Performance Of Random Forest Model Research Articles

Descriptor-augmented machine learning for enzyme-chemical interaction predictions

Descriptors play a pivotal role in enzyme design for the greener synthesis of biochemicals, as they could characterize enzymes and chemicals from the physicochemical and evolutionary perspective. This study examined the effects of various descriptors on the performance of Random Forest model used for enzyme-chemical relationships prediction. We curated activity data of seven specific enzyme families from the literature and developed the pipeline for evaluation the machine learning model performance using 10-fold cross-validation. The influence of protein and chemical descriptors was assessed in three scenarios, which were predicting the activity of unknown relations between known enzymes and known chemicals (new relationship evaluation), predicting the activity of novel enzymes on known chemicals (new enzyme evaluation), and predicting the activity of new chemicals on known enzymes (new chemical evaluation). The results showed that protein descriptors significantly enhanced the classification performance of model on new enzyme evaluation in three out of the seven datasets with the greatest number of enzymes, whereas chemical descriptors appear no effect. A variety of sequence-based and structure-based protein descriptors were constructed, among which the esm-2 descriptor achieved the best results. Using enzyme families as labels showed that descriptors could cluster proteins well, which could explain the contributions of descriptors to the machine learning model. As a counterpart, in the new chemical evaluation, chemical descriptors made significant improvement in four out of the seven datasets, while protein descriptors appear no effect. We attempted to evaluate the generalization ability of the model by correlating the statistics of the datasets with the performance of the models. The results showed that datasets with higher sequence similarity were more likely to get better results in the new enzyme evaluation and datasets with more enzymes were more likely beneficial from the protein descriptor strategy. This work provides guidance for the development of machine learning models for specific enzyme families.

Comparative analysis of multi-source data for machine learning-based LAI estimation in Argania spinosa

In this study, we conducted a comprehensive assessment of Leaf Area Index (LAI) estimation using three distinct sources of satellite data: Sentinel-2 imagery, drone imagery (UAVs), and Mohammed VI satellite data. The main objective was to identify the most reliable and precise dataset for predicting LAI, with a focus on evaluating the performance of Random Forest models. For Sentinel-2 imagery, our Random Forest model achieves a robust R-squared (R2) value of 0.89, signifying a strong alignment between predicted and measured LAI values. The associated root-mean-square error (RMSE) is 0.4, indicating high predictive accuracy. In the context of UAVs, our Random Forest model excels, exhibiting an impressive R2 value of 0.93, highlighting a substantial correlation between predicted and measured LAI. The RMSE for drone imagery stands at 0.37, showcasing exceptional predictive accuracy. Finally, the Random Forest model trained on Mohammed VI satellite data yields an R2 value of 0.92, underlining its strong fit with measured LAI values. The RMSE for Mohammed VI imagery is 0.39, further underscoring the model's exceptional predictive accuracy. This comparative analysis underscores the importance of selecting the most suitable satellite data source for LAI estimation in Argania spinosa. UAV imagery emerges as the most accurate choice, closely followed by Mohammed VI Satellite and Sentinel-2 imagery. These findings offer valuable insights for effective monitoring of Argania spinosa and advancing sustainable land management practices in rural ecosystems.

Tree species diversity mapping from spaceborne optical images: The effects of spectral and spatial resolution

