Accuracy Of Random Forest Model Research Articles

Lost in the Forest: Encoding categorical variables and the absent levels problem

AbstractLevels of a predictor variable that are absent when a classification tree is grown can not be subject to an explicit splitting rule. This is an issue if these absent levels are present in a new observation for prediction. To date, there remains no satisfactory solution for absent levels in random forest models. Unlike missing data, absent levels are fully observed and known. Ordinal encoding of predictors allows absent levels to be integrated and used for prediction. Using a case study on source attribution of Campylobacter species using whole genome sequencing (WGS) data as predictors, we examine how target-agnostic versus target-based encoding of predictor variables with absent levels affects the accuracy of random forest models. We show that a target-based encoding approach using class probabilities, with absent levels designated the highest rank, is systematically biased, and that this bias is resolved by encoding absent levels according to the a priori hypothesis of equal class probability. We present a novel method of ordinal encoding predictors via principal coordinates analysis (PCO) which capitalizes on the similarity between pairs of predictor levels. Absent levels are encoded according to their similarity to each of the other levels in the training data. We show that the PCO-encoding method performs at least as well as the target-based approach and is not biased.

Comparison and analysis of the effect of various machine learning algorithms on cancer prediction

Cancer is a serious disease that can have a big impact on the physical and mental health of patients. The incidence and mortality rate of cancer are increasing globally. Therefore, predicting the occurrence and treatment effect of cancer has become a hot research topic in the medical field. Machine learning algorithms can use vast amounts of data and algorithmic models to predict aspects of cancer occurrence, progression, and treatment effectiveness. This algorithm can learn and find patterns from large amounts of medical data, thereby improving the diagnosis and treatment of cancer. In this paper, naive Bayes model, logistic regression model, random forest model, KNN model and Ridge regression model were used to predict the data of cancer patients. The accuracy of naive Bayes model and logistic regression model is the highest, reaching 95%. The accuracy of random forest model and KNN model is 94%. The worst performing model was Ridge regression, with an accuracy of 93%. Naive Bayes algorithm and logistic regression algorithm perform well in solving binary classification problems because they are both probability-based classification algorithms. In this problem, the number of features is large, so the ridge regression algorithm can effectively process the data, but due to the small amount of data, it is easy to appear underfitting, so the performance is poor. In this paper, a variety of machine learning algorithms are used to predict cancer patients, and the accuracy, accuracy, recall rate and f1 parameters of each model are calculated and compared, which provides a basis for subsequent research.

Enhancing bank marketing strategies with ensemble learning: Empirical analysis.

In order to enhance market share and competitiveness, large banks are increasingly focusing on promoting marketing strategies. However, the traditional bank marketing strategy often leads to the homogenization of customer demand, making it challenging to distinguish among various products. To address this issue, this paper presents a customer demand learning model based on financial datasets and optimizes the distribution model of bank big data channels through induction to rectify the imbalance in bank customer transaction data. By comparing the prediction models of random forest model and support vector machine (SVM), this paper analyzes the ability of the prediction model based on ensemble learning to significantly enhance the market segmentation of e-commerce banks. The empirical results reveal that the accuracy of random forest model reaches 92%, while the accuracy of SVM model reaches 87%. This indicates that the ensemble learning model has higher accuracy and forecasting ability than the single model. It enables the bank marketing system to implement targeted marketing, effectively maintain the relationship between customers and banks, and significantly improve the success probability of product marketing. Meanwhile, the marketing model based on ensemble learning has achieved a sales growth rate of 20% and improved customer satisfaction by 30%. This demonstrates that the implementation of the ensemble learning model has also significantly elevated the overall marketing level of bank e-commerce services. Therefore, this paper offers valuable academic guidance for bank marketing decision-making and holds important academic and practical significance in predicting bank customer demand and optimizing product marketing strategy.

Remote Sensing Image Segmentation of Wetlands in Macau Based on Machine Learning

Wetlands are essential components of our natural ecosystems, performing critical functions that contribute to biodiversity conservation, water purification, climate regulation, and shoreline protection. Remote sensing classification of wetlands, wetland resource surveys and other wetland research is booming, for the development of environmental protection policies continue to provide reliable data. Macau is a coastal city with rich wetland resources, but there is no relevant research on wetlands in Macau. In this paper, we use the high spatial resolution remote sensing data from Gaofen-2 satellite to extract the wetlands of Macau SAR to fill the gap of related research. This paper used two distinct machine learning classifiers, including Random Forest (RF) and Support Vector Machine (SVM), to analyse the high spatial resolution remote sensing data of GF-2 satellite in order to extract wetlands in Macau Special Region Area. In this research, the accuracy of RF model is much higher than that of SVM model, comparing the indexes of overall accuracy (OA) and Kappa coefficient. After analysing the resultant images generated after segmentation, the image segmentation result of RF model is better than that of SVM model. As for the RF model, the overall accuracy (OA) is 0.9941 and Kappa coefficient is 0.9899.

