Regression Algorithm Research Articles

Bi-parametric MRI-based quantification radiomics model for the noninvasive prediction of histopathology and biochemical recurrence after prostate cancer surgery: a multicenter study.

To develop and evaluate the performance of a noninvasive radiomics combined model based on preoperative bi-parametric MRI to assess biochemical recurrence (BCR) risk factors and to predict biochemical recurrence free survival in PCa patients. Pretreatment bp-MRI and clinicopathology data of 666 (discovery cohort, 545; test cohort, 121) PCa patients from four centers between January 2015 to March 2023 were retrospectively included. To predict BCR, extracapsular extension (ECE), pelvic lymph node metastasis (PLNM), and Gleason Grade group (GG), the pred-BCR, pred-ECE, pred-PLNM, and pred-GG models were developed, respectively. Subsequently, a logistic regression algorithm was used to combine one or more radiomics models and clinicopathology variables into radiomics-clinicopathology combined models (M1, M2) and radiomics-clinical combined model without pathology results (M3) for predicting BCR. In the test cohort, the AUCs for the pred-BCR, pred-ECE, pred-PLNM, and pred-GG models were 0.841, 0.764, 0.896, and 0.698. Of the three combined models, M3 has the best prediction performance with an AUC of 0.884, M2 is the following with an AUC of 0.863, and M1 has the lowest performance with an AUC of 0.838 (95% CI 0.750-0.925) in the test cohort. Delong's test showed that the M3 was significantly higher (M1 vs. M3, p = 0.028; M2 vs. M3, p = 0.044). The combined model developed in this study, which is not dependent on pathologic biopsies, can noninvasively predict postoperative histopathology and BCR after PCa, therefore may provide decision support for follow-up and treatment strategies for patients in the postoperative period.

RRMSE-enhanced weighted voting regressor for improved ensemble regression.

Ensemble regression methods are widely used to improve prediction accuracy by combining multiple regression models, especially when dealing with continuous numerical targets. However, most ensemble voting regressors use equal weights for each base model's predictions, which can limit their effectiveness, particularly when there is no specific domain knowledge to guide the weighting. This uniform weighting approach doesn't consider that some models may perform better than others on different datasets, leaving room for improvement in optimizing ensemble performance. To overcome this limitation, we propose the RRMSE (Relative Root Mean Square Error) Voting Regressor, a new ensemble regression technique that assigns weights to each base model based on their relative error rates. By using an RRMSE-based weighting function, our method gives more importance to models that demonstrate higher accuracy, thereby enhancing the overall prediction quality. We tested the RRMSE Voting Regressor on six popular regression datasets and compared its performance with several state-of-the-art ensemble regression algorithms. The results show that the RRMSE Voting Regressor consistently achieves lower prediction errors than existing methods across all tested datasets. This improvement highlights the effectiveness of using relative error metrics for weighting in ensemble models. Our approach not only fills a gap in current ensemble regression techniques but also provides a reliable and adaptable method for boosting prediction performance in various machine learning tasks. By leveraging the strengths of individual models through smart weighting, the RRMSE Voting Regressor offers a significant advancement in the field of ensemble learning.

Multinomial logistic regression algorithm for the classification of patients with parkinsonisms

