Random Tree Model Research Articles

The performance comparison of the decision tree models on the prediction of seismic gravelly soil liquefaction potential based on dynamic penetration test

Seismic liquefaction has been reported in sandy soils as well as gravelly soils. Despite sandy soils, a comprehensive case history record is still lacking for developing empirical, semi-empirical, and soft computing models to predict this phenomenon in gravelly soils. This work compiles documentation from 234 case histories of gravelly soil liquefaction from across the world to generate a database, which will then be used to develop seismic gravelly soil liquefaction potential models. The performance measures, namely, accuracy, precision, recall, F-score, and area under the receiver operating characteristic curve, were used to evaluate the training and testing tree-based models’ performance and highlight the capability of the logistic model tree over reduced error pruning tree, random tree and random forest models. The findings of this research can provide theoretical support for researchers in selecting appropriate tree-based models and improving the predictive performance of seismic gravelly soil liquefaction potential.

Characterizing and Predicting the Quality of Milled Rice Grains Using Machine Learning Models

Physical classification is the procedure adopted by the rice unloading, delivery, storage, and processing units for the commercial characterization of the quality of the grains. This step occurs mostly by the conventional method, which demands more time and specialized labor, and the results are subjective since the evaluation is visual. In order to make the operation faster, more accurate, and less dependent, non-destructive technologies and computational intelligence can be applied to characterize grain quality. Therefore, this study aimed to characterize and predict the quality of whole, processed rice grains, as well as classify any defects present. This was achieved by sampling from the upper and lower points of four silo dryers with capacities of up to 40,000 sacks. The grain samples had moisture contents of 16%, 17%, 18%, and 19% and were subjected to drying-aeration until reaching 12% moisture content (w.b.). Near-infrared spectroscopy technology and Machine Learning algorithm models (Artificial Neural Networks, decision tree algorithms Quinlan’s algorithm, Random Tree, REPTree, and Random Forest) were employed for this purpose. By analyzing Pearson’s correlation statistics, a strong negative correlation (R2 = 0.98) was found between moisture content and the yield of whole grains. Conversely, a strong positive correlation (R2 = 0.97) was observed between moisture content and classified physical defects across the various characterized physicochemical constituents. These findings indicate the effectiveness of near-infrared spectroscopy technology. The Random Tree model (RandT) successfully predicted the grain quality outcomes and is therefore recommended as the model of choice, obtained Pearson’s correlation coefficient (r = 0.96), mean absolute error (MAE = 0.017), and coefficient of determination (R2 = 0.92). The results obtained here reveal that the combination of near-infrared spectroscopy technology and Machine Learning algorithm models is an excellent non-destructive alternative to manual physical classification for characterizing the physicochemical quality of whole and defective rice grains.

Estimation of forest canopy closure in northwest Yunnan based on multi-source remote sensing data colla-boration.

