Tree-based Algorithm Research Articles

Automated Image Analysis of Ultrasound Image using Machine Learning Models: Detection of Double Contours

Background: Musculoskeletal ultrasound (MSUS) is commonly used to detect double contour sign (DCS) to diagnose gout. MSUS is subjective and needs user training and experience for optimal visualisation. Machine learning (ML) models can be used for automatic detection of DCS. We sought to identify the optimal algorithm for detection of DCS on ultrasound images of the fist metatarsophalangeal joint (MTPJ1) using candidate ML algorithms. Methods: MSUS images of MTPJ1 of male patients acquired between January, 2020 and January, 2023 were included in this study. Diagnosis of gout was based on imaging, joint aspiration, serum uric acid level, and clinical evaluation. Images were acquired with a high frequency linear transducer (18 MHz) with a dedicated MSUS machine (Toshiba Xario 200). Patients diagnosed with calcium pyrophosphate crystal disease or gross deformities of MTPJ1 were excluded. Selected images were pre-processed using Keras-OCR. The following candidate ML algorithms were trained: Support Vector machine, tree-based algorithms like Random Forest and Decision Tree and boosting-based algorithms like Gradient Boosting and XGBoost and these were compared with a deep neural network (2D-CNN). Classification outcomes (DCS positive or negative) were compared before and after image augmentation. Classification performance was measured using sensitivity, specificity, and area under the receiver operating curve (AUROC) along with accuracy, precision and F1-score. Results: We used 410 high resolution ultrasound images of MTPJ1 of male patients, 311 patients with gout and positive DCS and 99 patients without gout and no DCS. On average, the final model using XGBoost for detection of DCS showed sensitivity of 94% and specificity of 93.7% and AUROC of 85%. Using XGBoost accuracy and precision were 84.74% and 87% respectively and recall and F1-scores were 94% and 90% respectively. Comparison of different models is given in Table 1. Classification performance improved after augmentation in almost all the models. Conclusions: The classification model based on a boosting based ML model (XGBoost) was trained with a relatively small dataset; however, it showed good accuracy. This proposed system was useful in correct classification of DCS in MSUS images with potential to aid imaging diagnostics of gout in a wider clinical setting.

Real-time EEG-based emotion recognition for neurohumanities: perspectives from principal component analysis and tree-based algorithms.

Within the field of Humanities, there is a recognized need for educational innovation, as there are currently no reported tools available that enable individuals to interact with their environment to create an enhanced learning experience in the humanities (e.g., immersive spaces). This project proposes a solution to address this gap by integrating technology and promoting the development of teaching methodologies in the humanities, specifically by incorporating emotional monitoring during the learning process of humanistic context inside an immersive space. In order to achieve this goal, a real-time emotion recognition EEG-based system was developed to interpret and classify specific emotions. These emotions aligned with the early proposal by Descartes (Passions), including admiration, love, hate, desire, joy, and sadness. This system aims to integrate emotional data into the Neurohumanities Lab interactive platform, creating a comprehensive and immersive learning environment. This work developed a ML, real-time emotion recognition model that provided Valence, Arousal, and Dominance (VAD) estimations every 5 seconds. Using PCA, PSD, RF, and Extra-Trees, the best 8 channels and their respective best band powers were extracted; furthermore, multiple models were evaluated using shift-based data division and cross-validations. After assessing their performance, Extra-Trees achieved a general accuracy of 94%, higher than the reported in the literature (88% accuracy). The proposed model provided real-time predictions of VAD variables and was adapted to classify Descartes' six main passions. However, with the VAD values obtained, more than 15 emotions can be classified (reported in the VAD emotion mapping) and extend the range of this application.