ReliefF-RF model application in communication security risk detection of distribution network

Abstract The construction of an intelligent power grid system has promoted innovation and development in the power grid industry, but the distribution network covers a wide range and has more access data, and the network attack risk is the focus of attention. In this regard, the stochastic forest model is introduced to build the distribution network risk detection model based on the software-defined network. The first is to study the power grid system based on a software-defined network, realize the analysis and extraction of the characteristics of abnormal power grid engineering data, and realize the diagnosis of risk data through a random forest model. At the same time, considering the long modeling time of the random forest model and the low classification accuracy in unbalanced samples, the feature selection model is introduced to optimize the random forest model, and the sampling weight function is set to improve the insufficient sampling accuracy of random forest model in small samples. The model classification effect is evaluated during the model performance test. The proposed model is the most accurate of the three data sets. In the three data classification tests of Satimage, Senbased, and Cleveland, the accuracy of MICRO Average is respectively 0.886, 0.986 and 0.856. At the same time, the proposed model is superior to other models in terms of training time and variance stability test. When the data set is 18000, the model accuracy is 0.968, which is better than the other two models. The research content has important reference value for maintaining communication security and improving system stability in distribution networks.

Relationship between plant species diversity and aboveground biomass in alpine grasslands on the Qinghai–Tibet Plateau: Spatial patterns and the factors driving them

Alpine grasslands are important ecosystems on the Qinghai–Tibet Plateau and are extremely sensitive to climate change. However, the spatial responses of plant species diversity and biomass in alpine grasslands to environmental factors under the background of global climate change have not been thoroughly characterized. In this study, a random forest model was constructed using grassland ground monitoring data with satellite remote sensing data and environmental variables to characterize the plant species diversity and aboveground biomass of grasslands in the Three-River Headwaters Region within the Qinghai–Tibet Plateau and analyze spatial variation in the relationship between the plant species diversity and aboveground biomass and their driving factors. The results show that (1) the selection of characteristic variables can effectively improve the accuracy of random forest models. The stepwise regression variable selection method was the most effective approach, with an R2 of 0.60 for the plant species diversity prediction model and 0.55 for the aboveground biomass prediction model, (2) The spatial distribution patterns of the plant species diversity and aboveground biomass in the study area were similar, they were both high in the southeast and low in the northwest and gradually decreased from east to west. The relationship between the plant species diversity and aboveground biomass varied spatially, they were mostly positively correlated (67.63%), but they were negatively correlated in areas with low and high values of plant species diversity and aboveground biomass, and (3) Analysis with geodetector revealed that longitude, average annual precipitation, and elevation were the main factors driving variation in the plant species diversity and aboveground biomass relationship. We characterized plant species diversity and aboveground biomass, as well as their spatial relationships, over a large spatial scale. Our data will aid biodiversity monitoring and grassland conservation management, as well as future studies aimed at clarifying the relationship between biodiversity and ecosystem functions.

Superiority of two-dimensional correlation spectroscopy combined with ResNet in species identification of bolete

Food safety is an important topic of social concern. There are many species of bolete on the market, and shoddy products often occur. This phenomenon not only disrupts the market order, infringes on the rights and interests of consumers, and even endangers the lives of consumers. Therefore, it is of great significance to find a comprehensive, efficient and modern technology to identify and evaluate it in order to ensure its quality and safety. Spectroscopy analysis is fast, nondestructive and green. In our research, a fast and reliable bolete species identification method was established by combining Fourier transform near infrared spectroscopy (FT-NIR) with random forest (RF) method. In order to adapt to the development of the times, a practical method beyond traditional spectral analysis is established. Therefore, we establish a residual convolution neural network model (ResNet). The results show that the classification accuracy of the model is low and the out of bag error (OOB) error is high. After data preprocessing, the accuracy of RF model can be significantly improved. ResNet model has absolute advantages in bolete species identification. It is hardly affected by factors such as data type and sample size. Its overall recognition ability is obviously better than RF model. By comparing the identification accuracy and market application prospect of the two models, this research believes that ResNet model is more suitable for the identification of bolete species. In addition, the method can also be extended to the further study of other food, medicinal plants and agricultural products.