Forest canopy closure (FCC) is an important parameter to evaluate forest resources and biodiversity. Using multi-source remote sensing collaborative means to achieve regional forest canopy closure inversion with low cost and high-precision is a research hotspot. Taking ICESat-2/ATLAS data as the main information source and combined with data of 54 measured plots, we estimated FCC value by the Bayesian optimization (BO) algorithm improved random forest (RF), K-nearest neighbor (KNN), and gradient boosting regression tree (GBRT) model at footprint-scale. Combined with multi-source remote sensing image Sentinel-1/2 and terrain factors, we estimated the regional-scale FCC value of Shangri-La in the northwest Yunnan based on deep neural network (DNN) optimized by BO algorithm. The results showed that six characteristic parameters (percentage of tree canopy, standard deviation of relative height of photons at the top of the canopy, minimum canopy height, difference between 98% canopy height and median canopy height in the segment, number of top canopy photons, apparent surface reflectance) out of the 50 parameters that were extracted from ATLAS lidar footprint had higher contribution rate after RF characteristic variable optimization, which could be used as model variable for footprint-scale remote sensing estimation. Among BO-RF, BO-KNN, and BO-GBRT models, the FCC results estimated by the BO-GBRT model were the best at footprint-scale. The coefficient of determination (R2) was 0.65, the root mean square error (RMSE) was 0.10, the mean absolute residual (RS) was 0.079, and the prediction accuracy (P) was 0.792 for leave-one-out cross validation. It could be used as the FCC estimation model of 74808 ATLAS footprints for forest in the study area. We used the ATLAS footprint-scale FCC value of forest as the large sample data of the regional-scale BO-DNN model and combined with multi-source remote sensing factors to estimate FCC in the study area, the accuracy of the 10-fold cross-validation BO-DNN model was R2=0.47, RMSE=0.22, P=0.558. The mean values of FCC in the study area estimated by BO-DNN model and ordinary Kriging (OK) interpolation were 0.46 and 0.52, respectively, and the values mainly distributed in 0.3-0.6, accounting for 77.8% and 81.4%, respectively. The FCC efficiency obtained directly by the OK interpolation method was higher (R2=0.26), but the prediction accuracy was significantly lower than the BO-DNN model (R2=0.49). The FCC high value was distributed from northwest to southeast in the study area, and the northern and southeastern regions were the main distribution areas of high and low FCC values, respectively. It had certain advantages to estimate mountain area FCC based on ICESat-2/ATLAS high-density footprint, and the estimation results of small sample data at footprint-scale could be used as large sample data of deep learning model at region-scale, which would provide a reference for the low-cost and high-precision to FCC estimation on the footprint-scale up to the extrapolated regional-scale.

Bioinformatic Analysis of the Potential Common Pathogenic Mechanisms for Psoriasis and Metabolic Syndrome.

The pathogeneses of psoriasis and metabolic syndrome are closely related; however, the underlying biological mechanisms are yet to be clarified. A psoriasis training set was downloaded from the Gene Expression Omnibus database and analyzed to identify the differentially expressed genes (|logFC|> 1 and adjust P < 0.05). Differentially expressed genes for metabolic syndrome were obtained from the GeneCards, Online Mendelian Inheritance in Man, and DisGeNET databases, and crosstalk genes were obtained for multiple enrichment analysis after identifying the disease intersection. Characteristic crosstalk genes were screened using the least absolute shrinkage and selection operator regression model and random forest tree model, and the genes with area under the receiver operating characteristic curve > 0.7 were selected for validation by the two validation sets. Differential analyses of immune cell infiltration were performed on psoriasis lesion and control samples using the CIBERSORT and ImmuCellAI methods, and correlation analyses were performed between the screened signature crosstalk genes and immune cell infiltration. Significant crosstalk genes were analyzed based on the psoriasis area and severity index and on the responses to biological agents. We found five signature genes (NLRX1, KYNU, ABCC1, BTC, and SERPINB4) were screened based on two machine learning algorithms, and NLRX1 was validated. The infiltration of multiple immune cells in psoriatic lesions and non-lesions was associated with NLRX1 expression. NLRX1 was found to be associated with psoriasis severity and response rate after the use of biologics. NLRX1 could be a significant crosstalk gene for psoriasis and metabolic syndrome.

Research on Identifying Important Factors and Prediction of Online Service Satisfaction for Mobile Phone Users

Our lives cannot do without the internet. How to improve the network quality has always been an essential problem. The paper explores the important factors of online service satisfaction and the best predicting model. Based on the data offered by Beijing Mobile Company, we identify main factors affecting online service satisfaction by calculating their mutual information values. The factors include signal problem factors, scene factors and software usage factors. Additionally, based on decision tree model and models with decision tree as base learner, we predict the online service satisfaction. The result shows that random tree model with One Vs Rest mode has the greatest accuracy among the models which offers telecommunications companies insight.

Comprehensive analysis and molecular map of Hippo signaling pathway in lower grade glioma: the perspective toward immune microenvironment and prognosis.