The Current Situation of the Implementation of College Admission System and the Way of Reform in the Digital Age

Abstract With the progress of science and technology and social changes, higher education in China is experiencing unprecedented challenges and opportunities. The digital era has not only changed teaching methods but has also profoundly affected the enrollment system of colleges and universities. The study uses a decision tree-based enrollment information mining algorithm and a gray prediction model to construct a data mining model of the current status of enrollment system implementation. Through the data mining model, the data mining and analysis of the student source status of college enrollment are used to propose targeted reform proposals. The number of female students in School Z has been increasing each year, while the number of male students has decreased significantly. The number of students with grades A and B is much lower than that of students with grades B and C. Among the special types of enrollment sources, the enrollment of national defense students and high-level athletes has decreased year by year, while the enrollment of the poverty special program has shown a yearly increasing trend. During the four years, the enrollment of national defense students was 61, 54, 46, and 39, and the enrollment of high-level athletes was 15, 14, 12, and 10. The enrollment in special programs for people experiencing poverty increased from 9 to 47 during the four years. The reform of the college enrollment system should focus on reforming the content of the examination, and government departments to do a good job of coordinating planning, improving laws and regulations, and putting an end to malpractice and irregularities.

A Novel Comparison of Charotar Region Wheat Variety Classification Techniques using Purely Tree-based Data Mining Algorithms

Techniques for classifying data using data mining are now a day prevalent in agriculture. The method of classifying seeds involves grouping various seed varieties according to their morphological characteristics. To accomplish categorization of the typical Charotar region (generally comprising Anand and Kheda districts of the Gujarat State of India) Gujarat Wheat (GW) varieties (TRITICUM – AESTIVUM) viz. GW 273, GW 496, GW 322, LOK-1, and GDW 1255 (TRITICUM – DURUM), Weka Explorer was used. The features used are area, perimeter, solidity, aspect ratio, major and minor axis of seed kernel, Hue, Saturation, Value, and SF1 (empirical). Features reduction was done using Information Gain (IG) and its modified version Gain Ratio (GR). This paper compares performance of Tree based data mining algorithms in classifying wheat varieties. For classification we used purely tree-based machine learning algorithms viz. J48, Random Forest, Hoeffding Tree, Logistic Model Tree (LMT), and REPTree. LMT- logistics regression method gives higher accuracy 96.4% compared to other classifiers. Hoeffding Tree classifiers stood second with 96% accuracy. For validation 10-fold cross validation was used. By reducing the number of folds in cross validation performance of most algorithms decreased except J48. The percentage of correctly classified instance increased for all algorithms when features were selected by GR except for J48.

Evaluating the performance of an AI-powered VBAC prediction system within a decision-aid birth choice platform for shared decision-making.

Vaginal birth after cesarean (VBAC) is generally regarded as a safe and viable birthing option for most women with prior cesarean delivery. Nonetheless, concerns about heightened risks of adverse maternal and perinatal outcomes have often dissuaded women from considering VBAC. This study aimed to assess the performance of an artificial intelligence (AI)-powered VBAC prediction system integrated into a decision-aid birth choice platform for shared decision-making (SDM). Employing a retrospective design, we collected medical records from a regional hospital in northern Taiwan from January 2019 to May 2023. To explore a suitable model for tabular data, we compared two prevailing modeling approaches: tree-based models and logistic regression models. We subjected the tree-based algorithm, CatBoost, to binary classification. Forty pregnant women with 347 records were included. The CatBoost model demonstrated a robust performance, boasting an accuracy rate of 0.91 (95% confidence interval (CI): 0.86-0.94) and an area under the curve of 0.89 (95% CI: 0.86-0.93), surpassing both regression models and other boosting techniques. CatBoost captured the data characteristics on the significant impact of gravidity and the positive influence of previous vaginal birth, reinforcing established clinical guidelines, as substantiated by the SHapley Additive exPlanations analysis. Using AI techniques offers a more accurate assessment of VBAC risks, boosting women's confidence in selecting VBAC as a viable birthing option. The seamless integration of AI prediction systems with SDM platforms holds a promising potential for enhancing the effectiveness of clinical applications in the domain of women's healthcare.

Using machine learning methods to predict all-cause somatic hospitalizations in adults: A systematic review.

