Tree-based Ensemble Models Research Articles

Development of clustered machine learning technique for the modeling of scour profile induced by propeller jets

In this study novel clustered machine learning (CML) models were developed by applying different base learners including multi linear regression (MLR), adaptive neuro-fuzzy inference system (ANFIS), classification and regression trees (CART), least-squares boosting (LSBoost), and random forest (RF) to predict scour profile induced by the jet of a propeller. Moreover, the results of CML models were compared with regular ML algorithms. The results show that the tree-based ensemble models, RF and LSBoost, are more accurate than CART. The ANIFS has the worst performance compared with other regular nonlinear methods, RMSE = 17.414 mm. The performance of MLR is also weak, RMSE = 35.63 mm and R2 = 0.088. However, the implementation of CML drastically increases the performance of MLR, RMSE = 15.053 mm, and R2 = 0.871. The CANFIS with RMSE = 8.757 mm, R2 = 0.976, and SI = 0 is ranked as the most accurate model. The results show that CLM enhances the performance of regular ML by up to 59.90%. The application of clustered ML is recommended for other complex problems. Finally, the results show that the derived models trained by other datasets very well predict measured scour profiles which datasets were not used in the training process.

A UCB-Based Tree Search Approach to Joint Verification-Correction Strategy for Large-Scale Systems

Verification planning is a sequential decision-making problem that specifies a set of verification activities (VAs) and correction activities (CAs) at different phases of system development. While VAs are used to identify errors and defects, CAs also play important roles in system verification as they correct the identified errors and defects. However, current planning methods only consider VAs as decision choices. Because VAs and CAs have different activity spaces, planning a joint verification-correction strategy (JVCS) is challenging, especially for large-scale systems. Here, we introduce a UCB-based tree search approach to search for near-optimal JVCSs. First, verification planning is simplified as repeatable bandit problems and an upper confidence bound rule for repeatable bandits (UCBRBs) is presented with the optimal regret bound. Next, a tree search algorithm is proposed to search for feasible JVCSs. A tree-based ensemble learning model is also used to extend the tree search algorithm to handle local optimality issues. The proposed approach is evaluated on the notional case of a communication system.

A predictive analytics model for crop suitability and productivity with tree-based ensemble learning

The increasing global population has heightened the importance of agriculture and the need for food security. This study focuses on developing and evaluating tree-based ensemble learning models for predicting crop suitability and productivity to achieve Agenda Zero Hunger by 2030. The study encompasses two primary objectives: (1) exploring and analyzing the relationships between environmental factors and crop suitability and productivity and (2) developing a predictive analytics tree-based ensemble learning model for crop suitability and productivity prediction. The proposed models are evaluated using a publicly available dataset from Kaggle. The experimental results demonstrate exceptional performance with an accuracy of 99.32%, precision of 99.34%, recall of 99.39%, and F1-score of 99.34%. Furthermore, our findings reveal that the choice of crop cultivation within a specific area heavily relies on factors like rainfall and potassium levels. Our findings contribute to the advancement of agricultural practices, enabling farmers to make informed decisions regarding crop selection, ultimately striving towards improved food security and sustainable agricultural practices aligned with the Zero Hunger Agenda 2030 goals.

Development of compositional-based models for prediction of heavy crude oil viscosity: Application in reservoir simulations

The properties of crude oil are of great importance for efficient recovery of oil from oil fields. The properties are primarily used in reservoir simulations for prediction of oil recovery in order to save time and obtain the best recovery. Among various crude oil properties, viscosity is the most important one which should be precisely simulated. In this work, a novel approach based on machine learning is developed for estimation of crude oil viscosity as function of input parameters. Multiple distinct tree-based ensemble models are applied on the available dataset in this work to predict heavy-oil viscosity. AdaBoost Decision Trees (ADA-DT), Random Forest (RF), and Extremely Randomized Trees (ERT) are selected tree-based ensembles that used in this work for the simulation of oil viscosity. An Isolation Forest is applied on the dataset to remove outliers and also the earthworm optimization algorithm (EWA) is employed to find the optimum values of models’ hyper-parameters. Optimized models of ADA-DT, ERT, and RF have RMSE error rates of 35.42, 27.02, and 58.71. Thus, ERT is selected as the best model of the dataset used in this work.