BackgroundAccurate differential diagnosis of neurodegenerative parkinsonisms is challenging due to overlapping early symptoms and high rates of misdiagnosis. To improve the diagnostic accuracy, we developed an integrated classification algorithm using multinomial logistic regression and Scaled Subprofile Model/Principal Component Analysis (SSM/PCA) applied to 18F-fluorodeoxyglucose positron emission tomography (FDG-PET) brain images. In this novel classification approach, SSM/PCA is applied to FDG-PET brain images of patients with various parkinsonisms, which are compared against the constructed undetermined images. This process involves spatial normalization of the images and dimensionality reduction via PCA. The resulting principal components are then used in a multinomial logistic regression model, which generates disease-specific topographies that can be used to classify new patients. The algorithm was trained and optimized on a cohort of patients with neurodegenerative parkinsonisms and subsequently validated on a separate cohort of patients with parkinsonisms.ResultsThe Area Under the Curve (AUC) values were the highest for progressive supranuclear palsy (PSP) (AUC = 0.95), followed by Parkinson’s disease (PD) (AUC = 0.93) and multiple system atrophy (MSA) (AUC = 0.90). When classifying the patients based on their calculated probability for each group, the desired tradeoff between sensitivity and specificity had to be selected. With a 99% probability threshold for classification into a disease group, 82% of PD patients, 29% of MSA patients, and 77% of PSP patients were correctly identified. Only 5% of PD, 6% of MSA and 6% of PSP patients were misclassified, whereas the remaining patients (13% of PD, 65% of MSA and 18% of PSP) are undetermined by our classification algorithm.ConclusionsCompared to existing algorithms, this approach offers comparable accuracy and reliability in diagnosing PD, MSA, and PSP with no need of healthy control images. It can also distinguish between multiple types of parkinsonisms simultaneously and offers the flexibility to easily accommodate new groups.

Risk prediction and early intervention strategies for persistent SARS-CoV-2 infection in patients with non-Hodgkin lymphoma: a retrospective cohort study

BackgroundPatients with non-Hodgkin lymphoma (NHL) face heightened mortality and accelerated disease progression when persistently infected with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). This critical situation underscores the urgent need to identify risk factors and establish early intervention strategies tailored to this vulnerable population. The primary aim of this study was to investigate the risk factors associated with persistent SARS-CoV-2 infection in NHL patients during the COVID-19 pandemic.MethodsA retrospective cohort study was conducted using data from January 2020 to June 2024, obtained from the Aerospace Center Hospital’s database, electronic health records, and laboratory archives. Inclusion criteria comprised patients with confirmed NHL and SARS-CoV-2 infection, with persistence defined as positive viral test results beyond 14 days after initial diagnosis. Patients with incomplete medical records or loss of follow-up were excluded. Predictive models were developed and refined using logistic regression and random forest algorithms. The models incorporated data on demographics, comorbidities, laboratory findings, and imaging results. Model performance was evaluated using accuracy, precision, and the area under the receiver operating characteristic curve (AUC-ROC). Validation was conducted on an independent dataset to ensure generalizability, and the best-performing model guided the development of a prediction tool for early risk assessment and intervention.ResultsKey risk factors for persistent SARS-CoV-2 infection in NHL patients included advanced age, hypertension, diabetes, immunosuppressed status, low lymphocyte count, elevated C-reactive protein, high body mass index, anemia, reduced CD4 + cell count, and the presence of lung lesions. The random forest model demonstrated superior predictive performance, achieving an AUC of 0.93. The study further highlighted that prompt antiviral therapy, adjustments to immunosuppressive regimens, and enhanced monitoring significantly reduced infection persistence.ConclusionsThis study identifies critical risk factors for persistent SARS-CoV-2 infection in NHL patients and underscores the importance of early intervention strategies. These findings may guide clinical decision-making to improve outcomes in this high-risk population.

Estimation Residential Real Estate Price Considering Spatial Variables using Machine Learning in GIS: A Case Study in Chon Buri, Thailand

This research investigates the effectiveness of machine learning (ML) techniques in predicting residential real estate prices, with a specific focus on integrating spatial variables within a Geographic Information System (GIS) framework. A case study in Chon Buri, Thailand, is utilized to compare the predictive accuracy of the Forest-based Classification and Regression (FBCR) algorithm, a GIS-specific ML tool, against established algorithms such as XGBoost (XGB) and Random Forest (RF). The methodology includes data collection through web scraping, rigorous preprocessing using the Interquartile Range (IQR) method for outlier detection, and spatial data integration via ArcGIS Pro. Model performance is evaluated using R², Mean Absolute Error (MAE), Mean Squared Error (MSE), and Root Mean Squared Error (RMSE). The results demonstrate that FBCR outperforms both RF and XGBoost in predicting real estate prices, evidenced by a higher R² value of 0.6412, lower RMSE of 1.529, and lower MAE of 1.131 on the testing dataset. This superior performance highlights FBCR's ability to effectively capture and model the complex spatial relationships that influence property prices. The study underscores the potential of GIS-integrated ML tools in enhancing the accuracy and reliability of real estate valuation, providing valuable insights for urban planning and property market analysis. Key Words: residential real estate, predict price, machine learning, forest-based classification and regression (FBCR), geographic information system (GIS), spatial integration.