In this review, we investigated how Machine Learning (ML) was utilized to predict all-cause somatic hospital admissions and readmissions in adults. We searched eight databases (PubMed, Embase, Web of Science, CINAHL, ProQuest, OpenGrey, WorldCat, and MedNar) from their inception date to October 2023, and included records that predicted all-cause somatic hospital admissions and readmissions of adults using ML methodology. We used the CHARMS checklist for data extraction, PROBAST for bias and applicability assessment, and TRIPOD for reporting quality. We screened 7,543 studies of which 163 full-text records were read and 116 met the review inclusion criteria. Among these, 45 predicted admission, 70 predicted readmission, and one study predicted both. There was a substantial variety in the types of datasets, algorithms, features, data preprocessing steps, evaluation, and validation methods. The most used types of features were demographics, diagnoses, vital signs, and laboratory tests. Area Under the ROC curve (AUC) was the most used evaluation metric. Models trained using boosting tree-based algorithms often performed better compared to others. ML algorithms commonly outperformed traditional regression techniques. Sixteen studies used Natural language processing (NLP) of clinical notes for prediction, all studies yielded good results. The overall adherence to reporting quality was poor in the review studies. Only five percent of models were implemented in clinical practice. The most frequently inadequately addressed methodological aspects were: providing model interpretations on the individual patient level, full code availability, performing external validation, calibrating models, and handling class imbalance. This review has identified considerable concerns regarding methodological issues and reporting quality in studies investigating ML to predict hospitalizations. To ensure the acceptability of these models in clinical settings, it is crucial to improve the quality of future studies.

Variable Importance Measures for Multivariate Random Forests

Multivariate random forests (or MVRFs) are an extension of tree-based ensembles to examine multivariate responses. MVRF can be particularly helpful where some of the responses exhibit sparse (e.g., zero-inflated) distributions, making borrowing strength from correlated features attractive. Tree-based algorithms select features using variable importance measures (VIMs) that score each covariate based on the strength of dependence of the model on that variable. In this paper, we develop and propose new VIMs for MVRFs. Specifically, we focus on the variable’s ability to achieve split improvement, i.e., the difference in the responses between the left and right nodes obtained after splitting the parent node, for a multivariate response. Our proposed VIMs are an improvement over the default naïve VIM in existing software and allow us to investigate the strength of dependence both globally and on a per-response basis. Our simulation studies show that our proposed VIM recovers the true predictors better than naïve measures. We demonstrate usage of the VIMs for variable selection in two empirical applications; the first is on Amazon Marketplace data to predict Buy Box prices of multiple brands in a category, and the second is on ecology data to predict co-occurrence of multiple, rare bird species. A feature of both data sets is that some outcomes are sparse — exhibiting a substantial proportion of zeros or fixed values. In both cases, the proposed VIMs when used for variable screening give superior predictive accuracy over naïve measures.

Using a Tree-Based Algorithm to Categorize Students who have Dropped out Evidence from Universities in Vietnam

Purpose: The purpose of this article is to identify and comprehend the mechanism of operation of the Decision Tree and Random Forest algorithms. comprehend precise implementation steps, better comprehend nature, and synthesize experience in the practice process. Methodology: The study was conducted based on an analysis of dropout data from Trade Union University, University of Social Labour, and Phenikaa University in Vietnam. Result: The article investigates and develops a machine learning model for classifying students as dropouts or not. Develop scales to measure the accuracy and dependability of the model. Value: The article assesses the model's outcomes, performance, and accuracy. Give your suggestions for future development and improvement of the topic. Propose several solutions to the challenge of identifying students as dropouts, based on the real scenario employed by the algorithm during the analysis.

