear soil spectral response, but with certain variations of reflectance intensity. Control samples showed a higher reflectance than MS and HS samples. This is attributed to the low pH and TC content. Low reflectance of MS and HS samples could be explained by their increased pH and TC values. Soil pH was the only parameter that showed a high R<sup>2</sup> and low root mean squared error (RMSE) after Savitzky Golay smoothing and the first derivation. In PLSR model, strong to very strong correlation and low RMSE were obtained. ANN model also showed a high R<sup>2</sup> and lower RMSE for all properties except pH. Both models showed satisfactory results for prediction of the studied soil properties. ANN model predicted EC, CaCO<sub>3</sub>, and TC better, while PLSR proved to be a better model for pH prediction.</p><p><strong>Key words:</strong> soil reflectance, fire severity, principal components, partial least squares regression, neural networks</p><p>Acknowledgements</p><p>This work was supported by Croatian Science Foundation through the project "Soil erosion and degradation in Croatia" (UIP-2017-05-7834) (SEDCRO).</p>

The selection of calibration method is one of the most important factors affecting the measurement accuracy with near infrared spectroscopy. In this research, the performance of two general calibration methods, namely, linear partial least squares regression and nonlinear back propagation artificial neural networks for firmness predicting of ‘Mazafati’ date fruit was investigated. A total of 175 date samples harvested during fruit ripening were selected as the data set. Optical scanning was performed with a fiber optic near infrared spectrometer with a range of 900–1700 nm. The inputs of back propagation artificial neural networks were the first five principal components resulting from the principal component analysis, namely, principal component analysis artificial neural network modeling or the first six latent variables obtained from partial least squares regression, namely, latent variable artificial neural network modeling. Both the leave-one-out cross validation and test set validation showed the priority of latent variable artificial neural network model with respect to partial least squares and also principal component analysis artificial neural network models. The best latent variable artificial neural network model could predict the firmness of Mazafati date fruits with Rp 2 of 0.90, RMSEP of 1.30 N, and SDRp of 3.28. The results also recommended the adoption of nonlinear latent variable artificial neural network modeling for applying light scattering properties of fruits in order to predict indirect property of firmness with near infrared spectroscopy.

The odor emitted from a wastewater treatment plant (WWTP) is an important environmental problem. An estimation of odor emission rate is difficult to detect and quantify. To address this, various approaches including the development of emission factors and measurement using a closed chamber have been employed. However, the evaluation of odor emission involves huge manpower, time, and cost. An artificial neural network (ANN) is recognized as an efficient method to find correlations between nonlinear data and prediction of future data based on these correlations. Due to its usefulness, ANN is used to solve complicated problems in various disciplines of sciences and engineering. In this study, a method to predict the odor concentration in a WWTP using ANN was developed. The odor concentration emitted from a WWTP was predicted by the ANN based on water quality data such as biological oxygen demand, dissolved oxygen, and pH. The water quality and odor concentration data from the WWTP were measured seasonally in spring, summer, and autumn and these were used as input variations to the ANN model. The odor predicted by the ANN model was compared with the measured data and the prediction accuracy was estimated. Suggestions for improving prediction accuracy are presented.

With the development of internet of things technology community time-series data modeling is a new hot topic of scholars all over the world, modeling ability with its artificial neural network under the environment of Internet of things technology community complex nonlinear time series data system and the application ability good data, application and network in nonlinear time series data forecasting technology community environment get massive practice. The subject of the nonlinear network technology community time-series data prediction several bottlenecks exist in neural network analysis, artificial neural network is presented from the perspective of nonlinearity IOT application of forecast technology community time-series data to optimize the prediction bottleneck neural network environment on the based. Therefore, through the simulation community data of artificial neural network show that the practicality and efficiency of the algorithm.

System modeling is a scientific method that combines theory with experimental studies and has an important place in research activities. With the system model, the data to be obtained through real tests and experiments are provided more economically in terms of cost and the critical points of the system are provided with time savings. Some system models are very difficult to obtain using only analytical equations and methods. At this point, artificial neural networks are an alternative way to model complex, uncertain, nonlinear systems. Artificial neural network is an artificial intelligence system that takes the human brain as an example, learns from existing examples, can produce results with noisy, incomplete, non-linear data, and can make predictions and generalizations with high speed and accuracy after learning once. In this study, RT 512 liquid level control system produced by GUNT Hamburg, an experimental process control system for educational purposes, was modeled with an artificial neural network. In order to create the dynamic model, an input-output data set was created by operating the system in open-loop mode. In this set, the level change seen in the liquid level tube against the given control sign has been taken into account. For this process, a certain number of output data was obtained for a certain number of input data by using computer, Arduino, MCP4725 DAC, current/voltage, voltage/current converters. In the developed ANN model, the relationship between the regression curves and the model output and the test data taken from the system was observed and high accuracy was obtained.

Chapter 9 - Artificial neural network and its applications: Unraveling the efficiency for hydrogen production

Recently, Artificial Neural Network (ANN) has been developed as an alternative to classical statistical methods in animal production. The methods can do classification or prediction by analyzing the information in the data set with the help of the neural network without requiring any preconditions (for example, distribution of data, non-linear data, highly correlated variables, etc.). In this context, we hypothesized that ANN, which is not only used in large and complex data sets but also estimates better in small data sets compared to classical statistical methods. The ability of ANN of Bayesian Regularization (BR-ANN) or Levenberg-Marquardt (LM-ANN) algorithms and Multiple Regression (MR) model to predict meat quality traits were assessed in a comparative study. The multilayer ANN algorithms obtained prediction data of meat quality measurements from pre-slaughter information using 1, 2, 4, 6 and 8 neurons in the hidden layer applied 10 times for each model. The performance of the methods was assessed according to the coefficient of determination (R2) criteria, root mean squared error (RMSE) and residual prediction deviation (RPD). The comparison of the findings of BR-ANN and ML-ANN algorithms showed a similar ability to predict meat quality traits (error of prediction and R2 values between 0.32-2.72 and 0.19-0.49, respectively). However, MR model predictions had lower performance than ANN algorithms, resulting in a wider error of prediction interval (0.4-3.44) and low R2 (0.16-0.44). The RPD for meat quality traits was fair for BR-ANN and LM-ANN but was poor for MR. Based on our results, the ANN algorithms produced more reasonable prediction values than the MR model. ANN algorithms can be used as an acceptable alternative method for simple physical measurements of meat quality. ANN algorithms can be used reliably in small data sets. Key words: Bayesian networks, cattle, meat science