Clinical combined PET/CT radiomics model prediction of benefit from platinum-based chemotherapy and chemoradiotherapy in patients with small cell lung cancer.

To develop and validate a clinical combined radiomics model for predicting the treatment response and long-term survival prognosis of small cell lung cancer (SCLC) patients receiving platinum-based chemotherapy, as well as survival outcomes following chemoradiotherapy. A total of 98 SCLC patients treated with platinum-based first-line chemotherapy were included in this study. Five prediction models for assessing the short-term efficacy of platinum-based first-line chemotherapy were developed using a logistic regression algorithm. The performance of the models was assessed by calculating the area under the curve of the receiver operating characteristic curves. For predicting progression-free survival (PFS) and overall survival in the platinum-based chemotherapy group and the chemoradiotherapy group, the optimal cutoff value was determined using X-tile software. Kaplan-Meier survival curves were plotted, and the log-rank test was used to compare survival outcomes. Among the five models for predicting short-term efficacy, the clinical combined positron emission tomography/computed tomography (PET/CT) radiomics model performed the best, achieving areas under the curve of 0.832 and 0.833 for the training and test sets, respectively. The Kaplan-Meier survival analysis indicated that both the high-scoring Combine group and high-scoring PET/CT group were significantly associated with worse PFS and worse overall survival in the platinum-only chemotherapy group. Additionally, the high-scoring CT group was significantly associated with worse PFS in the chemoradiotherapy group. The clinical combined PET/CT radiomics model can noninvasively and accurately predict the response to platinum-based treatments in SCLC as well as long-term survival prognosis, which can contribute to personalized treatment strategies and guide precision therapy for SCLC patients.

Mitochondrial mt12361A>G increased risk of metabolic dysfunction-associated steatotic liver disease among non-diabetes.

Insulin resistance, lipotoxicity, and mitochondrial dysfunction contribute to the pathogenesis of metabolic dysfunction-associated steatotic liver disease (MASLD). Mitochondrial dysfunction impairs oxidative phosphorylation and increases reactive oxygen species production, leading to steatohepatitis and hepatic fibrosis. Artificial intelligence (AI) is a potent tool for disease diagnosis and risk stratification. To investigate mitochondrial DNA polymorphisms in susceptibility to MASLD and establish an AI model for MASLD screening. Multiplex polymerase chain reaction was performed to comprehensively genotype 82 mitochondrial DNA variants in the screening dataset (n = 264). The significant mitochondrial single nucleotide polymorphism was validated in an independent cohort (n = 1046) using the Taqman® allelic discrimination assay. Random forest, eXtreme gradient boosting, Naive Bayes, and logistic regression algorithms were employed to construct an AI model for MASLD. In the screening dataset, only mt12361A>G was significantly associated with MASLD. mt12361A>G showed borderline significance in MASLD patients with 2-3 cardiometabolic traits compared with controls in the validation dataset (P = 0.055). Multivariate regression analysis confirmed that mt12361A>G was an independent risk factor of MASLD [odds ratio (OR) = 2.54, 95% confidence interval (CI): 1.19-5.43, P = 0.016]. The genetic effect of mt12361A>G was significant in the non-diabetic group but not in the diabetic group. mt12361G carriers had a 2.8-fold higher risk than A carriers in the non-diabetic group (OR = 2.80, 95%CI: 1.22-6.41, P = 0.015). By integrating clinical features and mt12361A>G, random forest outperformed other algorithms in detecting MASLD [training area under the receiver operating characteristic curve (AUROC) = 1.000, validation AUROC = 0.876]. The mt12361A>G variant increased the severity of MASLD in non-diabetic patients. AI supports the screening and management of MASLD in primary care settings.