AbstractIncreasingly available spaceborne sensors provide unprecedented opportunities for large‐scale, timely and continuous tree species diversity (TSD) monitoring. However, given differences in spectral and spatial resolutions, the choice of sensor is not always straightforward. In this work, we investigated the effects of spatial and spectral resolutions for four spaceborne sensors (RapidEye, Landsat‐8, Sentinel‐2 and PlanetScope) on TSD mapping in an area of approximately 4000 km2 within the Black Forest, Germany. We employed a random forest (RF) regression model to predict Shannon–Wiener diversity based on seven types of spectral heterogeneity metrics (texture, coefficient of variation, Rao's Q, convex hull volume, spectral angle mapper, convex hull area and spectral species diversity) and a full survey dataset from 135 one‐ha sample plots. We compared the RF model's performance across sensors and spatial resolutions. Our results demonstrated that the Sentinel‐2‐based TSD model achieved the highest accuracy (mean R2: 0.477, mean root‐mean‐square error (RMSE): 0.274). The RapidEye‐based TSD model produced lower accuracy (mean R2: 0.346, mean RMSE: 0.303), but it was better than the PlanetScope‐ and Landsat‐based TSD models. The 10 m (for Sentinel‐2 and RapidEye) and 15 m (for PlanetScope) were the best spatial resolutions for predicting TSD. The NIR band was the most favourable spectral band for predicting TSD. Texture metrics and Rao's Q outperformed the other spectral heterogeneity metrics. Our results highlighted that spaceborne optical imagery (especially Sentinel‐2) can be successfully used for large‐scale TSD mapping but that the choice of sensors can significantly affect the resulting mapping accuracy in temperate montane forests.

Application of radiomics model based on lumbar computed tomographyin diagnosis of elderly osteoporosis.

A metabolic bone disease characterized by decreased bone formation and increased bone resorption is osteoporosis. It can cause pain and fracture of patients. The elderly are prone to osteoporosis and are more vulnerable to osteoporosis. In this study, radiomics are extracted from computed tomography (CT)images to screen osteoporosis in the elderly. Collect the plain scan CT images of lumbar spine, cut the region of interest of the image and extract radiomics features, use Lasso regression to screen variables and adjust complexity, use python language to model random forests, support vector machines, K nearest neighbor, and finally use receiver operating characteristic curve to evaluate the performance of the model, including precision, recall, accuracy and area under the curve (AUC). For the model, 14 radiolomics features were selected. The diagnosis performance of random forest model and support vector machine is good, all around 0.9. The AUC of K nearest neighbor model in training set and test set is 0.828 and 0.796, respectively. We selected the plain scan CT images of the elderly lumbar spine to build radiomics features model, which has good diagnostic performance and can be used as a tool to assist the diagnosis of osteoporosis in the elderly.

Early detection of pine shoot beetle attack using vertical profile of plant traits through UAV-based hyperspectral, thermal, and lidar data fusion

Pine shoot beetle (PSB) is one of the most damaging forest insects of Yunnan pine plantations in southwest China. However, the subtle symptoms of heterogeneous tree crowns make it difficult to accurately detect at early stage of PSB attack. Here, we evaluated the potential of a combination of plant traits (PTs) and vegetation indices (VIs) to distinguish different levels of tree damage by integrating hyperspectral, thermal imagery, and light detection and ranging (lidar) data based on unmanned airborne vehicle (UAV) systems. A voxelization method was used to fuse hyperspectral reflectance, temperature, and lidar point cloud data. Subsequently, PTs such as pigments (i.e., chlorophyll, carotenoid, and anthocyanin contents) were retrieved from a radiative transfer model inversion, and structural, fluorescence, and thermal traits, as well as VIs, were derived from the fusion data. We developed a novel analytical approach using random forest (RF) algorithm with predictors from different spatial distributions (i.e., horizontal directions, vertical layers, and vertical clusters) to compare the performance of tree severity classification. The results showed that the difference of both PTs and VIs in vertical layers between different severity levels are more than that in horizontal directions. The performance of RF model with predictors of the vertical layers (OA = 74 %, kappa = 0.65) was better than that using the predictive variables in horizontal directions (OA = 69 % and kappa = 0.60) of tree crowns. Using the vertical clustering features RF model increased the accuracy (OA = 78 % and kappa = 0.70), especially for slightly and moderately damaged trees, with improvements of 10 % and 12 %, respectively. Among all variables analyzed, chlorophylls were the most important predictor, followed by photochemical reflectance index and structural traits. Our work demonstrates the effectiveness of using fused UAV-based multi-sensor data for early detection of PSB attack, and can be applied other potential forest diseases and insect monitoring.

Impacts of sample ratio and size on the performance of random forest model to predict the potential distribution of snail habitats.