Quantifying the Efficacy of Magnetic Nanoparticles for MRI and Hyperthermia Applications via Machine Learning Methods.

Magnetic nanoparticles are a prospective class of materials for use in biomedicine as agents for magnetic resonance imagining (MRI) and hyperthermia treatment. However, synthesis of nanoparticles with high efficacy is resource-intensive experimental work. In turn, the use of machine learning (ML) methods is becoming useful in materials design and serves as a great approach to designing nanomagnets for biomedicine. In this work, for the first time, an ML-based approach is developed for the prediction of main parameters of material efficacy, i.e., specific absorption rate (SAR) for hyperthermia and r1 /r2 relaxivities in MRI, with parameters of nanoparticles as well as experimental conditions as descriptors. For that, a unique database with more than 980 magnetic nanoparticles collected from scientific articles is assembled. Using this data, several tree-based ensemble models are trained to predict SAR, r1 and r2 relaxivity. After hyperparameter optimization, models reach performances of R2 = 0.86, R2 = 0.78, and R2 = 0.75, respectively. Testing the models on samples unseen during the training shows no performance drops. Finally, DiMag, an open access resource created to guide synthesis of novel nanosized magnets for MRI and hyperthermia treatment with machine learning and boost development of new biomedical agents, is developed.

Establishment and interpretation of the gamma pass rate prediction model based on radiomics for different intensity-modulated radiotherapy techniques in the pelvis

Backgroundand objectives: Implementation of patient-specific quality assurance (PSQA) is a crucial aspect of precise radiotherapy. Various machine learning-based models have showed potential as virtual quality assurance tools, being capable of accurately predicting the dose verification results of fixed-beam intensity-modulated radiation therapy (IMRT) or volumetric modulated arc therapy (VMAT) plans, thereby ensuring safe and efficient treatment for patients. However, there has been no research yet that simultaneously integrates different IMRT techniques to predict the gamma pass rate (GPR) and explain the model.Methods: Retrospective analysis of the 3D dosimetric verification results based on measurements with gamma pass rate criteria of 3%/2 mm and 10% dose threshold of 409 pelvic IMRT and VMAT plans was carried out. Radiomics features were extracted from the dose files, from which the XGBoost algorithm based on SHapley Additive exPlanations (SHAP) values was used to select the optimal feature subset as the input for the prediction model. The study employed four different machine learning algorithms, namely, random forest (RF), adaptive boosting (AdaBoost), extreme gradient boosting (XGBoost), and light gradient boosting machine (LightGBM), to construct predictive models. Sensitivity, specificity, F1 score, and AUC value were calculated to evaluate the classification performance of these models. The SHAP values were utilized to perform a related interpretive analysis on the best performing model.Results: The sensitivities and specificities of the RF, AdaBoost, XGBoost, and LightGBM models were 0.96, 0.82, 0.93, and 0.89, and 0.38, 0.54, 0.62, and 0.62, respectively. The F1 scores and area under the curve (AUC) values were 0.86, 0.81, 0.88, and 0.86, and 0.81, 0.77, 0.85, and 0.83, respectively. The explanation of the model output based on SHAP values can provide a reference basis for medical physicists when adjusting the plan, thereby improving the efficiency and quality of treatment plans.Conclusion: It is feasible to use a machine learning method based on radiomics to establish a gamma pass rate classification prediction model for IMRT and VMAT plans in the pelvis. The XGBoost model performs better in classification than the other three tree-based ensemble models, and global explanations and single-sample explanations of the model output through SHAP values may offer reference for medical physicists to provide high-quality plans, promoting the clinical application and implementation of GPR prediction models, and providing safe and efficient personalized QA management for patients.