Elastic net with Bayesian Density Estimation model for feature selection for photovoltaic energy prediction

Accurate forecasting of photovoltaic (PV) generated electricity is essential for efficiently managing and integrating Renewable Energy (RE) into electricity distribution systems. This research investigation optimizes Feature Selection (FS) and prediction results for PV energy prediction by applying Bayesian Density Estimation (BDE) with Elastic Net (ELNET) regression analysis. This phenomenon and unacceptable outcomes are prevalent when applying conventional regression algorithms on datasets with significant results and addressing predictor multicollinearity. Improved FS and multicollinearity control has been rendered feasible by ELNET, which integrates the best features of Ridge and Lasso regression. ELNET eliminates these challenges through the implementation of L1 and L2 penalties. Non-parametric prediction Bayesian Density Estimation (BDE) is comprehensive data regarding residual distributions and predictor impacts. By incorporating ELNET’s regularisation and FS abilities with BDE’s statistical prediction and adaptability, the recommended ELNET-BDE is proposed to attain more accurate and reliable predictions. This technique has been used to assess massive data sets developing from Visakhapatnam, India, incorporating historical PV energy generation combined with definite Meteorological Factors (MF). Considering comprehensive data preliminary processing, FS, and validation, ELNET-BDE outperforms existing methods. Research investigations demonstrate that the ELNET-BDE model attains significantly lower Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) than contesting Machine Learning (ML) algorithms like Artificial Neural Network (ANN), Support Vector Machine (SVM), Random Forest (RF), and Gradient Boosting Machines (GBM). Compared with distinct FS techniques, RMSE can be minimized by up to 15% and MAE by up to 20%. The findings specify a substantial improvement in accuracy in prediction, emphasizing how the model can be used for improving solar power grid integration and energy for improved RE management.

Estimation of energy production of solar panels installed in agricultural areas with machine learning algorithms

Predicting solar power generation is used to ensure that solar power plants operate with optimum efficiency, meet the demands of the energy grid and stabilize energy prices. This study aims to predict the medium-term electricity generation of photovoltaic panels with machine learning algorithms. Boosting Regression, Decision Tree Regression, K-Nearest Neighbors Regression, Neural Network Regression, Random Forest Regression, Regularized Linear Regression, and Support Vector Machine Regression algorithms were evaluated. Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and R-Squared (R²) were calculated. It was found that the Random Forest algorithm has the best prediction metrics. A hypothesis was formulated to evaluate the difference between the actual energy generation of the photovoltaic panels and the predicted energy by the Random Forest algorithm. The hypothesis was evaluated by the Mann-Whitney U hypothesis test and the p-value was calculated as greater than 0.05. It was concluded that there is no significant difference between the predicted energy by the Random Forest (RF) algorithm and the actual energy generated by photovoltaic panels. Based on the results of this study, we recommend using the Random Forest algorithm for medium-term energy generation prediction for photovoltaic solar panels.

Predicting viscosity of multi-walled carbon nanotube/water nanofluids using gaussian process regression and emperor penguin optimizer algorithm

Abstract Thermal management is essential in many industries like energy, transportation, and HVAC systems. Since thermal management is so important, there is a need for improved heat transfer fluids, such as nanofluids. The current study uses machine learning (ML) approaches to predict the viscosity of multi-walled carbon nanotube (MWCNT)-water nanofluids. The dataset comprised 446 experimental data points with characteristics such as weight concentration, temperature, shear time, shear stress, and viscosity were used in current research. The dataset was evaluated with a Gaussian Process Regression (GPR) model and the hyperparameters were further optimized via Emperor Penguin Optimizer (EPO).With the achieved values of R2 of 0.9995, RMSE of 0.0016, and MAPE of 1.89%, the proposed model GPR-EPO, yielded better predictive performance than other machine learning models such as Gradient Boosting Regressor, XGBoost, and Extra Trees Regressor. Additionally, for validation the GPR-EPO model was compared with conventional model like Batchelor and Einstein, it was found more precise and yielded better predictive performance. This study highlights the significant role of AI-driven technique in predicting nanofluid viscosity with accuracy as well as reducing the experimental efforts. The GPR method was found to have the best performance by using radial basis function (RBF) kernel and optimized the hyperparameters with EPO algorithm. Such models can serve as a valuable tool for engineers and researchers to investigate nanofluids and develop efficient thermal management systems.