The activation of YAP/TAZ transcriptional co-activators, downstream effectors of the Hippo/YAP pathway, is commonly observed in human cancers, promoting tumor growth and invasion. The aim of this study was to use machine learning models and molecular map based on the Hippo/YAP pathway to explore the prognosis, immune microenvironment and therapeutic regimen of patients with lower grade glioma (LGG). SW1783 and SW1088 cell lines were used as in vitro models for LGG, and the cell viability of the XMU-MP-1 (a small molecule inhibitor of the Hippo signaling pathway) treated group was evaluated using a Cell Counting Kit-8 (CCK-8). Univariate Cox analysis on 19 Hippo/YAP pathway related genes (HPRGs) was performed to identify 16 HPRGs that exhibited significant prognostic value in meta cohort. Consensus clustering algorithm was used to classify the meta cohort into three molecular subtypes associated with Hippo/YAP Pathway activation profiles. The Hippo/YAP pathway's potential for guiding therapeutic interventions was also investigated by evaluating the efficacy of small molecule inhibitors. Finally, a composite machine learning models was used to predict individual patients' survival risk profiles and the Hippo/YAP pathway status. The findings showed that XMU-MP-1 significantly enhanced the proliferation of LGG cells. Different Hippo/YAP Pathway activation profiles were associated with different prognostic and clinical features. The immune scores of subtype B were dominated by MDSC and Treg cells, which are known to have immunosuppressive effects. Gene Set Variation Analysis (GSVA) indicated that subtypes B with a poor prognosis exhibited decreased propanoate metabolic activity and suppressed Hippo pathway signaling. Subtype B had the lowest IC50 value, indicating sensitivity to drugs that target the Hippo/YAP pathway. Finally, the random forest tree model predicted the Hippo/YAP pathway status in patients with different survival risk profiles. This study demonstrates the significance of the Hippo/YAP pathway in predicting the prognosis of patients with LGG. The different Hippo/YAP Pathway activation profiles associated with different prognostic and clinical features suggest the potential for personalized treatments.

Credit Card Default Prediction based on Machine Learning Techniques

In recent years, with the development of society and economy, credit cards have been popularized due to their low interest rate and easy payment. However, with the advent of the epidemic era, the unemployment rate has increased, making the probability of credit card defaults rising. The prediction of credit card default helps banks and financial institutions balance the risk and economic interests, contributes to the stable and healthy development of the financial industry, and plays an important role in bank credit control. Therefore, this paper addresses the credit card default prediction problem by using Random forest, Decision tree, LightGBM, XGBoost, Logistic regression, and Adaboost models to make predictions and compare the results. The outcomes demonstrate that LightGBM algorithm has the most outstanding prediction score, and its AUC value can reach 0.78 and recall rate reaches 0.95.

A diagnostic model based on color vision examination for dysthyroid optic neuropathy using Hardy-Rand-Rittler color plates.

To investigate color vision deficiency and the value of Hardy-Rand-Rittler (HRR) color plates in monitoring dysthyroid optic neuropathy (DON) to improve the diagnosis of DON. The participants were divided into DON and non-DON (mild and moderate-to-severe) groups. All the subjects underwent HRR color examination and comprehensive ophthalmic examinations. The random forest and decision tree models based on the HRR score were constructed by R software. The ROC curve and accuracy of different models in diagnosing DON were calculated and compared. Thirty DON patients (57 eyes) and sixty non-DON patients (120 eyes) were enrolled. The HRR score was lower in DON patients than in non-DON patients (12.1 ± 6.2 versus 18.7 ± 1.8, p < 0.001). The major color deficiency was red-green deficiency in DON using HRR test. The HRR score, CAS, RNFL, and AP100 were found to be important factors in predicting DON from random forest and selected by decision tree to construct the multifactor model. The sensitivity, specificity, and the area under the curve (AUC) of the HRR score were 86%, 72%, and 0.87, respectively. The HRR score decision tree had a sensitivity, specificity, and AUC of 93%, 57%, and 0.75, respectively, with an accuracy of 82%. The data of the multifactor decision tree were 90%, 89%, and 0.93 for sensitivity, specificity, and AUC, respectively, with an accuracy of 91%. The HRR test was valid as screening method for DON. The multifactor decision tree based on the HRR test improved the diagnostic efficacy for DON. An HRR score of less than 12 and red-green deficiency may be characteristic of DON.