Few studies have considered the impacts of sample size and sample ratio of presence and absence points on the results of random forest (RF) testing. We applied this technique for the prediction of the spatial distribution of snail habitats based on a total of 15,000 sample points (5,000 presence samples and 10,000 control points). RF models were built using seven different sample ratios (1:1, 1:2, 1:3, 1:4, 2:1, 3:1, and 4:1) and the optimal ratio was identified via the Area Under the Curve (AUC) statistic. The impact of sample size was compared by RF models under the optimal ratio and the optimal sample size. When the sample size was small, the sampling ratios of 1:1, 1:2 and 1:3 were significantly better than the sample ratios of 4:1 and 3:1 at all four levels of sample sizes (p<0.01) and there was no significant difference among the ratios of 1:1, 1:2 and 1:3 (p>0.05). The sample ratio of 1:2 appeared to be optimal for a relatively large sample size with the lowest quartile deviation. In addition, increasing the sample size produced a higher AUC and a smaller slope and the most suitable sample size found in this study was 2400 (AUC=0.96). This study provides a feasible idea to select an appropriate sample size and sample ratio for ecological niche modelling (ENM) and also provides a scientific basis for the selection of samples to accurately identify and predict snail habitat distributions.

Spatiotemporal variation in soil methane uptake in a cool-temperate immature deciduous forest

Atmospheric methane (CH4) concentration has been increasing recently, contributing to global warming. As natural sinks, forest soils are expected to mitigate the atmospheric CH4 rise. However, it had been difficult to measure soil CH4 flux continuously and accurately because of the limited stability and precision of CH4 analyzers in the field. In this study, we measured hourly CH4 flux with plant roots (Root) and without roots (Trench) during the growing season in a regenerating deciduous forest using an automated chamber system with an up-to-date analyzer. Combined with a Random Forest (RF) approach, we studied the spatiotemporal variation and control of soil CH4 uptake rates. The results showed that the soil was a CH4 sink throughout the experimental period, and the existence of roots significantly enhanced CH4 uptake, mainly through improving soil aeration. The CH4 uptake rate varied seasonally according to the variations in soil gaseous diffusion caused by soil moisture and temperature differences. In addition, soil CH4 uptake showed a significant spatial variation, mainly resulting from spatial difference in soil porosity, soil carbon and nitrogen contents, and fine root biomass. The RF models showed high performance in soil CH4 flux prediction using the soil O2 diffusion coefficient and soil temperature as explanatory variables. The performance of RF models using ordinary variables of soil water content or water-filled pore space (WFPS) was equal to or slightly better than that of models using the diffusion coefficient. The higher importance of ambient CH4 concentration in Trench chambers indicates an increase in soil CH4 uptake at higher CH4 concentrations, which is predicted in the future. Although there are limitations, we believe that a machine learning approach, such as RF, using a large amount of continuous data with high temporal resolution, has great potential for investigating the dynamic variation in soil CH4 flux.

128-LB: Accessing Glucose Fluctuation—The Role of FPG, PPG, and HbA1c

Background: Time in range (TIR) has been widely used clinically to investigate the glucose fluctuation in diabetes management. Studies have been done in exploring the relationship between TIR and HbA1c to estimate TIR without multi-invasive measurements. Beyond the univariate correlation model, this study incorporated fasting plasma glucose (FPG), postprandial glucose (PPG) obtained from glucose tolerance test, which can to some extent reflect glucose fluctuations, and revealed the advantage and the potential of the multi-variate linear or no-linear model to estimate TIR. Method: This study involved 399 Chinese T2DM patients. TIR were calculated from continuous glucose monitoring (CGM) data with a coverage of at least 70%. FPG, PPG were measured by oral or mixed-meal glucose tolerance test. Univariate, multivariate linear regression and non-linear random forest (RF) regression were used in regression analysis. Model performance was assessed using MAE, RMSE and MSE. Results: Regression analysis between glucose tolerance test data and CGM suggests that FPG were strongly linked to trough glucose while PPG were strongly correlated with peak glucose. Regression analysis revealed a stronger correlation between TIR and a triad of FPG, PPG, HbA1c than single HbA1c variable. And the performance of RF model is superior (correlation = 0.79, MAE = 14.77), among multivariate linear (0.73, 16.67) and univariate linear (0.62, 21.29) model. Conclusion: Our study shows that FPG and PPG are closely associated to trough and peak glucose in CGM data, as they provide a more complete picture of glucose fluctuations compared to HbA1c alone. The TIR estimation model that based on FPG, PPG, and HbA1c outperformed univariate model. However, the performance of RF based non-linear model suggests a more complex relationship between those parameters, indicating personalized models for different populations is needed in TIR estimating without long-term invasive glucose monitoring. Disclosure J. Ma: None. X. Su: None. L. Feng: Employee; Hua Medicine. B. Ding: None. R. Yan: None. R. Sun: None. Y. Zhang: None. Y. Duan: None. Funding Nanjing Municipal Health Commission (ZKX22038)