An Efficient and Scalable RFID Anti-Collision Algorithm on Optimal Partition and Collided Block Bit-Mapping

In RFID systems, many anti-collision algorithms, driven by the concept of rescheduling the response sequence between the reader and unidentified tags, have been put forward to solve tag collision problem, including ALOHA-based, tree-based and hybrid algorithms. In this paper, we propose a novel RFID anti-collision algorithm called EAQ-CBB, which adopts three main approaches: tag population estimation based on collided bit detection method, optimal partitions and trimmed query tree based on the strategy of collided block bit-mapping (QTCBB). The relatively accurate estimation of tag backlog and optimal partition ensure a great reduction of collisions in the initial phase. For each collided partition, a QTCBB process is introduced immediately, which eliminates all the empty slots and significantly reduces the collided slots. Simulation results show that EAQ-CBB performs good stability and scalability when the key parameters change. Compared with the existing algorithms, such as DFSA, QTI, T-GDFSA and CT, EAQ-CBB outperforms the others with high system throughput, low normalized latency and low normalized overhead at a low cost of energy, which makes it easier to be used widely in the efficient-aware and energy-aware applications.

GLMM and GLMMTree for Modelling Poverty in Indonesia

GLMMTree is a tree-based algorithm that can detect interaction and find subgroups in the GLMM to improve fixed effect estimation. This study uses GLMM trees in real data applications of poverty in Indonesia. Using this data, we found that the GLMMTree algorithm method performs similarly to GLMM. 2 significant predictors affect poverty in Indonesia: the unemployment rate and the GRDP at a constant price. GLMMTree algorithm enriches the analysis by finding two variables, namely the percentage of households with electricity lighting access and the percentage of households with clean drinking water sources, that interact with predictor variables in the model.

Data-driven machine learning model of a Selective Catalytic Reduction on Filter (SCRF) in a heavy-duty diesel engine: A comparison of Artificial Neural Network with Tree-based algorithms

The Selective Catalytic Reduction on Filter (SCRF) system is yet to be deployed in current heavy-duty diesel engine aftertreatment system. Due to the thermal, space and cost benefits of the SCRF, it could become a useful component of the after-treatment system of a heavy-duty diesel engine, as regulators continue to demand an even cleaner environment. Ammonia cross-sensitivity of NOx sensors at the post-SCRF location poses challenges in measuring NOx emission accurately at this location and in turn affects the NOx conversion efficiency calculations. Developing a model that could replace the NOx sensor helps to mitigate the ammonia cross-sensitivity challenge as well as provides a medium to measure post-SCRF NOx concentration and NOx conversion efficiency while saving the cost on NOx sensors. This work focuses on a data-driven approach to developing a model for predicting NOx conversion efficiency across the SCRF using Artificial Neural Network, Bootstrap Forest, and Boosted Tree methods. Further, the three modeling techniques were also compared for accuracy and computation cost.

Early prediction of mortality at sepsis diagnosis time in critically ill patients by using interpretable machine learning.

This study applied machine learning for the early prediction of 30-day mortality at sepsis diagnosis time in critically ill patients. Retrospective study using data collected from the Medical Information Mart for Intensive Care IV database. The data of the patient cohort was divided on the basis of the year of hospitalization, into training (2008-2013), validation (2014-2016), and testing (2017-2019) datasets. 24,377 patients with the sepsis diagnosis time < 24h after intensive care unit (ICU) admission were included. A gradient boosting tree-based algorithm (XGBoost) was used for training the machine learning model to predict 30-day mortality at sepsis diagnosis time in critically ill patients. Model performance was measured in both discrimination and calibration aspects. The model was interpreted using the SHapley Additive exPlanations (SHAP) module. The 30-day mortality rate of the testing dataset was 17.9%, and 39 features were selected for the machine learning model. Model performance on the testing dataset achieved an area under the receiver operating characteristic curve (AUROC) of 0.853 (95% CI 0.837-0.868) and an area under the precision-recall curves of 0.581 (95% CI 0.541-0.619). The calibration plot for the model revealed a slope of 1.03 (95% CI 0.94-1.12) and intercept of 0.14 (95% CI 0.04-0.25). The SHAP revealed the top three most significant features, namely age, increased red blood cell distribution width, and respiratory rate. Our study demonstrated the feasibility of using the interpretable machine learning model to predict mortality at sepsis diagnosis time.