Modeling the spatial variation of calcium carbonate equivalent to depth using machine learning techniques

Inorganic carbon is the largest source of carbon in terrestrial surface, particularly in arid and semiarid regions, including the Chahardowli Plain in western Iran. Inorganic carbon plays an equal or greater role than organic soil carbon in these areas, although less attention has been paid in quantifying their variability. The objective of this study was to model and map calcium carbonate equivalent (CCE) presenting inorganic carbon in soil using machine learning and digital soil mapping techniques. Chahardowli Plain in foothills of the Zagros Mountains in the southeast of Kurdistan Province in Iran was taken as a case study area. CCE was measured at 0-5, 5-15, 15-30, 30-60, and 60-100cm depths following GloalSoilMap.net project specifications. A total of 145 samples were collected from 30 soil profiles using the conditional Latin hypercube (cLHS) method of sampling. Relationships between CCE and environmental predictors were modeled using random forest (RF) and decision tree (DT) models. In general, the RF model performed slightly superior than the DT model. The mean value of CCE increased with soil depth, from 3.5% (0-5cm) to 63.8% (30-60cm). Remote sensing (RS) variables and terrestrial variables were equally important. The importance of RS variables was higher at the surface than terrestrial variables, and vice versa. The most significant variables were Channel Network Base Level (CNBL) variable and Difference Vegetation Index (DVI) with the same variable importance value (21.1%). In areas affected by river activities, the use of the CNBL and vertical distance to channel networks (VDCN) as variables in digital soil mapping (DSM) could increase the accuracy of soil property prediction maps. The VDCN played a principal role in soil distribution in the study area by affecting the rate of discharge and, thus, erosion and sedimentation. A high percentage of carbonate in parts of the region could exacerbate nutrient deficiencies for most crops and provide information for sustainably managing agricultural activity.

Association of Long-term exposure to air pollution and residential greenness with lipid profile: Mediating role of inflammation

Lipidemic effect of air pollutants are still inconsistent and their joint effects are neglected. Meanwhile, identified inflammation pathways in animal have not been applied in epidemiological studies, and beneficial effect of residential greenness remained unclear. Therefore, we used data from typically air-polluted Chinese cities to answer these questions. Particulate matter (PM) with a diameter of ≤ 1 µm (PM1), PM with a diameter of ≤ 2.5 µm (PM2.5), PM with a diameter of ≤ 10 µm (PM10), sulphur dioxide (SO2), nitrogen dioxide (NO2), and ozone (O3) were predicted by space–time extremely randomized trees model. Residential greenness was reflected by Normalized Difference Vegetation Index (NDVI). Total cholesterol (TC), triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C) were measured, and atherogenic coefficient (AC) and TG/HDL-C (TGH) ratio were calculated to indicate lipid metabolism. Generalized additive mixed model and quantile g-computation were respectively conducted to investigate individual and joint lipidemic effect of air pollutants. Covariates including demographical characteristics, living habits, meteorological factors, time trends, and disease information were considered to avoid confounding our results. Complement C3 and high-sensitivity C-reactive protein (hsCRP) were analyzed as potential mediators. Finally, association between NDVI and lipid markers were explored. We found that long-term air pollutants exposure were positively associated with lipid markers. Complement C3 mediated 54.72% (95% CI: 0.30, 63.10) and 72.53% (95% CI: 0.65, 77.61) of the association between PM1 and TC and LDL-C, respectively. We found some significant associations of lipid markers with NDVI1000 m rather than NDVI500 m. BMI, disease status, smoke/drink habits are important effect modifiers. Results are robust in sensitive analysis. Our study indicated that air pollutants exposure may detriment lipid metabolism and inflammation may be the potential triggering pathways, while greenness may exert beneficial effects. This study provided insights for the lipidemic effects of air pollution and greenness.