A Novel Sensing Feature Extraction Based on Mold Temperature and Melt Pressure for Plastic Injection Molding Quality Assessment

Injection molding is one of the polymer molding methods. Product quality mainly can be affected by temperature and pressure. To observe the process of the melt forming in the mold, it is required to monitor the signal with the sensor. To achieve the purpose of robust molding, this research provides a new design-process window architecture based on machine learning of the injection product quality assessment. The process window is also known as the forming area map. The monitored value for defective products will be above or below the variable limit. The process is set to the center of this window so that any variation within the window will still result in an acceptable product. There are some potential problems in the current process. First, the conventional process window uses a small number of process variables as monitoring indicators, and it cannot provide stronger monitoring indicators. Second, most of the conventional process windows are 2-D and there are many monitoring windows, so the information that can be monitored is limited. Therefore, product quality cannot be effectively monitored without proper selection of important factors. Finally, the current process window can only provide information in the process and cannot make product quality prediction. In order to solve the above problems, a machine learning method is proposed to design the process window. The random forest classifier and regressor are used to predict the product quality. In order to improve the usability of the factor, the design of the proposed new variable is derived from the classical formula of thermodynamics (heat conduction). Using existing factors and proposing time-based feature changes can better represent the molding process as a whole. Based on the experimental results, the proposed variables can effectively increase the predictive performance of random forest model for quality assessment. The research results make a satisfactory result to the quality assessment of industrial injection molding. The accuracy of the prediction results is 100%. Moreover, it can be found from the 3-D process window that the data are very concentrated and can be clearly classified by the hyperplane, which is defined as the shortest distance between two clusters, and this space is the boundary of the process window. Therefore, compared with the past, the judgment boundary based on the random forest process window is no longer a simple boundary. Random forests can provide more accurate visualization boundaries.

Empirical evaluation of internal validation methods for prediction in large-scale clinical data with rare-event outcomes: a case study in suicide risk prediction

BackgroundThere is increasing interest in clinical prediction models for rare outcomes such as suicide, psychiatric hospitalizations, and opioid overdose. Accurate model validation is needed to guide model selection and decisions about whether and how prediction models should be used. Split-sample estimation and validation of clinical prediction models, in which data are divided into training and testing sets, may reduce predictive accuracy and precision of validation. Using all data for estimation and validation increases sample size for both procedures, but validation must account for overfitting, or optimism. Our study compared split-sample and entire-sample methods for estimating and validating a suicide prediction model.MethodsWe compared performance of random forest models estimated in a sample of 9,610,318 mental health visits (“entire-sample”) and in a 50% subset (“split-sample”) as evaluated in a prospective validation sample of 3,754,137 visits. We assessed optimism of three internal validation approaches: for the split-sample prediction model, validation in the held-out testing set and, for the entire-sample model, cross-validation and bootstrap optimism correction.ResultsThe split-sample and entire-sample prediction models showed similar prospective performance; the area under the curve, AUC, and 95% confidence interval was 0.81 (0.77–0.85) for both. Performance estimates evaluated in the testing set for the split-sample model (AUC = 0.85 [0.82–0.87]) and via cross-validation for the entire-sample model (AUC = 0.83 [0.81–0.85]) accurately reflected prospective performance. Validation of the entire-sample model with bootstrap optimism correction overestimated prospective performance (AUC = 0.88 [0.86–0.89]). Measures of classification accuracy, including sensitivity and positive predictive value at the 99th, 95th, 90th, and 75th percentiles of the risk score distribution, indicated similar conclusions: bootstrap optimism correction overestimated classification accuracy in the prospective validation set.ConclusionsWhile previous literature demonstrated the validity of bootstrap optimism correction for parametric models in small samples, this approach did not accurately validate performance of a rare-event prediction model estimated with random forests in a large clinical dataset. Cross-validation of prediction models estimated with all available data provides accurate independent validation while maximizing sample size.