Retrieval of Desert Microwave Land Surface Emissivity Based on Machine Learning Algorithms

Based on the community radiative transfer model, ensemble tree-based random forest algorithm, and extreme gradient boosting tree algorithm, this study established a random forest retrieval model (RF) and an extreme gradient boosting tree retrieval model (XGBoost) for the microwave land surface emissivity by using ERA5 reanalysis data and the observed brightness temperature of 10.65 GHz vertical polarization from FY3C Microwave Radiation Imager-I. In addition, an optimized Bayesian regularized neural network retrieval model (M2_30NN) has also been established on the basis of the original neural network land surface emissivity retrieval model (M1_20NN). The results show that compared with the simulated brightness temperature of the original land surface emissivity, the simulated brightness temperature of the land surface emissivity from each retrieval model is not only significantly improved in the correlation coefficient with the observed brightness temperature (5.92% (M1_20NN), 4.23% (M2_30NN), 14.21% (RF), 13.07% (XGBoost)), but also in the evaluation indexes of root mean square error, mean absolute error and explained variance regression score in the training datasets. Furthermore, in terms of testing datasets and spatiotemporal independence test datasets, the retrieval results of RF and XGBoost models can capture the spatial distribution patterns that are consistent with observations well, and also show great numerical improvement compared with the original model. In general, the XGBoost retrieval model is the best, followed by the RF retrieval model.

PREDICTING 10 YEAR MORTALITY WITH MACHINE LEARNING: EVIDENCE FROM NATIONAL SOCIAL LIFE, HEALTH, AND AGING PROJECT

Abstract The current academic literature extensively used linear logistic models with social, behavioral, and psychological status to predict mortality. However, few address the interdependency of predictors and the imbalance of targets that adversely bias the results. Using the National Social Life Health and Aging Project (NSHAP), we developed two machine learning models predicting the 10-year mortality of older adults in the US. We first used tree-based algorithms of Decision Tree (DT) that account for the interdependency of the social features and decide the splitting nodes and thresholds using entropy gain conditional on the previous splitting predictor to discern disposition status. Second, we used the Fuzzy Support Vector Machine (FSVM) that regards every sample as a node in high-dimensional vector space and splits the nodes with an optimum plane by finding the best linear combination of features to get an optimum prediction accuracy. Additionally, FSVM addresses the target imbalance problem by conducting a more delicate classification of samples with close predicted probabilities of being alive and deceased. Compared to the accuracy rates achieved by the Logistic Regression, our algorithms perform better on the entire population and the population near the class boundary. We also discussed the social and demographic characteristics of the cases whose disposition statuses were either wrongly predicted as deceased or alive by our algorithms. The findings serve important purposes for public health practitioners in accurately understanding the risk and protective factors of mortality in aging.

Accurate prediction of pressure losses using machine learning for the pipeline transportation of emulsions

One of the significant challenges to designing an emulsion transportation system is predicting frictional pressure losses with confidence. The state-of-the-art method for enhancing reliability in prediction is to employ artificial intelligence (AI) based on various machine learning (ML) tools. Six traditional and tree-based ML algorithms were analyzed for the prediction in the current study. A rigorous feature importance study using RFECV method and relevant statistical analysis was conducted to identify the parameters that significantly contributed to the prediction. Among 16 input variables, the fluid velocity, mass flow rate, and pipe diameter were evaluated as the top predictors to estimate the frictional pressure losses. The significance of the contributing parameters was further validated by estimation error trend analyses. A comprehensive assessment of the regression models demonstrated an ensemble of the top three regressors to excel over all other ML and theoretical models. The ensemble regressor showcased exceptional performance, as evidenced by its high R2 value of 99.7 % and an AUC-ROC score of 98 %. These results were statistically significant, as there was a noticeable difference (within a 95 % confidence interval) compared to the estimations of the three base models. In terms of estimation error, the ensemble model outperformed the top base regressor by demonstrating improvements of 6.6 %, 11.1 %, and 12.75 % for the RMSE, MAE, and CV_MSE evaluation metrics, respectively. The precise and robust estimations achieved by the best regression model in this study further highlight the effectiveness of AI in the field of pipeline engineering.

