Random Forest Regression Research Articles

Comparative Analysis of Machine Learning Models for Daytime Power Generation Prediction

This paper proposes to evaluate how different machine learning techniques can be used to predict daytime power generation based on the ”Daily Power Generation Data” dataset. As a result of six models, which contain Random Forest Regressor, Decision Tree Regressor, Nearest Neighbors, Linear Regression, MLP Regressor, and SVR, a clear understanding has been accomplished by assessing the performance using multiple metrics. First of all, the Random Forest Regressor turned out to be the best in terms of the Mean Squared Error (MSE) of 3.57×10−6, which was the lowest among the three ML models. The introduction of the paper highlights the role of precise planning of the power market and the consecutive sections describing the topic mathematically. The table below, with a total list of performance issues, explains why the Random Forest Regressor is the superior full-proof model using the lowest MSE, highest explained variance, and great resistance to outlying samples. The paper thus gave various useful approval criteria that, to a great extent, we can choose the best model out of them because the Random Forest Regressor was in a position to get the highest performance metrics.

A multivariate analysis to explain residue errors in pathogen concentration in wastewater-based epidemiology.

With the beginning of the COVID-19 pandemic, wastewater-based epidemiology (WBE), which according to Larsen et al. (2021), describes the science of linking pathogens and chemicals found in wastewater to population-level health, received an enormous boost worldwide. The basic procedure in WBE is to analyse pathogen concentrations and to relate these measurements to cases from clinical data. This prediction of cases is subject to large errors, due to various factors such as dilution effects, decay or wastewater matrix and inhibitors. In this study we used different models to identify the most important, what we call, wastewater-based epidemiologically relevant parameters (WBERP) to describe these errors. We used linear regression and random forest regression as base models for predicting cases and random forest regression also to analyse the importance of different WBERP. Two catchments, one with a large proportion of combined sewers and one with separate sewers, served as study areas. Our results show that the most important information to be included in any model are the variants of concern (VOCs), a time-variable parameter. The performance for both catchments is improved by ~30% in terms of root mean square error when the VOCs are used as additional information. For practical applications, this is a real drawback as it means that every time a new pathogen variant becomes dominant, we need to know the specific behaviour of the variant in the wastewater and its detection in order to interpret the WBE data correctly. This limits the predictive capabilities of such systems, perhaps not in terms of dynamics but for quantitative statements. The addition of other physicochemical parameters and faecal markers only marginally improved the results. Furthermore, there were differences in the importance of the parameters between the catchments, which limits the generalisability of the conclusions. The results show that more complex wastewater matrices (high proportion of combined sewer system) influence the relationship between pathogen concentration and medical cases more than those of less complex wastewater matrices (separate sewer system).

Data-DrivenWeather Prediction: Integrating Deep Learning and Ensemble Models for Robust Weather Forecasting

Accurate weather forecasting is critical for sectors like agriculture, transportation, disaster management, and public safety. This paper presents a comprehensive methodology integrating traditional machine learning models, deep learning techniques, and ensemble learning approaches to enhance the precision and reliability of weather predictions. Using a combination of four datasets—two for classification and two for regression—the study evaluates various machine learning models such as Decision Trees, Support Vector Machines, and KNearest Neighbors, alongside ensemble methods like Bagging and AdaBoost. Additionally, deep learning models, particularly the Multilayer Perceptron (MLP), are employed to handle complex weather patterns. The Random Forest Regressor and Bagging Regressor consistently outperformed other models in terms of accuracy, precision, and F1-score. By comparing the performance of these models across different weather datasets, this research demonstrates the effectiveness of cross-validation and the importance of optimizing hyperparameters. The findings contribute valuable insights into enhancing the robustness and efficiency of weather forecasting systems, with potential applications in environmental monitoring, event planning, and climate change analysis.The findings indicate that Random Forest Regression consistently outperformed the other machine learning models evaluated. For ensemble learning, the Bagging Regressor was the top performer. In deep learning, the Multilayer Perceptron without cross-validation delivered outstanding performance. For the classification datasets, Random Forest achieved the highest accuracy, precision, and F-score. Our study also highlights the importance of cross-validation to prevent overfitting and ensure model robustness, as well as the impact of class imbalance on overall performance metrics.