Machine Learning and digital Imaging for Spatiotemporal Monitoring of Stress Dynamics in the clonal plant Carpobrotus edulis: Uncovering a Functional Mosaic.

Rapid, large-scale monitoring is critical to understanding spatiotemporal plant stress dynamics, but current physiological stress markers are costly, destructive, and time-consuming. This study aimed to evaluate the potential of machine learning to non-destructively predict leaf betalains-yellow to reddish pigments unique to Caryophyllales species-for the first time, and to explore betalains' intra-individual variation on a clonal species and its role to respond to stressful periods. We characterized the betalainic profile of an invasive clonal plant for the first time, Carpobrotus edulis (L.) NE Br. (the cape fig), via HPLC. We measured multiple stress markers over a year, including betalain content using our optimized method, where the species is spreading. Additionally, 3,735 digital images at the leaf level were taken. Machine learning regression algorithms were trained to predict betalain accumulation from digital images, outperforming classic spectroradiometer measurements. Betalain content increased sharply in non-reproductive ramets during extreme abiotic conditions in summer and during senescence in reproductive ramets. The stress markers revealed a strong intra-individual functional mosaic, underscoring the importance of spatiotemporal dimensions in stress tolerance. We developed a scalable, non-destructive tool for betalain research that integrates digital imaging with machine learning. This approach opens new possibilities for understanding spatiotemporal stress responses, particularly in clonal plant systems, using artificial intelligence.

Prediction of temperature and humidity in footwear warehouses based on improved arithmetic optimization twin support vector machine for regression

Abstract Temperature and humidity fluctuations in storage environments significantly impact the material properties and quality of footwear. Temperature and humidity fluctuations in storage environments significantly impact the material properties and quality of footwear. Traditional warehouse management approaches typically rely on fixed environmental control systems, which struggle to adapt to real-time dynamic changes, leading to high energy consumption, inefficient management, and potential impacts on footwear quality. To address the aforementioned issues, the study employed an improved arithmetic optimization algorithm (AOA) to optimize the twin support vector machine for regression algorithm (LAOA-TSVR) and develop the corresponding prediction model. First, SPSS was used for correlation analysis to identify the primary factors influencing temperature and humidity in the warehouse. Second, to validate the model’s prediction accuracy, the algorithm was compared with three common models: backpropagation neural network (BP), radial basis function neural network (RBF), and twin support vector regression (TSVR). The results demonstrate that the LAOA-TSVR exhibits fast convergence, high prediction accuracy, and robust generalization ability. Finally, the model was applied in a real warehouse environment. The results indicate that when the temperature and humidity errors in the warehouse remain within ±10% and ±5%, the endpoint hit rates for warehouse temperature reach 96.8% and 88.1%. while those for warehouse humidity reach 93.6% and 86.3%, respectively. For ±10% errors, the dual hit rate for warehouse temperature and humidity is 92.7%. The model establishes a theoretical foundation for accurately predicting temperature and humidity in warehouse environments, thereby facilitating the optimization of control strategies and improving warehouse management efficiency.

Clinical validation of a novel hand dexterity measurement device.