Prey ingestion rates revealed by back-mounted accelerometers in Eurasian spoonbills

BackgroundQuantifying foraging success in space and time and among individuals is essential for answering many ecological questions and may guide conservation efforts. However, collecting this information is challenging for species that forage on mobile prey and are difficult to observe visually, for example, because they forage in inaccessible areas or at night. In such cases, the use of tracking devices that simultaneously collect location and acceleration data may provide a solution if foraging success can be extracted successfully. The aim of this study was to assess how well searching for and ingesting prey, among other behaviours, could be distinguished from 20 Hz acceleration data collected by GPS/ACC-trackers mounted on the back of Eurasian spoonbills Platalea leucorodia. Upon capturing a prey, spoonbills make a distinct movement with their head and back to throw the prey from the tip of the bill into the throat.MethodsWe compared the behavioural classification performance of random forest models that were trained and tested on video-annotated acceleration data segments of different (fixed or flexible) lengths. The best-performing model was then applied to 4 years of data of spoonbills foraging in the Wadden Sea during the breeding season, to explore seasonal and annual variation in prey ingestion rates.ResultsHighest classification accuracies (as indicated by the F-measure, a balanced measure of precision and sensitivity) of foraging behaviours were achieved by analysing short fixed-length segments (0.4–0.8 s) or “flexibly-cut” segments. The F-measure was very high (> 0.90) for searching, standing, sitting and flying (distinguishing active and passive flight), 0.73 for ingesting prey and 0.65 for walking. False positive and negative prey ingestions were equally likely and most often confused with searching, resulting in a close match between the predicted and observed prey ingestion rates. Application of the best-performing model revealed strong seasonal patterns in prey ingestion rates in the Wadden Sea that varied between years.ConclusionsWe show that prey ingestion rates of spoonbills can be fairly accurately estimated from acceleration data. These results are promising for the use of spoonbills equipped with GPS/ACC-trackers as monitors of spatial and temporal variation in the availability of small fish and shrimp, which is key to understand the foraging and migratory movements of spoonbills and provides information on the quality of (coastal) wetlands.

Exploiting Sentinel-2 dataset to assess flow intermittency in non-perennial rivers

Knowledge about the frequency and duration of each flowing status of non-perennial rivers is severely limited by the small number of streamflow gauges and reliable prediction of surface water presence by hydrological models. In this study, multispectral Sentinel-2 images were used to detect and monitor changes in water surface presence along three non-perennial Mediterranean rivers located in southern Italy. Examining the reflectance values of water, sediment and vegetation covers, the bands in which these classes are most differentiated were identified. It emerged that the false-color composition of the Sentinel-2 bands SWIR, NIR and RED allows water surfaces to be clearly distinguished from the other components of the river corridor. From the false-color composite images, it was possible to identify the three distinct flowing status of non-perennial rivers: “flowing” (F), “ponding” (P) and “dry” (D). The results were compared with field data and very high-resolution images. The flowing status was identified for all archive images not affected by cloud cover. The obtained dataset allowed to train Random Forest (RF) models able to fill temporal gaps between satellite images, and predict the occurrence of one of the three flowing status (F/P/D) on a daily scale. The most important predictors of the RF models were the cumulative rainfall and air temperature data before the date of satellite image acquisition. The performances of RF models were very high, with total accuracy of 0.82–0.97 and true skill statistic of 0.64–0.95. The annual non-flowing period (phases P and D) of the monitored rivers was assessed in range 5 to 192 days depending on the river reach. Due to the easy-to-use algorithm and the global, freely available satellite imagery, this innovative technique has large application potential to describe flowing status of non-perennial rivers and estimate frequency and duration of surface water presence.