A hybrid hw-rfr forecasting model: case of moroccan pharmaceutical sector

Sales forecasting is an essential element of effective supply chain management, particularly in the pharmaceutical sector where continuous availability of drugs is crucial. This article examines sales forecasts for fluoxetine, an antidepressant available on the Moroccan market under six trade names and 14 different forms. The main objective of this study is to compare the effectiveness of four forecasting models, namely Prophet Facebook, ARIMA, GRU and Holt-Winters through their accuracy, and to propose a hybrid model that will contribute to improving the accuracy of demand forecasts. Each model was applied individually to predict future sales, and evaluated using MAPE, MAE and RMSE metrics. Next, a hybrid model, integrating Holt-Winters and Random Forest Regressor methods, was developed to leverage the robustness of traditional models while improving predictive performance through machine learning techniques. The results of the study show that traditional models, such as ARIMA and Holt-Winters, offer a solid basis for sales forecasting. However, the hybrid HW-RFR (Holt-Winters Random Forest Regressor) model stands out for a significant improvement in forecast accuracy, demonstrating great robustness to fluctuations in fluoxetine demand. This article highlights the potential of hybrid models for forecasting pharmaceutical sales. The improved forecast accuracy achieved with the HW-RFR model provides stakeholders with more reliable information, enabling them to make informed decisions to optimize pharmaceutical supply chain management

Toxicity assessment and i-QSTTR analysis of ionic liquids on D. magna, D. rerio, and R. subcapitata.

The study aimed to assess the impacts of ionic liquids (ILs) as innovative alternatives to traditional organic solvents on aquatic environments and human health. Five machine learning methods, including multiple linear regression (MLR), partial least squares regression (PLS), random forest regression (RF), support vector regression (SVR), and extreme gradient boosting (XGBoost), were used to construct the prediction models of the toxicity of ILs to D. magna, D. rerio, and R. subcapitata. Rigorous validation criteria were implemented to evaluate the robustness and predictive accuracy of these models. The results indicated SVR and XGBoost models demonstrated superior predictive performance. In addition, for these three species of D. magna, D. rerio, and R. subcapitata. The six interspecies quantitative structure-toxicity-toxicity (i-QSTTR) models were developed to analyze the cross-species toxicity responses of ILs. The results revealed a strong interspecies correlation in the toxicity of ILs to D. magna and D. rerio, as well as between D. rerio and R. subcapitata. However, the correlation between D. magna and R. subcapitata was weaker, indicating significant differences in the responses of ILs toxicity between these two aquatic species. This study not only filled the data gap in the biotoxicity of ILs but also provided an important theoretical basis for their safe application.

Revealing the driving factors of urban wetland park cooling effects using Random Forest regression and SHAP algorithm

Revealing the driving factors of urban wetland park cooling effects using Random Forest regression and SHAP algorithm

Prediction of impurity concentrations in AlN single crystals by absorption at 230 nm using random forest regression

A sample matrix of C, O, and Si doping in PVT-AlN is presented, and links between concentration and absorption coefficient are demonstrated. A trained random forest model offers a promising approach for the prediction of C, O, and Si content.

Regionally differentiated responses of chlorophyll-a concentrations to reduced human activity during COVID-19 lockdown in the San Francisco Bay area.

The COVID-19 lockdown created a unique opportunity to study the impact of reduced human activities on water quality. This study aimed to explore how changes in human activities, specifically reduced traffic emissions, influenced water quality in the San Francisco Bay Area from 2019 to 2021. Using chlorophyll-a (Chl-a) concentration as an indicator of water quality and NO₂ concentration as a proxy for traffic emissions, we analyzed the effects of reduced emissions on water quality across different regions of the Bay. This study combined traffic flow, satellite remote sensing, and meteorological data. Pearson correlation analysis was used to assess the time-lag effects between NO₂ and Chl-a concentrations. Additionally, a random forest regression model was applied to analyze the drivers of Chl-a concentrations across different regions. The findings revealed that NO₂'s impact on Chl-a exhibited significant spatial heterogeneity: positive correlations increased in the Upper Bay region, a general positive correlation was observed in the Central Bay region, the Coastal Zone shifted from positive to negative correlations, and there was no significant correlation in the Lower South Bay region. The random forest regression analysis demonstrated that the main drivers of Chl-a varied by region. In Upper Bay and Central Bay, NO₂ concentration was the most significant driver of Chl-a, whereas temperature was the primary influence in the Coastal Zone and Lower South Bay. These findings offer insights into the spatially variable impacts of traffic emissions on water quality and suggest strategies for targeted management of water quality in urban estuarine environments.