Clinical Hematochemical Parameters in Differential Diagnosis between Pediatric SARS-CoV-2 and Influenza Virus Infection: An Automated Machine Learning Approach.

The influenza virus and the novel beta coronavirus (SARS-CoV-2) have similar transmission characteristics, and it is very difficult to distinguish them clinically. With the development of information technologies, novel opportunities have arisen for the application of intelligent software systems in disease diagnosis and patient triage. A cross-sectional study was conducted on 268 infants: 133 infants with a SARS-CoV-2 infection and 135 infants with an influenza virus infection. In total, 10 hematochemical variables were used to construct an automated machine learning model. An accuracy range from 53.8% to 60.7% was obtained by applying support vector machine, random forest, k-nearest neighbors, logistic regression, and neural network models. Alternatively, an automated model convincingly outperformed other models with an accuracy of 98.4%. The proposed automated algorithm recommended a random tree model, a randomization-based ensemble method, as the most appropriate for the given dataset. The application of automated machine learning in clinical practice can contribute to more objective, accurate, and rapid diagnosis of SARS-CoV-2 and influenza virus infections in children.

A Novel Methodology for Human Kinematics Motion Detection Based on Smartphones Sensor Data Using Artificial Intelligence

Kinematic motion detection aims to determine a person’s actions based on activity data. Human kinematic motion detection has many valuable applications in health care, such as health monitoring, preventing obesity, virtual reality, daily life monitoring, assisting workers during industry manufacturing, caring for the elderly. Computer vision-based activity recognition is challenging due to problems such as partial occlusion, background clutter, appearance, lighting, viewpoint, and changes in scale. Our research aims to detect human kinematic motions such as walking or running using smartphones’ sensor data within a high-performance framework. An existing dataset based on smartphones’ gyroscope and accelerometer sensor values is utilized for the experiments in our study. Sensor exploratory data analysis was conducted in order to identify valuable patterns and insights from sensor values. The six hyperparameters, tunned artificial indigence-based machine learning, and deep learning techniques were applied for comparison. Extensive experimentation showed that the ensemble learning-based novel ERD (ensemble random forest decision tree) method outperformed other state-of-the-art studies with high-performance accuracy scores. The proposed ERD method combines the random forest and decision tree models, which achieved a 99% classification accuracy score. The proposed method was successfully validated with the k-fold cross-validation approach.

Exploring the potential of artificial intelligence techniques in prediction of the removal efficiency of vortex tube silt ejector

A vortex tube silt ejector is a curative hydraulic structure used to remove sediment deposits from canals and is recognized as one of the most efficient substitutes for physically removing canal sediment. The spatially varied flow in the channel and the rotational flow behavior in the tube make the silt removal process complex. It is even harder to accurately predict the silt removal efficiency by traditional models accurately. However, artificial intelligence (AI) and machine learning approaches have emerged as robust substitutes for studying complex processes. Therefore, this research makes use of AI approaches; support vector machine (SVM), random forest (RF), random tree (RT), and multivariate adaptive regression spline (MARS) to compute the vortex tube silt ejection efficiency using the laboratory data sets. The outcomes of the artificial intelligence (AI)-based techniques also were compared with traditional models. It was found that the RT model (root mean square error, RMSE = 2.165, Nash Sutcliffe efficiency, NSE = 0.98) outperforms the other applied approaches which had relatively more significant result errors. The sensitivity analysis of the process depicts the extraction ratio as the key parameter in the computation of vortex tube silt ejector removal efficiency. The findings of the AI-based approaches discussed in the current study might be helpful for hydraulic engineers as well as researchers in the assessment of the removal efficiency of vortex tube silt ejectors.