Sequential feature selection for heart disease detection using random forest

Heart disease identification is one of the most challenging task that requires highly experienced cardiologists. However, in developing nations such as Ethiopia, there are a few cardiologists and heart disease detection is more challenging. As an alternative solution to cardiologist, this study proposed a more effective model for heart disease detection by employing random forest and sequential feature selection (SFS). SFS is an effective approach to improve the performance of random forest model on heart disease detection. SFS removes unrelated features in heart disease dataset that tends to mislead random forest model on heart disease detection. Thus, removing inappropriate and duplicate features from the training set with sequential feature selection approach plays significant role in improving the performance of the proposed model. The proposed feature selection approach is evaluated using real world clinical heart disease dataset collected from University of California Irvine (UCI) data repository. Empirical test on validation set reveals that the proposed model performs well as compared to the existing methods. Overall, the state of-the-art heart disease detection model with classification accuracy of 98.53% is proposed for heart disease detection using SFS and random forest model.

Predicting and Mapping Potential Fire Severity for Risk Analysis at Regional Level Using Google Earth Engine

Despite being a natural ecological process, wildfires are dramatic events that, accelerated by global change, could negatively affect ecosystem services depending on their severity level. However, because of data processing constraints, fire severity has been mostly neglected in risk analysis (especially at regional levels). Indeed, previous studies addressing fire severity focused mainly on analyzing single fire events, preventing the projection of the results over large areas. Although, building and projecting robust models of fire severity to integrate into risk analysis is of main importance to best anticipate decisions. Here, taking advantage of free data-processing platforms, such as Google Earth Engine, we use more than 1000 fire records from Western Italy and Southern France in the years 2004–2017, to assess the performance of random forest models predicting the relativized delta normalized burn ratio (rdNBR) used as proxy of fire severity. Furthermore, we explore the explanatory capacity and meaning of several variables related to topography, vegetation, and burning conditions. To show the potentialities of this approach for operational purposes, we projected the model for one of the regions (Sardinia) within the study area. Results showed that machine learning algorithms explain up to 75% of the variability in rdNBR, with variables related to vegetation amount and topography being the most important. These results highlight the potential usefulness of these tools for mapping fire severity in risk assessments.

Automatic Parkinson’s disease detection based on the combination of long-term acoustic features and Mel frequency cepstral coefficients (MFCC)

In recent years, acoustic signals have found increasing popularity among the scientific community as biomarkers for detecting Parkinson’s disease. Literature highlights that 90% of people with Parkinson’s disease have vocal impairment symptoms at earlier stages of the disease. Standard Parkinson’s disease diagnosis depends on medical imaging or observation of motor symptoms such as gait disturbance and muscle rigidity which may appear at later stages. Several acoustic feature extraction algorithms, such as long-term and short-term features, were developed to train machine learning models to detect Parkinson’s disease using a single algorithm. Although the literature demonstrates promising results, there has not been any exploitation of the potential of combining solely long-term and short-term algorithms to the best of the authors’ knowledge. This paper presents a novel method to investigate the effect of a dataset incorporating long-term and short-term features known as Mel frequency cepstral coefficients (MFCC) on the performance of random forest model for Parkinson’s disease detection. The performance of random forest has been compared between three feature sets: MFCC features, long-term features, and a combination of MFCC with long-term features. The combined features improved the detection accuracy to 88.84%, while independent sets of MFCC and long-term features, were 84.12% and 84.00%, respectively. The findings of the proposed method indicates the effectiveness of the long-term features and MFCC combination in improving the overall performance of Parkinson’s disease detection model.