A Novel Hybrid Machine Learning Framework for Wind Speed Prediction

The growing urgency of environmental challenges and the depletion of fossil fuels have accelerated the search for sustainable and renewable energy sources. Wind energy, for example, is an important source of green electricity. However, using wind power is challenging due to the variability and unpredictability of wind patterns. Consequently, the ability to predict wind power in advance is crucial. The integration of artificial intelligence within the renewable energy sector could provide a viable solution to this challenge. In this study, we investigate the potential of machine learning to improve wind power forecasting by conducting a comparison of three regression models: K-Nearest Neighbor regression, Random Forest regression, and Support Vector regression. These models are combined with a feature selection technique to forecast wind power. Additionally, we propose a novel hybrid approach that combines these machine learning models with Multiple Linear Regression to address the complexities of wind energy forecasting. The performance of the models is evaluated using the R² score, Mean Absolute Error, and Root Mean Squared Error. The dataset for this study was generated from a numerical simulation conducted at a location with a latitude of 22.55° N and a longitude of -14.33° E. The findings demonstrate that the proposed hybrid model outperforms the individual machine learning models in terms of prediction accuracy. This study provides a solid foundation for future research and development in wind energy forecasting.

Unlocking Mechanisms for Soil Organic Matter Accumulation: Carbon Use Efficiency and Microbial Necromass as the Keys.

Soil microorganisms transform plant-derived C (carbon) into particulate organic C (POC) and mineral-associated C (MAOC) pools. While microbial carbon use efficiency (CUE) is widely recognized in current biogeochemical models as a key predictor of soil organic carbon (SOC) storage, large-scale empirical evidence is limited. In this study, we proposed and experimentally tested two predictors of POC and MAOC pool formation: microbial necromass (using amino sugars as a proxy) and CUE (by 18O-H2O approach). Soil sampling (0-10 and 10-20 cm depth) was conducted along a climatic transect of 900 km on the Loess Plateau, including cropland, grassland, shrubland, and forest ecosystems, to ensure the homogeneous soil parent material. We found the highest POC and MAOC accumulation occurred in zones of MAT between 5°C and 10°C or MAP between 300 and 500 mm. Microbial necromass C was more positively related to POC than MAOC (p < 0.05), suggesting that microbial residues may improve POC pool more strongly compared to MAOC pool. Random forest and linear regression analyses showed that POC increased with fungal necromass C, whereas bacterial necromass C drove MAOC. Microbial CUE was coupled with MAOC (p < 0.05) but decoupled with POC and SOC (p > 0.05). The POC have faster turnover rate due to the lack of clay protection, which may lead to the rapid turnover of microbial necromass and thus their decoupling from the CUE. In this sense, the SOC accumulation is driven by microbial necromass, whereas CUE explains MAOC dynamics. Our findings highlight the insufficiency of relying solely on microbial carbon use efficiency (CUE) to predict bulk SOC storage. Instead, we propose that CUE and microbial necromass should be used together to explain SOC dynamics, each influencing distinct C pools.

Reconstructing urban vegetation evolution in China using multimodal deep learning and 30-years Landsat archive

To retrospect the impact of urbanization on vegetation, consistent and reliable maps that provide urban vegetation estimates are required. In this study, we make the first attempt to employ deep learning for urban vegetation mapping of 2120 cities in China using a 30-years Landsat archive (UV-30). We present a multimodal deep learning (MDL) model to combine features from 5 categories, namely reflective bands, remote sensing indices, texture variables, topographical variables, and time labels. We normalized elevation data locally to remove the inter-city differences, while retaining topographic details. We used time labels to account for spectral variability over time and Landsat sensors. The performance of the MDL models was compared with a vanilla deep neural network, random forest, and support vector regression. Results indicated that the fusion of different categories of features improved the prediction accuracy. The MDL performed best in most validation cities, and our results were superior to the fractional maps aggregated from high-resolution datasets in homogeneous urban landscapes. We further analyzed urban vegetation dynamics across the urban core-new town-outskirt gradients of different geographical regions and city sizes. We found a gradually diminishing de-greening trend in urban China, especially in urban cores and new towns. Nevertheless, we observed a more pronounced degradation of urban vegetation in the south regions of China compared to the north, and smaller cities show no evident urban vegetation increase. The suggested framework is valuable for urban vegetation assessments and facilitates a thorough understanding and management of urban ecosystems.