A Hybrid Ensemble Model for Solar Irradiance Forecasting: Advancing Digital Models for Smart Island Realization

This study introduces HYTREM, a hybrid tree-based ensemble learning model conceived with the sustainable development of eco-friendly transportation and renewable energy in mind. Designed as a digital model, HYTREM primarily aims to enhance solar power generation systems’ efficiency via accurate solar irradiance forecasting. Its potential application extends to regions such as Jeju Island, which is committed to advancing renewable energy. The model’s development process involved collecting hourly solar irradiance and weather-related data from two distinct regions. After data preprocessing, input variables configuration, and dataset partitioning into training and testing sets, several tree-based ensemble learning models—including extreme gradient boosting, light gradient boosting machine, categorical boosting, and random forest (RF)—were employed to generate prediction values in HYTREM. To improve forecasting accuracy, separate RF models were constructed for each hour. Experimental results validated the superior performance of HYTREM over state-of-the-art models, demonstrating the lowest mean absolute error, root mean square error (RMSE), and normalized RMSE values across both regions. Due to its transparency and efficiency, this approach suits energy providers with limited computational resources. Ultimately, HYTREM is a stepping stone towards developing advanced digital twin systems, highlighting the importance of precise forecasting in managing renewable energy.

Classification of imbalanced data using machine learning algorithms to predict the risk of renal graft failures in Ethiopia

IntroductionThe prevalence of end-stage renal disease has raised the need for renal replacement therapy over recent decades. Even though a kidney transplant offers an improved quality of life and lower cost of care than dialysis, graft failure is possible after transplantation. Hence, this study aimed to predict the risk of graft failure among post-transplant recipients in Ethiopia using the selected machine learning prediction models.MethodologyThe data was extracted from the retrospective cohort of kidney transplant recipients at the Ethiopian National Kidney Transplantation Center from September 2015 to February 2022. In response to the imbalanced nature of the data, we performed hyperparameter tuning, probability threshold moving, tree-based ensemble learning, stacking ensemble learning, and probability calibrations to improve the prediction results. Merit-based selected probabilistic (logistic regression, naive Bayes, and artificial neural network) and tree-based ensemble (random forest, bagged tree, and stochastic gradient boosting) models were applied. Model comparison was performed in terms of discrimination and calibration performance. The best-performing model was then used to predict the risk of graft failure.ResultsA total of 278 completed cases were analyzed, with 21 graft failures and 3 events per predictor. Of these, 74.8% are male, and 25.2% are female, with a median age of 37. From the comparison of models at the individual level, the bagged tree and random forest have top and equal discrimination performance (AUC-ROC = 0.84). In contrast, the random forest has the best calibration performance (brier score = 0.045). Under testing the individual model as a meta-learner for stacking ensemble learning, the result of stochastic gradient boosting as a meta-learner has the top discrimination (AUC-ROC = 0.88) and calibration (brier score = 0.048) performance. Regarding feature importance, chronic rejection, blood urea nitrogen, number of post-transplant admissions, phosphorus level, acute rejection, and urological complications are the top predictors of graft failure.ConclusionsBagging, boosting, and stacking, with probability calibration, are good choices for clinical risk predictions working on imbalanced data. The data-driven probability threshold is more beneficial than the natural threshold of 0.5 to improve the prediction result from imbalanced data. Integrating various techniques in a systematic framework is a smart strategy to improve prediction results from imbalanced data. It is recommended for clinical experts in kidney transplantation to use the final calibrated model as a decision support system to predict the risk of graft failure for individual patients.

A machine learning and explainable artificial intelligence triage-prediction system for COVID-19