Comparison of nomogram and machine‐learning methods for predicting the survival of non‐small cell lung cancer patients

AbstractBackgroundMost patients with advanced non‐small cell lung cancer (NSCLC) have a poor prognosis. Predicting overall survival using clinical data would benefit cancer patients by allowing providers to design an optimum treatment plan. We compared the performance of nomograms with machine‐learning models at predicting the overall survival of NSCLC patients. This comparison benefits the development and selection of models during the clinical decision‐making process for NSCLC patients.MethodsMultiple machine‐learning models were used in a retrospective cohort of 6586 patients. First, we modeled and validated a nomogram to predict the overall survival of NSCLC patients. Subsequently, five machine‐learning models (logistic regression, random forest, XGBoost, decision tree, and light gradient boosting machine) were used to predict survival status. Next, we evaluated the performance of the models. Finally, the machine‐learning model with the highest accuracy was chosen for comparison with the nomogram at predicting survival status by observing a novel performance measure: time‐dependent prediction accuracy.ResultsAmong the five machine‐learning models, the accuracy of random forest model outperformed the others. Compared with the nomogram for time‐dependent prediction accuracy with a follow‐up time ranging from 12 to 60 months, the prediction accuracies of both the nomogram and machine‐learning models changed as time varied. The nomogram reached a maximum prediction accuracy of 0.85 in the 60th month, and the random forest algorithm reached a maximum prediction accuracy of 0.74 in the 13th month.ConclusionsOverall, the nomogram provided more reliable prognostic assessments of NSCLC patients than machine‐learning models over our observation period. Although machine‐learning methods have been widely adopted for predicting clinical prognoses in recent studies, the conventional nomogram was competitive. In real clinical applications, a comprehensive model that combines these two methods may demonstrate superior capabilities.

Remote estimates of soil organic carbon using multi-temporal synthetic images and the probability hybrid model

Soil organic carbon (SOC) plays a key role in soil function, ecosystem services, and the global carbon cycle. Digital SOC mapping is essential for agricultural production management. Digital SOC mapping based on multi-source remote sensing data has been integrated well into prediction models and methodological approaches on different mapping scales. However, the mixed use of synthetic images and the application of the hybrid model considering the soil classification probability are rare. Here, we propose a probability hybrid model to estimate and map SOC content and distribution. Multi-temporal synthetic images were used to build the probability hybrid model. One hundred forty topsoil samples were collected in Suihua, a city in a typical black soil region in China. Cloud-free Sentinel-2 images were acquired from the bare soil periods between 2018 and 2021. Minimum, maximum, mean, and median synthetic images were calculated using single images from the same period. The random forest and support vector machine models were used for discriminating soil class and calculating soil classification probability, and then random forest regression model was applied to SOC mapping. Soil class mapping and classification probability were performed for Phaeozems, Chernozems, and Cambisols of the World Reference Base for Soil Resources (WRB) based on soil texture, climatic factors, and the normalized differential vegetation index (NDVI). Based on soil class mapping results, global models using single temporal images and multi-temporal images, the hybrid model using multi-temporal images for the three soil classes as well as a probability hybrid model using multi-temporal images were built and compared their SOC predictions. Our study showed that (1) Phaeozems, Chernozems and Cambisols could be classified accurately, and the overall validation accuracy of random forest model was 91.67%. (2) The correlations of SOC with bands and band indices improved using multi-temporal images, and the use of mixed synthetic images was better than the use of only one synthetic image. (3) The hybrid model performed far better than the global models, and the probability hybrid model led to the highest prediction accuracy, a validation R2 of 0.77 and an RMSE of 2.30 g kg−1. (4) The probability hybrid model was more accurate than the original hybrid model for digital SOC mapping, and the SOC distribution in boundary regions was smoother and more continuous. Our results suggest that the probability hybrid model has a large potential for SOC prediction and mapping.