Prediction of subpectoral direct-to-implant breast reconstruction failure based on random forest and logistic regression algorithms: a multicenter study in Chinese population

BackgroundFew studies have been conducted on direct-to-implant (DTI) breast reconstruction failure, and consistent conclusions are lacking. Thus, this study aimed to comprehensively analyze the risk factors of reconstruction failure. MethodsPatients who underwent DTI breast reconstruction after mastectomy at a single center between July 18, 2014, and January 13, 2020, were retrospectively included in this study. Two algorithms, random forest and logistic regression, were employed to construct models that analyzed the complications and risk factors of reconstruction failure. Subsequently, a multicenter external validation was performed for both models. ResultsThere were 538 patients in the model construction group and 91 patients in the multicenter external validation group, with 23 and five reconstruction failure outcomes, respectively. Random forest analysis revealed that infection and wound dehiscence were the most significant factors leading to reconstruction failure. Multivariate logistic regression analysis indicated that body mass index (BMI), infection, and wound dehiscence were correlated with reconstruction failure. The risk of failure was 3.35% higher in overweight (BMI > 24Kg/m2) patients, 9.6% higher in patients with infection, and 42.5% higher in patients with wound dehiscence than that in the control group. The internal validation receiver operating characteristic (ROC) value for the random forest model was 0.990, whereas the external validation ROC was 0.736. The internal and external validation ROCs values for the logistic regression model were 0.995 and 0.826, respectively. ConclusionWound dehiscence and infection were the most significant risk factors for DTI breast reconstruction failure, and preoperative weight control was also important.

Cellular senescence-associated genes in rheumatoid arthritis: Identification and functional analysis.

Rheumatoid arthritis (RA), a long-term autoinflammatory condition causing joint damage and deformities, involves a multifaceted pathogenesis with genetic, epigenetic, and immune factors, including early immune aging. However, its precise cause remains elusive. Cellular senescence, a hallmark of aging marked by a permanent halt in cell division due to damage and stress, is crucial in aging and related diseases. In our study, we analyzed RA microarray data from the Gene Expression Omnibus (GEO) and focused on cellular senescence genes from the CellAge database. We started by selecting five RA datasets from GEO. Next, we pinpointed 29 differentially expressed genes (DEGs) linked to cellular senescence in RA, aligning them with genes from CellAge. We explored the roles of these DEGs in cellular senescence through Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis. We then pinpointed three key genes (DHX9, CYR61, and ITGB) using random forest and LASSO Cox regression machine learning techniques. An integrated diagnostic model was created using these genes. We also examined the variance in immune cell infiltration and immune checkpoint gene expression between RA and normal samples. Our methodology's predictive accuracy was confirmed in external validation cohorts. Subsequently, RA samples were classified into three distinct subgroups based on the cellular senescence-associated DEGs, and we compared their immune landscapes. Our findings reveal a significant impact of cellular senescence-related DEGs on immune cell infiltration in RA samples. Hence, a deeper understanding of cellular senescence in RA could offer new perspectives for diagnosis and treatment.

Random Forest Regressor for Predicting Sensory Texture of Emotional Designed Packaging Films

Random Forest Regressor for Predicting Sensory Texture of Emotional Designed Packaging Films

Mechanism- and data-driven algorithms of electrical energy consumption accounting and prediction for medium and heavy plate rolling

&lt;p&gt;Energy consumption accounting and prediction in the medium and thick plate rolling process are crucial for controlling costs, improving production efficiency, optimizing equipment management, and enhancing the market competitiveness of enterprises. Starting from the perspective of integrating process mechanism and industrial big data, we overcame the difficulties brought by complex and highly nonlinear coupling of process variables, proposed a rolling power consumption accounting algorithm based on time slicing method, and gave a calculation method for the additional power consumption of the main motor for rough rolling and finishing rolling (auxiliary system power consumption, power loss, main motor power consumption deviation); with the help of SIMS model, forward recursion, and reverse recursion pass rolling force estimation strategies are proposed, and the rated power consumption of the main motor was predicted. Furthermore, a random forest regression model of additional power consumption based on data was established, and then a prediction algorithm for the comprehensive power consumption of billet rolling was given. Experiments showed the effectiveness of the proposed method.&lt;/p&gt;

Efficient Physical Truncation of Low-Frequency ATEM Problems in Specific Geometries by Using Random Forest Regression Based PMM Model

Efficient Physical Truncation of Low-Frequency ATEM Problems in Specific Geometries by Using Random Forest Regression Based PMM Model

Mechanism- and data-driven algorithms of electrical energy consumption accounting and prediction for medium and heavy plate rolling