MACHINE LEARNING MODEL FOR THE PREDICTION OF CONDITION OF MUSEUM OBJECTS

An accurate prediction of the future condition of museum objects is crucial for developing appropriate proactive maintenance and preservation strategies. Despite this, there are very few such damage models that can be used in practice. The main reasons, for this lack of deterioration models, include complexity of deterioration problem and lack of understanding of the degradation mechanisms affecting various materials and objects, and lack of reliable quantitative approaches. In the article, we discuss the machine learning model, which predicts the future condition of museum objects. For this purpose, the model uses the data of MuIS (Estonian Museum Information System). To predict deterioration, we experimented primarily with various tree-based machine learning algorithms, such as the decision tree, the random forest, and XGBoost. The best results were obtained using the decision forest algorithm, which was able to identify 92% of deteriorating museum objects with 50% accuracy. The machine learning model provides a way to model ageing processes of museum objects over the course of time and thus better plan the preservation work of museums.

Estimation of sugarcane evapotranspiration from remote sensing and limited meteorological variables using machine learning models

Rapid and accurate crop evapotranspiration (ETc) estimation is essential for water resource management of irrigated lands. Nevertheless, due to the significant uncertainties in the inputs of conventional ETc equations and the inaccessibility of their proper spatial and temporal distribution values, precision agriculture requires a more accurate evaluation of ETc using efficient methods. Combining remote sensing (RS) data and machine learning (ML) techniques has provided a considerable capacity for estimating ETc, which can address these challenges. This study aims to develop a framework for ETc estimation based on the four most popular tree-based ML algorithms, namely M5-pruned (M5P), random forest regression (RFR), gradient-boosted regression trees (GBRT), and extreme gradient boosting (XGBoost). For this purpose, multisource RS data and meteorological and ground measurements (Meteo-GM) were used during three growing seasons (2018–2021) in irrigated croplands of the Sugarcane & By-Products Development Company in Khuzestan, southwest Iran. Since the unavailability of optical sensors, e.g., Landsat-8 data, in cloudy conditions restricts their enduring application, an alternative approach employing the Sentinel-1 Synthetic Aperture Radar (SAR) data for ETc estimation was also proposed. A set of 23 scenarios of various unique combinations of the available parameters was developed to ensure that a proper decision could be made for ETc estimation under any conditions. Eventually, the contribution of the input variables to estimate ETc was evaluated using the Sobol sensitivity analysis. The results showed that the RFR algorithm provided the most accurate ETc estimation in all scenarios, followed by the XGBoost, GBRT, and M5P algorithms (R2 = 0.92–0.99, RMSE = 2.02–0.32 mm d-1). The Meteo-GM models’ performances were improved by incorporating optical and thermal infrared (TIR) (R2 = 0.99, RMSE = 0.32 mm d-1) and SAR (R2 = 0.98, RMSE = 0.65 mm d-1) RS data. In addition, Sobol’s sensitivity analysis revealed that maximum temperature, wind speed, pan evaporation, leaf area index, moisture stress index, and inverse dual-pol diagonal distance were the most influential input variables in the ETc estimation. This study can be used as a reference to simplify the ETc calculations using a novel application of ML as a helpful basis for precision irrigation management from the point to the regional scale in any weather conditions.