COVID-19 is a respiratory disease caused by the SARS-CoV-2 contagion, severely disrupted the healthcare infrastructure. Various countries have developed COVID-19 vaccines that have effectively prevented the severe symptoms caused by the virus to a certain extent. However, a small section of people continues to perish. Artificial intelligence advances have revolutionized healthcare diagnosis and prognosis infrastructure. In this study, we predict the severity of COVID-19 using heterogenous Machine Learning and Deep Learning algorithms by considering clinical markers, vital signs, and other critical factors. This study extensively reviews various classifier architectures to predict the COVID-19 severity. We built and evaluated multiple pipelines entailing combinations of five state-of-the-art data-balancing techniques (Synthetic Minority Oversampling Technique (SMOTE), Adaptive Synthetic, Borderline SMOTE, SMOTE with Tomek links, and SMOTE with Edited Nearest Neighbor (ENN)) and twelve heterogeneous classifiers such as Logistic Regression, Decision Tree, Random Forest, Support Vector Machine, K-Nearest Neighbors, Naïve Bayes, Xgboost, Extratrees, Adaboost, Light GBM, Catboost, and 1-D Convolution Neural Network. The best-performing pipeline consists of Random Forest trained on Borderline SMOTE balanced data that produced the highest recall of 83%. We deployed Explainable Artificial Intelligence tools such as Shapley Additive Explanations and Local Interpretable Model-agnostic Explanations, ELI5, Qlattice, Anchor, and Feature Importance to demystify complex tree-based ensemble models. These tools provide valuable insights into the significance of critical features in the severity prediction of a COVID-19 patient. It was observed that changes in respiratory rate, blood pressure, lactate, and calcium values were the primary contributors to the increase in severity of a COVID-19 patient. This architecture aims to be an explainable decision-support triaging system for medical professionals in countries lacking advanced medical technology and infrastructure to reduce fatalities.

Multicentric evaluation of a machine learning model to streamline the radiotherapy patient specific quality assurance process

PurposePatient-specific quality assurance (PSQA) is performed to ensure that modulated treatment plans can be delivered as intended, but constitutes a substantial workload that could slow down the radiotherapy process and delay the start of clinical treatments. In this study, we investigated a machine learning (ML) tree-based ensemble model to predict the gamma passing rate (GPR) for volumetric modulated arc therapy (VMAT) plans. Materials and methods5622 VMAT plans from multiple treatment sites were selected from a database of Institution 1 and the ML model trained using 19 metrics. PSQA analyses were performed automatically using criteria 3%/1 mm (global normalization, absolute dose, 10% threshold) and 95% action limit. Model’s performance was evaluated on an out-of-sample test set of Institution 1 and on two independent sets of measurements collected at Institution 2 and Institution 3. Mean absolute error (MAE), as well as the model’s sensitivity and specificity, were computed. ResultsThe model obtained a MAE of 2.33%, 2.54% and 3.91% for the three Institutions, with a specificity of 0.90, 0.90 and 0.68, and a sensitivity of 0.61, 0.25, and 0.55, respectively. Small positive median values of the residuals (i.e., the difference between measurements and predictions) were observed for each Institution (0.95%, 1.66%, and 3.42%). Thus, the model’s predictions were, on average, close to the real values and provided a conservative estimation of the GPR. ConclusionsML models can be integrated into clinical practice to streamline the radiotherapy workflow, but they should be center-specific or thoroughly verified within centers before clinical use.

Early stage diabetes prediction using decision tree-based ensemble learning model

Diabetes is a lifelong disease that has undesirable effects on various organs, such as long-term organ damage, functional disorder, and finally failure of the organ. Diabetes must be treated under the supervision of a doctor. Diabetes is known as a disease that can be seen in many people today and is becoming widespread due to life conditions. If a person with diabetes does not receive any treatment at an early stage, the patient's body can react with serious complications. In addition to the medical methods used in the diagnosis of diabetes, this disease can be detected by an artificial intelligence approach. This research aims to establish the most influential variable among the many variables causing diabetes and to design a model that will predict diabetes to help doctors analyze the disease with selected machine learning methods. In this study, Decision Tree, Bagging with Decision Tree, Random Forest and Extra Tree algorithms were used for the proposed model and the highest accuracy values were obtained with the Extra Trees algorithm with 99.2%.

Comparison of tree-based model with deep learning model in predicting effluent pH and concentration by capacitive deionization

Capacitive deionization (CDI) is an emerging technique for water treatment and electroadsorption processes (i.e., brackish water desalination). Various numerical modeling methods have been developed to predict and optimize the performance of CDI, and artificial intelligence techniques have recently been applied to overcome the limitations of numerical modelings, such as the difficulty in handling all complexities in the environment. However, such a complex neural network (i.e., deep learning (DL)) has limitations in that it is difficult to design a structure, takes a long time to train, and requires massive computer resources. Therefore, in this study, a tree-based model that is more effective than a neural network model for processing tabular data was developed to predict effluent pH and concentration in the CDI process. The tree-based ensemble models had a remarkably lower computational cost (100 times less than the DL model) with almost the same prediction accuracy (R2 = 0.998 for the steady random forest model and R2 = 0.986 for the DL model) using a binary feature concept. These findings will contribute to further examining the use of tree-based models for predicting and optimizing the CDI process to reduce computing capacity and minimize modeling complexity.