The lack of sensitive objective outcome measures for hand dexterity is a barrier for clinical assessment of neurological conditions and has negatively affected clinical trials. Here, we clinically validate a new method for measuring hand dexterity, a novel hand worn sensor that digitises the Finger Tapping Test. The device was assessed in a cohort of 180 healthy controls and 51 people with Amyotrophic Lateral Sclerosis (ALS) and compared against rating scales and traditional measures (Nine Hole Peg test and grip dynamometry). 14 features were extracted from the device and using a logistic regression algorithm, a 0-100 dexterity performance score was generated for each participant, which accounted for age/sex differences. The device returned objective ratings of a participant's hand dexterity (dominant, non-dominant and overall score). The average overall dexterity performance score in all healthy participants was 88 ± 17 (mean ± standard deviation). The overall dexterity score was statistically significantly worse in participants with ALS (age/sex matched healthy subset: 80 ± 20, ALS: 45 ± 32, p-value < 0.0001). The device also had a higher completion rate, (94% dominant hand) compared to the traditional measures (82% dominant hand). This test and scoring system have been validated and the regression model was developed using a framework that is potentially applicable to any relevant condition. This device could act as an objective outcome measure in clinical trials and may be useful in improving patient care.

A New Passive Islanding Detection Method Based on Rate of Change of Frequency

Abstract To address the issues of low accuracy and significant detection blind zones inherent in traditional passive islanding detection methods, we propose an innovative islanding detection method that leverages the rate of frequency change. This method is formulated by meticulously analyzing the frequency variation patterns at the Point of Common Coupling (PCC) before and after islanding occurs. The frequency-locked loop control mechanism is employed to track the grid frequency using reactive power negative feedback. Subsequently, a high-pass filter is utilized to extract high-frequency components, which are then fitted using the local weighted linear regression algorithm to generate the high-frequency components of the frequency variation curve. By calculating the slope of this frequency fitting curve, the fluctuation changes in the slope curve are utilized to detect the occurrence of islanding. MATLAB/Simulink simulation results validate that the proposed method can accurately detect islanding effects under diverse conditions and eliminate detection blind zones, demonstrating robust applicability.&amp;#xD;

Simplifying Field Traversing Efficiency Estimation Using Machine Learning and Geometric Field Indices

Enhancing agricultural machinery field efficiency offers substantial benefits for farm management by optimizing the available resources, thereby reducing cost, maximizing productivity, and supporting sustainability. Field efficiency is influenced by several unpredictable and stochastic factors that are difficult to determine due to the inherent variability in field configurations and operational conditions. This study aimed to simplify field efficiency estimation by training machine learning regression algorithms on data generated from a farm management information system covering a combination of different field areas and shapes, working patterns, and machine-related parameters. The gradient-boosting regression-based model was the most effective, achieving a high mean R2 value of 0.931 in predicting field efficiency, by taking into account only basic geometric field indices. The developed model showed also strong predictive performance for indicative agricultural fields located in Europe and North America, reducing considerably the computational time by an average of 73.4% compared to the corresponding analytical approach. Overall, the results of this study highlight the potential of machine learning for simplifying field efficiency prediction without requiring detailed knowledge of a plethora of variables associated with agricultural operations. This can be particularly valuable for farmers who need to make informed decisions about resource allocation and operational planning.

Comparison and Optimization of Machine Learning Methods for Fault Detection in District Heating and Cooling Systems

In this study, the methods used for the detection of sub-station pollution failures in District Heating and Cooling (DHC) systems are analyzed. In the study, high, medium and low-level pollution situations are considered and machine learning methods are applied for the detection of these failures. Random Forest, Decision Tree, Logistic Regression and CatBoost Regression algorithms are compared within the scope of the analysis. The models are trained to perform fault detection at different pollution levels. In order to improve the model performance, hyperparameter optimization was performed with random search optimization and the most appropriate values were selected. The results show that the CatBoost Regression algorithm provides the highest accuracy and overall performance compared to other methods. The CatBoost model stood out with an accuracy of 0.9832 and a superior performance. These findings reveal that CatBoost-based approaches provide an effective solution in situations requiring high accuracy, such as contamination detection in DHC systems. The study makes an important contribution as a reliable fault detection solution in industrial applications.