Retrieval of Daytime Surface Upward Longwave Radiation Under All-Sky Conditions With Remote Sensing and Meteorological Reanalysis Data

Surface upward longwave radiation (SULR) is a key parameter that regulates surface radiation budget balance and matter-energy exchange. However, the state-of-the-art SULR retrieval methods based on remotely sensed data are only effective under clear skies, which mean that existing methods are unable to generate spatiotemporal continuous SULR product at regional or global scale. Herein, taking the advantage of long-pending abundant ground-based radiation observations, satellite products and meteorological reanalysis data, a data-driven random forest (RF) method is proposed to retrieve the instantaneous SULR under all-sky conditions. Based on spectral samples of different surface types and simulation results from the moderate resolution atmospheric transmission (MODTRAN), spectral transformation is carried out to transform SULR of various measured domains into the defined 4~100 μm domain at first. SULR and surface downward shortwave radiation (SDSR) observations from seven stations of the Surface Radiation Budget Network (SURFRAD) and nine stations of the Baseline Surface Radiation Network (BSRN) are used in model’s training and testing procedures, and the RF model achieves a high accuracy with the root-mean-square error (RMSE) of 10.45 W/m² on test set. In model evaluation, ground measurements from 14 stations of FLUXNET have been used, and the overall RMSE is 18.40 W/m². In the actual application process, SDSR is estimated by remotely sensed data of Meteosat Second Generation (MSG). The accuracy of RF model has been validated with the observations from five stations of BSRN in 2021, and RMSEs are 17.00, 10.94, 12.17, 27.89 and 12.54 W/m², respectively. Validation result shows that the data-driven method is capable of estimating SULR under all-sky conditions with a high accuracy. Finally, sensitivity analysis has been carried out, and the established RF model keeps robust even though there are great uncertainties among input parameters.

Influence of plot and sample sizes on aboveground biomass estimations in plantation forests using very high resolution stereo satellite imagery

AbstractRegular biomass estimations for natural and plantation forests are important to support sustainable forestry and to calculate carbon-related statistics. The application of remote sensing data to estimate biomass of forests has been amply demonstrated but there is still space for increasing the efficiency of current approaches. Here, we investigated the influence of field plot and sample sizes on the accuracy of random forest models trained with information derived from Pléiades very high resolution (VHR) stereo images applied to plantation forests in an arid environment. We collected field data at 311 locations with three different plot area sizes (100, 300 and 500 m2). In two experiments, we demonstrate how plot and sample sizes influence the accuracy of biomass estimation models. In the first experiment, we compared model accuracies obtained with varying plot sizes but constant number of samples. In the second experiment, we fixed the total area to be sampled to account for the additional effort to collect large field plots. Our results for the first experiment show that model performance metrics Spearman’s r, RMSErel and RMSEnor improve from 0.61, 0.70 and 0.36 at a sample size of 24–0.79, 0.51 and 0.15 at a sample size of 192, respectively. In the second experiment, highest accuracies were obtained with a plot size of 100 m2 (most samples) with Spearman’s r = 0.77, RMSErel = 0.59 and RMSEnor = 0.15. Results from an analysis of variance type-II suggest that the overall most important factors to explain model performance metrics for our biomass models is sample size. Our results suggest no clear advantage for any plot size to reach accurate biomass estimates using VHR stereo imagery in plantations. This is an important finding, which partly contradicts the suggestions of earlier studies but requires validation for other forest types and remote sensing data types (e.g. LiDAR).

Experimental study and modelling discharge coefficient of trapezoidal and rectangular piano key weirs

Crest length is an important parameter in influencing the discharge handling capacity of a weir. Nonlinear weirs with longer crests are cost effective alternatives for those existing dam structures which are more susceptible to failure due to loss of storage capacity by reservoir silting problem, and insufficiency of the structure in evacuating the updated flow due to the limited space. Piano key weir is a type of nonlinear weir designed in the form of piano keys, over-hanged from both the upstream and the downstream with sloping floors founded on a base or footprint. These weirs can be easily placed over gravity dams due to smaller footprint than labyrinth weirs. The present study’s focus is on the comparative analysis of identical configurations of trapezoidal and rectangular piano key (PK) weirs. The importance of (crest length to width) L/W ratio and weir height (P) in affecting the discharge efficiency of both types of PK weirs is investigated in the experimental study. Furthermore, soft computing approaches are applied to the current data set obtained from both types of weirs by considering discharge coefficient (C_{text{d}}) as a function of dimensionless geometric variables of PK weirs. The modelling performance of random forest regression and M5 tree approach is tested in order to estimate the values of discharge coefficient. The results conclude higher predictive accuracy of random forest model over M5 tree model.