A Gibbsian Random Tree with Nearest Neighbour Interaction

We revisit the random tree model with nearest-neighbour interaction as described in Dunlop and Mardin (J Stat Phys 189:38, 2022), enhancing growth. When the underlying free Bienaymé–Galton–Watson (BGW) model is sub-critical, the (non-Markov) model with interaction exhibits a phase transition between sub- and super-critical regimes. In the critical regime, using tools from dynamical systems, we show that the partition function of the model approaches a limit at rate $$n^{-1}$$ in the generation number n. In the critical regime with almost sure extinction, we also prove that the mean number of external nodes in the tree at generation n decays like $$n^{-2}$$ . Finally, we give a spin representation of the random tree, opening the way to tools from the theory of Gibbs states, including FKG inequalities. We extend the construction in Dunlop and Mardin (J Stat Phys 189:38, 2022), when the law of the branching mechanism of the free BGW process has unbounded support.

Ancient glass classification based on random forest algorithm and decision tree model

The texture change of ancient glass depends on its chemical composition and preservation environment, and it is highly susceptible to weathering by the burial environment. In this paper, a classification model based on random forest algorithm is established to solve the problem of analyzing the classification laws of high potassium glass and lead-barium glass; a decision tree model based on decision tree is established to solve the problem of subclass differentiation of glass categories and analyzing their rationality and sensitivity. The reasonableness of the model and the accuracy of the data are tested by the plain Bayesian algorithm.

A Comparative Study, Prediction and Development of Chronic Kidney Disease Using Machine Learning on Patients Clinical Records

Chronic Kidney Disease (CKD) has become a major problem in modern times, and it is dubbed the silent assassin due to its delayed signs. To overcome these critical issues, early identification may minimize the prevalence of chronic diseases, though it is quite difficult because of different kinds of limitations in the dataset. The novelty of our study is that we extracted the best features from the dataset in order to provide the best classification models for diagnosing patients with chronic kidney disease. In our study, we used CKD patients’ clinical datasets to predict CKD using some popular machine learning algorithms. After handling missing values, K-means clustering has been performed. Then feature selection was done by applying the XGBoost feature selection algorithm. After selecting features from our dataset, we have used a variety of machine learning models to determine the best classification models, including Neural Network (NN), Random Forest (RF), Support Vector Machine (SVM), Random Tree (RT), and Bagging Tree Model (BTM). Accuracy, Sensitivity, Specificity, and Kappa values were used to evaluate model performance.

Prediction Model of Type 2 Diabetes Mellitus for Oman Prediabetes Patients Using Artificial Neural Network and Six Machine Learning Classifiers

The early diagnosis of type 2 diabetes mellitus (T2DM) will provide an early treatment intervention to control disease progression and minimise premature death. This paper presents artificial intelligence and machine learning prediction models for diagnosing T2DM in the Omani population more accurately and with less processing time using a specially created dataset. Six machine learning algorithms: K-nearest neighbours (K-NN), support vector machine (SVM), naive Bayes (NB), decision tree, random forest (RF), linear discriminant analysis (LDA), and artificial neural networks (ANN) were applied in MATLAB. All data used were clinical data collected manually from a prediabetes register and the Al Shifa health system of South Al Batinah Province in Oman. The results were compared with the most widely used Pima Indian Diabetes dataset. Eleven clinical features were taken into consideration for predicting T2DM. The random forest and decision tree models performed better than all the other algorithms, providing an accuracy of 98.38% for Oman data. When the same model and number of features were used, the accuracy obtained with the Oman dataset exceeded PID by 9.1%. The analysis showed that T2DM diagnosis efficiency increased with more features, which is of help in the case of many missing values.