A proposed tree-based explainable artificial intelligence approach for the prediction of angina pectoris

Cardiovascular diseases (CVDs) are a serious public health issue that affects and is responsible for numerous fatalities and impairments. Ischemic heart disease (IHD) is one of the most prevalent and deadliest types of CVDs and is responsible for 45% of all CVD-related fatalities. IHD occurs when the blood supply to the heart is reduced due to narrowed or blocked arteries, which causes angina pectoris (AP) chest pain. AP is a common symptom of IHD and can indicate a higher risk of heart attack or sudden cardiac death. Therefore, it is important to diagnose and treat AP promptly and effectively. To forecast AP in women, we constructed a novel artificial intelligence (AI) method employing the tree-based algorithm known as an Explainable Boosting Machine (EBM). EBM is a machine learning (ML) technique that combines the interpretability of linear models with the flexibility and accuracy of gradient boosting. We applied EBM to a dataset of 200 female patients, 100 with AP and 100 without AP, and extracted the most relevant features for AP prediction. We then evaluated the performance of EBM against other AI methods, such as Logistic Regression (LR), Categorical Boosting (CatBoost), eXtreme Gradient Boosting (XGBoost), Adaptive Boosting (AdaBoost), and Light Gradient Boosting Machine (LightGBM). We found that EBM was the most accurate and well-balanced technique for forecasting AP, with accuracy (0.925) and Youden's index (0.960). We also looked at the global and local explanations provided by EBM to better understand how each feature affected the prediction and how each patient was classified. Our research showed that EBM is a useful AI method for predicting AP in women and identifying the risk factors related to it. This can help clinicians to provide personalized and evidence-based care for female patients with AP.

CUSTOMS FRAUD DETECTION USING EXTREMELY BOOSTED NEURAL NETWORK (XBNET)

Noticing the vital role of cross-border trade has made Customs plays a crucial role not only in maintaining supply chain but also in securing government revenue from intentional fraud. Given the huge volume of international trade and limited workforce, Customs across the world must implement efficient and effective risk management. This paper proposes XBNet, an ensemble of tree-based algorithms with deep learning algorithms, to detect fraud in import activity. The strength of XBNet is combining gradient-boosted trees with neural networks where the weights, bias, and loss are adjusted simultaneously with the importance features of each tree in each layer. The object of this study is Import Declaration data from four Customs Offices, and the model is set for each Customs office to capture fraud patterns related to their region. We compared the model with two different parameters and concluded the models with learning rates = 1%, number of hidden layers = 2, activation function = sigmoid, and number of epochs = 100 as the most suitable for the Belawan, Merak, and Makassar and model with number of hidden layers = 2, number of epochs = 50 and other parameters are set the same as the most suitable for Tanjung Emas.

Dual sampling linear regression ensemble to predict wheat yield across growing seasons with hyperspectral sensing

In the domain of crop yield assessment, the combination of remote sensing and advanced regression techniques has gained significant attention. This study focused on predicting wheat yield across growing seasons using hyperspectral reflectance data. The performance of two decision tree-based algorithms, deep forest (DF) and random forest (RF), as well as a novel ensemble method called dual sampling linear regression ensemble (DSLRE), was evaluated. Canopy hyperspectral reflectance data were collected at two stages of wheat growth: early grain filling (EGF) and mid-grain filling (MGF), in three distinct growing environments. Prediction accuracy was assessed using the Pearson correlation coefficient (PCC) and mean absolute error (MAE). The results showed that the DF algorithm outperformed the RF algorithm in most scenarios. In modeling framework A, DF achieved the highest accuracy at the MGF stage using truncated reflectance (PCC = 0.60, MAE = 0.82 t/ha). In modeling framework B, DF performed better than RF in three out of four conditions, while RF exhibited higher prediction accuracy during the EGF stage using raw reflectance (PCC = 0.59, MAE = 1.48 t/ha). Moreover, the DSLRE algorithm outperformed DF and RF in six of the eight modeling scenarios and demonstrated robustness in coping with noisy features. The heritability analysis indicated that DSLRE predictions effectively captured the effects of genetic factors on yield variation. These findings illustrate the significance of the proposed DSLRE algorithm in improving yield prediction, providing valuable guidance for effective crop yield management.