&lt;p&gt;Energy consumption accounting and prediction in the medium and thick plate rolling process are crucial for controlling costs, improving production efficiency, optimizing equipment management, and enhancing the market competitiveness of enterprises. Starting from the perspective of integrating process mechanism and industrial big data, we overcame the difficulties brought by complex and highly nonlinear coupling of process variables, proposed a rolling power consumption accounting algorithm based on time slicing method, and gave a calculation method for the additional power consumption of the main motor for rough rolling and finishing rolling (auxiliary system power consumption, power loss, main motor power consumption deviation); with the help of SIMS model, forward recursion, and reverse recursion pass rolling force estimation strategies are proposed, and the rated power consumption of the main motor was predicted. Furthermore, a random forest regression model of additional power consumption based on data was established, and then a prediction algorithm for the comprehensive power consumption of billet rolling was given. Experiments showed the effectiveness of the proposed method.&lt;/p&gt;

Developing a Prototype Machine Learning Model to Predict Quality of Life Measures in People Living With HIV.

In the realm of Evidence-Based Medicine, introduced by Gordon Guyatt in the early 1990s, the integration of machine learning technologies marks a significant advancement towards more objective, evidence-driven healthcare. Evidence-Based Medicine principles focus on using the best available scientific evidence for clinical decision-making, enhancing healthcare quality and consistency by integrating this evidence with clinician expertise and patient values. Patient-Reported Outcome Measures (PROMs) and Patient-Reported Experience Measures (PREMs) have become essential in evaluating the broader impacts of treatments, especially for chronic conditions like HIV, reflecting patient health and well-being comprehensively. The study aims to leverage Machine Learning (ML) technologies to predict health outcomes from PROMs/PREMs data, focusing on people living with HIV. Our research utilizes a ML Random Forest Regression to analyze PROMs/PREMs data collected from over 1200 people living with HIV through the NAVETA telemedicine system. The findings demonstrate the potential of ML algorithms to provide precise and consistent predictions of health outcomes, indicating high reliability and effectiveness in clinical settings. Notably, our ALGOPROMIA ML model achieved the highest predictive accuracy for questionnaires such as MOS30 VIH (Adj. R² = 0.984), ESTAR (Adj. R² = 0.963), and BERGER (Adj. R² = 0.936). Moderate performance was observed for the P3CEQ (Adj. R² = 0.753) and TSQM (Adj. R² = 0.698), reflecting variability in model accuracy across instruments. Additionally, the model demonstrated strong reliability in maintaining standardized prediction errors below 0.2 for most instruments, with probabilities of achieving this threshold being 96.43% for WHOQoL HIV Bref and 88.44% for ESTAR, while lower probabilities were observed for TSQM (44%) and WRFQ (51%). The results from our machine learning algorithms are promising for predicting PROMs and PREMs in AIDS settings. This work highlights how integrating ML technologies can enhance clinical pharmaceutical decision-making and support personalized treatment strategies within a multidisciplinary integration framework. Furthermore, leveraging platforms like NAVETA for deploying these models presents a scalable approach to implementation, fostering patient-centered, value-based care.

Random forest regression feature importance for climate impact pathway detection

Random forest regression feature importance for climate impact pathway detection

The RADARSAT Constellation Mission for Soil Moisture Retrieval of Bare Soil by Compact Polarimetry and Random Forest Regression

The RADARSAT Constellation Mission (RCM) performance evaluation is currently in progress for core Synthetic Aperture Radar (SAR) applications. This study aims to investigate the retrieval of Soil Moisture Content (SMC) in bare soil with RCM compact polarimetry and Random Forest Regression (RFR). The focus is on RH (right circular transmit and linear horizontal receive signal) and RV (right circular transmit and linear vertical receive signal) backscattering, which are the primary RCM Compact Polarimetric (CP) products. SMC retrieval is pursued over a wide range of radar incidence angles. Then, an attempt is made to retrieve SMC at higher radar incidence angles only. Furthermore, soil moisture maps are produced and used for analyzing the captured soil moisture variability. CP SAR images acquired with the RCM SC30MCP mode over three Canadian experimental sites are considered in our study. The sites are equipped with calibrated Real-Time In-Situ Soil Monitoring for Agriculture (RISMA) stations. A RFR retrieval algorithm was able to predict SMC with a correlation of 0.75 when compared to in-situ soil moisture measurements. A Root Mean Square Error (RMSE) = 5.9%, a bias = −1.5%, and an unbiased RMSE (ubRMSE) = 5.7% are achieved. A degradation in performance is reported for SMC retrieval under higher radar incidence angles. Results of our study indicate promising performance for capturing near-surface soil moisture variability under bare soil conditions.