The fact that bitumens behave as non-Newtonian fluids results in non-linear relationships between their near-infrared (NIR) spectra and the physico-chemical properties that define their consistency (viz. penetration and viscosity). Determining such properties using linear calibration techniques [e.g. partial least-squares regression (PLSR)] entails the previous transformation of the original variables by use of non-linear functions and employing the transformed variables to construct the models. Other properties of bitumens such as density and composition exhibit linear relationships with their NIR spectra. Artificial neural networks (ANNs) enable modelling of systems with a non-linear property-spectrum relationship; also, they allow one to determine several properties of a sample with a single model, so they are effective alternatives to linear calibration methods. In this work, the ability of ANNs simultaneously to determine both linear and non-linear parameters for bitumens without the need previously to transform the original variables was assessed. Based on the results, ANNs allow the simultaneous determination of several linear and non-linear physical properties typical of bitumens.

According to current trends in the development of electricity markets, distribution and transmission system operators must purchase electricity to cover their losses in the networks and the wholesale electricity market. By reducing the error in forecasting losses by 1%, this will reduce the cost of compensating for imbalances in the amount of 131.2 million per year, which will reduce tariffs for distribution and transmission of electricity. The study describes a comparative analysis of different architectures of artificial neural networks of deep learning for short-term forecasting of nodal electrical load. A comparison of the results of forecasting artificial neural network architectures and classical forecasting methods was performed. Data from the Northwestern region of the United States and Turkish power system were used. The results of the study show that neural networks of deep learning are superior to classical methods.

Earlier methods concentrated on constructing a single CNN model, while the ensemble learning paradigm has received less attention. Based on our survey results, we chose to utilize an ensemble model comprising multiple CNN models to predict pneumonia diagnosis from x-rays. We proposed employing hybrid model algorithms to predict early pneumonitis in health care. Hybrid model algorithms are increasingly used in healthcare due to their ability to combine the strengths of multiple algorithms to achieve better performance than a single algorithm. In the case of early detection of pneumonitis, this is particularly important, as it is a serious condition that requires prompt diagnosis and treatment to prevent further complications. Artificial neural networks are well suited to process complex and non-linear data, which is important in healthcare where patient data can be highly heterogeneous. Decision trees can identify the most important features for predicting pneumonitis and can be used to generate rules for clinical decision-making. Support vector machines can be used for classification tasks, which is important in identifying patients who are at high risk for pneumonitis. By combining these algorithms in a hybrid model, it may be possible to achieve better performance than using a single algorithm alone. For example, an artificial neural network could be used to pre-process the data and identify the most important features, which are then fed into a decision tree for rule generation. The resulting rules could then be used to classify patients using a support vector machine.

Predicting corporate bankruptcy accurately is crucial for reducing financial risks and protecting stakeholders. Logistic Regression (LR) has long been valued for its interpretability, while Artificial Neural Networks (ANNs) have gained prominence for capturing complex, nonlinear financial relationships. This study synthesizes comparative evidence to assess their relative strengths and limitations. A systematic review of peer-reviewed studies published between 2018 and 2024 was conducted using Scopus, Web of Science, and IEEE Xplore. Studies were included if they directly compared ANNs and LR in bankruptcy prediction, reporting performance metrics such as accuracy, recall, and handling of class imbalance. Attention was also given to the application of visualization and interpretability tools. The findings indicate that ANNs generally outperform LR in predictive performance, with accuracy reaching up to 96% and recall above 95%, especially when advanced preprocessing techniques such as the Synthetic Minority Oversampling Technique (SMOTE) and feature selection are applied. ANNs are particularly effective in modelling nonlinear and high-dimensional data but face challenges related to overfitting, computational demand, and lack of transparency. LR, though often less accurate, remains robust for its statistical rigor, ease of implementation, and interpretability. In certain contexts, LR has also demonstrated higher precision, showing that performance may vary depending on the characteristics and quality of the underlying data. Overall, ANNs are preferable in complex, data-rich environments, while LR is more suitable where transparency and regulatory compliance are critical. Future research should explore hybrid models that integrate ANN’s predictive power with LR’s explainability, supported by explainable AI and real-time decision-support tools.

Early plant disease diagnosis through handheld UV-Vis transmittance spectrometer with DD-SIMCA one-class classification and MCR-ALS bilinear decomposition

An artificial neural network for mixed frequency data

Using scientific knowledge, weather forecasters can predict what the atmosphere will be like in a particular place. It predicts snow, cloud cover, rain, temperature and wind speed. Since weather predictions consist of multidimensional and nonlinear data, they are one of the world's most challenging problems. Various machine learning algorithms and methods have been used in data mining for weather prediction, including Support Vector Machines, supervised and unsupervised machine learning algorithms, artificial neural networks, FPGrowth Algorithms, Hadoop with Map Reduce, K-medoids, and Naive Bayes. This survey briefly explains the methods used to build weather forecasting models to assist researchers in choosing the appropriate method for their model.