Monitoring Variables Influence on Random Forest Models to Forecast Injuries in Short-Track Speed Skating.

Injuries limit the athletes' ability to participate fully in their training and competitive process. They are detrimental to performance, affecting the athletes psychologically while limiting physiological adaptations and long-term development. This study aims to present a framework for developing random forest classifier models, forecasting injuries in the upcoming 1 to 7 days, to assist the performance support staff in reducing injuries and maximizing performance within the Canadian National Female Short-Track Speed Skating Program. Forty different variables monitored daily over two seasons (2018–2019 and 2019–2020) were used to develop two sets of forecasting models. One includes only training load variables (TL), and a second (ALL) combines a wide array of monitored variables (neuromuscular function, heart rate variability, training load, psychological wellbeing, past injury type, and location). The sensitivity (ALL: 0.35 ± 0.19, TL: 0.23 ± 0.03), specificity (ALL: 0.81 ± 0.05, TL: 0.74 ± 0.03) and Matthews Correlation Coefficients (MCC) (ALL: 0.13 ± 0.05, TL: −0.02 ± 0.02) were computed. Paired T-test on the MCC revealed statistically significant (p < 0.01) and large positive effects (Cohen d > 1) for the ALL forecasting models' MCC over every forecasting window (1 to 7 days). These models were highly determined by the athletes' training completion, lower limb and trunk/lumbar injury history, as well as sFatigue, a training load marker. The TL forecasting models' MCC suggests they do not bring any added value to forecast injuries. Combining a wide array of monitored variables and quantifying the injury etiology conceptual components significantly improve the injury forecasting performance of random forest models. The ALL forecasting models' performances are promising, especially on one time windows of one or two days, with sensitivities and specificities being respectively above 0.5 and 0.7. They could add value to the decision-making process for the support staff in order to assist the Canadian National Female Team Short-Track Speed Skating program in reducing the number of incomplete training days, which could potentially increase performance. On longer forecasting time windows, ALL forecasting models' sensitivity and MCC decrease gradually. Further work is needed to determine if such models could be useful for forecasting injuries over three days or longer.

Terrain Variables Based Spatial Mapping of Soil Properties in a Watershed of Himalayan Landscape Using Random Forest Model

Soil information, including their spatial variability is vital for devising various soil and land management policies and strategies as well as assessing the environmental impact of different land use changes. Among the various soil forming factors, terrain plays a vital role in determining the spatial distribution of soil properties especially in hilly terrains due to its impact on soil processes as well as the redistribution of soil materials. The study was carried out for assessing the applicability and efficiency of various terrain parameters for prediction as well as spatial mapping of different soil properties using Random Forest (RF) modeling approach, in a mid-Himalayan watershed located in Tehri Garhwal district, Uttarakhand. Field data collection was done by collecting soil samples from 0-15cm depth (surface layer) at 68 different locations in the study area, following a transect survey method. Collected samples were pre-processed and analyzed in the laboratory for estimating various soil properties such as sand, silt and clay contents as well as soil nutrients such as Soil Organic Carbon (SOC) and nitrogen (N). Fifteen terrain parameters derived by digital terrain analysis of CartoDEM of 10m spatial resolution were used as predictor variables during RF modeling. Pearson correlation analysis revealed poor linear relationship of soil properties with various terrain parameters. Variable importance ranking revealed that compound topographic index, elevation, heat load index, integrated moisture index and mean curvature are the most important variables, explaining maximum variability in different properties. The performance of RF models for predicting various soil properties were evaluated based on the coefficient of determination (R2), Root Mean Square Error (RMSE) as well as Lin’s Concordance Correlation Coefficient (CCC). The models were found to perform well with R2 and CCC values of 0.47 and 0.50 for sand, 0.52 and 0.55 for silt, 0.44 and 0.57 for clay, 0.41 and 0.36 for organic carbon and 0.55 and 0.34 for nitrogen, respectively. The validated models were further employed for generating the spatial distribution maps of various soil properties. Such attempts combining various environmental variables with machine learning based models may aid us in mapping the spatial distribution of various soil properties in the hilly mountainous terrains with considerable accuracies.