Hyperspectral estimation of chlorophyll content in grapevine based on feature selection and GA-BP

Leaf chlorophyll content (LCC) is a key indicator for assessing the growth of grapes. Hyperspectral techniques have been applied to LCC research. However, quantitative prediction of grape LCC using this technique remains challenging due to baseline drift, spectral peak overlap, and ambiguity in the sensitive spectral range. To address these issues, two typical crop leaf hyperspectral data were collected to reveal the spectral response characteristics of grape LCC using standardization by variables (SNV) and multiple far scattering correction (MSC) preprocessing variations. The sensitive spectral range is determined by Pearson’s algorithm, and sensitive features are further extracted within that range using Extreme Gradient Boosting (XGBoost), Recursive Feature Elimination (RFE), and Principal components analysis (PCA). Comparison of the prediction ability of Random Forest Regression (RFR) algorithm, Support Vector Machine Regression (SVR) model, and Genetic Algorithm-Based Neural Network (GA-BP) on grape LCC based on sensitive features. A SNV-RFE-GA-BP framework for predicting hyperspectral LCC in grapes is proposed, where =0.835 and NRMSE = 0.091. The analysis results show that SNV and MSC treatments improve the correlation between spectral reflectance and LCC, and different feature screening methods have a greater impact on the model prediction accuracy. It was shown that SNV-based processed hyperspectral data combined with GA-BP has great potential for efficient chlorophyll monitoring in grapevine. This method provides a new framework theory for constructing a hyperspectral analytical model of grapevine key growth indicators.

Insight into the Relationships Between Chemical, Protein and Functional Variables in the PBP/GOBP Family in Moths Based on Machine Learning.

During their lives, insects must cope with a plethora of chemicals, of which a few will have an impact at the behavioral level. To detect these chemicals, insects use several protein families located in their main olfactory organs, the antennae. Inside the antennae, odorant-binding proteins (OBPs), as the most studied protein family, bind volatile chemicals to transport them. Pheromone-binding proteins (PBPs) and general-odorant-binding proteins (GOPBs) are two subclasses of OBPs and have evolved in moths with a putative olfactory role. Predictions for OBP-chemical interactions have remained limited, and functional data collected over the years unused. In this study, chemical, protein and functional data were curated, and related datasets were created with descriptors. Regression algorithms were implemented and their performance evaluated. Our results indicate that XGBoostRegressor exhibits the best performance (R2 of 0.76, RMSE of 0.28 and MAE of 0.20), followed by GradientBoostingRegressor and LightGBMRegressor. To the best of our knowledge, this is the first study showing a correlation among chemical, protein and functional data, particularly in the context of the PBP/GOBP family of proteins in moths.

Quantum machine learning regression optimisation for full-scale sewage sludge anaerobic digestion

Quantum machine learning regression optimisation for full-scale sewage sludge anaerobic digestion

Machine learning and numerical simulations for predicting critical crack conditions in wooden panels

PurposeArtworks made of hygroscopic materials, like wooden panel paintings, are susceptible to environmental conditions. Traditional panel paintings typically consist of a wooden panel coated with layers of gesso, paint and varnish. Due to environmental fluctuations, the gesso layer and the wood panel may respond differently to moisture changes, triggering potential fractures. The investigation of such phenomena is of high interest, but it is still scarcely studied by engineers.Design/methodology/approachThe proposed study aimed to create a simplified 3D finite element model for paintings to identify environmental conditions that could exceed critical strain levels. A penny-shaped crack within the gesso layer was modelled and, after applying a given deformation, the strain energy density failure criterion was used to assess if the crack was in a critical state.FindingsVarious combinations of geometric parameters of the model were explored, and to save computational time and cost, machine learning algorithms (namely extreme gradient boosting machines and Gaussian process regression algorithms) were introduced. The analyses were carried out on different panel paintings 3D models obtained by varying the wooden species and the boundary conditions, for exploring a wide number of combinations.Originality/valueMoreover, the integration of machine learning can potentially reduce the reliance on numerical simulations and offer new insights into the conservation of artworks, a field in which such tools are still scarcely exploited.