Credal-Decision-Tree-Based Ensembles for Spatial Prediction of Landslides

Spatial landslide susceptibility assessment is a fundamental part of landslide risk management and land-use planning. The main objective of this study is to apply the Credal Decision Tree (CDT), adaptive boosting Credal Decision Tree (AdaCDT), and random subspace Credal Decision Tree (RSCDT) models to construct landslide susceptibility maps in Zhashui County, China. The observed 169 historical landslides were classified into two groups: 70% (118 landslides) for training and 30% (51 landslides) for validation. To compare and validate the performance of the three models, the receiver operating characteristic (ROC) curve and the area under the ROC curve (AUC) were utilized. Specifically, the success rates of the CDT model, AdaCDT model, and RSCDT model were 0.788, 0.821, and 0.847, respectively, while the corresponding prediction rates were 0.771, 0.802, and 0.861, respectively. In sum, the two ensemble models can effectively improve the performance accuracy of an individual CDT model, and the RSCDT model was proven to be superior to the other two models. Therefore, ensemble models are capable of being novel and promising approaches for the spatial prediction and zonation of a certain region’s landslide susceptibility.

Using Random Forest and Random Tree model to Predict the splitting tensile strength for the concrete with basalt fiber reinforced concrete

The best model for predicting concrete splitting tensile strength using Basalt Fiber Reinforced Concrete (BFRC) was found utilising Random Forest (RF) and Random Tree (RT) implemented in this study. A total of 74 datasets were collected for this investigation from various academic papers. The entire data set is split into 51 training data sets and 23 testing data sets. The software which used for analysis in WEKA software. Cement, fine aggregate/crushed sand, coarse aggregate, water, superplasticizer, fly ash, Basalt Fiber Reinforced Concrete BFRC, diameter, length, and curing time are the input factors, while the compressive strength of the concrete containing BFRC is the output variable. Three performance assessment indices are used to assess the performance of the created models: Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Correlation Coefficient (CC). Containing CC (0.9889, 0.9579) and lower MAE (0.0886, 0.1842), RMSE (0.1430, 0.2406) for the training and testing data sets, the Random Forest was shown to be the best model for predicting the splitting tensile strength of concrete with BFRC. And by analysing the sensitivity it shows that the curing time is the most sensitive input among all the inputs.

Modeling of the mass flow rate of natural gas flow stream using genetic/decision tree/kernel-based data-intelligent approaches

The large consumption of natural gas, one of the most important energy sources in the world, necessitates reliable, precise, and accurate calculation of gas flow rate and amount in order to use this resource in an efficient and sustainable way. The present computational study investigates the possibilities of several soft-computing strategies in estimating the mass flow rate of natural gas flow stream (kg/h) (output variable) based on four input variables of orifice plate diameter ratio, differential pressure of orifice plate (kPa), operating pressure of the natural gas (bar), and operating temperature of the natural gas (°C). A genotype/phenotype genetic algorithm (gene expression programming (GEP) technique), two decision tree-based methods (random forest (RF), random tree (RT) models), and two kernel-based approaches (Gaussian process regression (GPR) and support vector machines (SVM) methods) were applied for the first time to predict gas mass flow rate. Coefficient of correlation (CC), mean absolute error (MAE), root mean square error (RMSE), Scattering index (SI), Nash–Sutcliffe efficiency (NSE), and mean absolute relative error (MARE) were computed as the statistical performance evaluators to determine of the best-performing soft-computing approach. The performance assessment indices corroborated the superiority of the Pearson VII universal kernel function-based GPR model (GPR-PUKF) model (CC = 0.9997, MAE = 64.8091 kg/h, RMSE = 248.7584 kg/h, SI = 0.0237, and NSE = 0.9993 for the testing dataset) over other data-intelligent models in predicting the gas mass flow rate. In addition, statistical results revealed that the predictions of the RF method were better than those of the GEP- and RT-based models, but the GEP approach showed the lowest performance among all applied models. Although the CC values of all models were satisfactory (>0.993), the percentile deviation of GPR model (1.7325%) from the actual values showed competitive lower values, indicating its superior performance than other models (GEP = 15.1436%, RF = 6.5403%, RT = 9.5576%, and SVM = 3.2107%). This study highlighted the significance of employing advanced soft-computing approaches in determining the mass flow rate of natural gas, a vital source of energy, as well as its value to the gas sector.