Exploring the Use of Waste Marble Powder in Concrete and Predicting Its Strength with Different Advanced Algorithms.

Recently, the high demand for marble stones has progressed in the construction industry, ultimately resulting in waste marble production. Thus, environmental degradation is unavoidable because of waste generated from quarry drilling, cutting, and blasting methods. Marble waste is produced in an enormous amount in the form of odd blocks and unwanted rock fragments. Absence of a systematic way to dispose of these marble waste massive mounds results in environmental pollution and landfills. To reduce this risk, an effort has been made for the incorporation of waste marble powder into concrete for sustainable construction. Different proportions of marble powder are considered as a partial substitute in concrete. A total of 40 mixes are prepared. The effectiveness of marble in concrete is assessed by comparing the compressive strength with the plain mix. Supervised machine learning algorithms, bagging (Bg), random forest (RF), AdaBoost (AdB), and decision tree (DT) are used in this study to forecast the compressive strength of waste marble powder concrete. The models’ performance is evaluated using correlation coefficient (R2), root mean square error, and mean absolute error and mean square error. The achieved performance is then validated by using the k-fold cross-validation technique. The RF model, having an R2 value of 0.97, has more accurate prediction results than Bg, AdB, and DT models. The higher R2 values and lesser error (RMSE, MAE, and MSE) values are the indicators for better performance of RF model among all individual and ensemble models. The implementation of machine learning techniques for predicting the mechanical properties of concrete would be a practical addition to the civil engineering domain by saving effort, resources, and time.

Research on stock trend prediction method based on optimized random forest

AbstractAs a complex hot problem in the financial field, stock trend forecasting uses a large amount of data and many related indicators; hence it is difficult to obtain sustainable and effective results only by relying on empirical analysis. Researchers in the field of machine learning have proved that random forest can form better judgements on this kind of problem, and it has an auxiliary role in the prediction of stock trend. This study uses historical trading data of four listed companies in the USA stock market, and the purpose of this study is to improve the performance of random forest model in medium‐ and long‐term stock trend prediction. This study applies the exponential smoothing method to process the initial data, calculates the relevant technical indicators as the characteristics to be selected, and proposes the D‐RF‐RS method to optimize random forest. As the random forest is an ensemble learning model and is closely related to decision tree, D‐RF‐RS method uses a decision tree to screen the importance of features, and obtains the effective strong feature set of the model as input. Then, the parameter combination of the model is optimized through random parameter search. The experimental results show that the average accuracy of random forest is increased by 0.17 after the above process optimization, which is 0.18 higher than the average accuracy of light gradient boosting machine model. Combined with the performance of the ROC curve and Precision–Recall curve, the stability of the model is also guaranteed, which further demonstrates the advantages of random forest in medium‐ and long‐term trend prediction of the stock market.

Grain Yield Estimation in Rice Breeding Using Phenological Data and Vegetation Indices Derived from UAV Images

The accurate estimation of grain yield in rice breeding is crucial for breeders to screen and select qualified cultivars. In this study, a low-cost unmanned aerial vehicle (UAV) platform mounted with an RGB camera was carried out to capture high-spatial resolution images of rice canopy in rice breeding. The random forest (RF) regression techniques were used to establish yield models by using (1) only color vegetation indices (VIs), (2) only phenological data, and (3) fusion of VIs and phenological data as inputs, respectively. Then, the performances of RF models were compared with the manual observation and CERES-Rice model. The results indicated that the RF model using VIs only performed poorly for estimating yield; the optimized RF model that combined the use of phenological data and color VIs performed much better, which demonstrated that the phenological data significantly improved the model performance. Furthermore, the yield estimation accuracy of 21 rice cultivars that were continuously planted over three years in the optimal RF model had no significant difference (p &gt; 0.05) with that of the CERES-Rice model. These findings demonstrate that the RF model, by combining phenological data and color Vis, is a potential and cost-effective way to estimate yield in rice breeding.