Decision Tree-Based Ensemble Model for Predicting National Greenhouse Gas Emissions in Saudi Arabia

Greenhouse gas (GHG) emissions must be precisely estimated in order to predict climate change and achieve environmental sustainability in a country. GHG emissions are estimated using empirical models, but this is difficult since it requires a wide variety of data and specific national or regional parameters. In contrast, artificial intelligence (AI)-based methods for estimating GHG emissions are gaining popularity. While progress is evident in this field abroad, the application of an AI model to predict greenhouse gas emissions in Saudi Arabia is in its early stages. This study applied decision trees (DT) and their ensembles to model national GHG emissions. Three AI models, namely bagged decision tree, boosted decision tree, and gradient boosted decision tree, were investigated. Results of the DT models were compared with the feed forward neural network model. In this study, population, energy consumption, gross domestic product (GDP), urbanization, per capita income (PCI), foreign direct investment (FDI), and GHG emission information from 1970 to 2021 were used to construct a suitable dataset to train and validate the model. The developed model was used to predict Saudi Arabia’s national GHG emissions up to the year 2040. The results indicated that the bagged decision tree has the highest coefficient of determination (R2) performance on the testing dataset, with a value of 0.90. The same method also has the lowest root mean square error (0.84 GtCO2e) and mean absolute percentage error (0.29 GtCO2e), suggesting that it exhibited the best performance. The model predicted that GHG emissions in 2040 will range between 852 and 867 million tons of CO2 equivalent. In addition, Shapley analysis showed that the importance of input parameters can be ranked as urbanization rate, GDP, PCI, energy consumption, population, and FDI. The findings of this study will aid decision makers in understanding the complex relationships between the numerous drivers and the significance of diverse socioeconomic factors in defining national GHG inventories. The findings will enhance the tracking of national GHG emissions and facilitate the concentration of appropriate activities to mitigate climate change.

Multi-objective optimization of fly ash-slag based geopolymer considering strength, cost and CO2 emission: A new framework based on tree-based ensemble models and NSGA-II

Fly ash-slag based geopolymer has excellent mechanical performance with low carbon footprints, which has emerged as a promising alternative to Portland cement. The optimization of geopolymers requires trade-offs between multiple objectives (strength, cost, and CO2 emission) while considering a large number of highly nonlinear variables. To solve this multi-objective optimization (MOO) problem, this study developed a MOO model combining the tree-based ensemble learning algorithm and non-dominated sorting genetic algorithm (NSGA-II). The MOO model was trained on a database collected from published literature with 676 mixture proportions. The chemical components as well as their contents and the curing conditions were selected as the input variables, while the uniaxial compressive strength (UCS) was the output variable. The UCS of fly ash-slag based geopolymers were modeled using the random forest regressor, extra trees regressor, gradient boosted regressor and extreme gradient boosting regressor. The results show that the gradient boosted regressor has the highest prediction accuracy with R (0.966) and RMSE (5.295 MPa) on the testing set. The analysis of the variable importance by the developed model indicates that the curing age, contents of slag and sodium silicate are more important to the strength development of the binary blended geopolymer. The MOO model developed based on the gradient-boosted regressor and NSGA-II successfully found the cost-UCS and CO2 emission-UCS Pareto fronts of the bi-objective optimization problem, as well as the cost-UCS-CO2 emission Pareto front of the tri-objective optimization problem. This developed framework improves the efficiency of geopolymer design and can be applied to the mixture optimization of other construction materials.

Efficient model for detecting application layer distributed denial of service attacks

The increasing advancement of technologies and communication infrastructures has been posing threats to the internet services. One of the most powerful attack weapons for disrupting web-based services is the distributed denial of service (DDoS) attack. The sophisticated nature of attack tools being created and used for launching attacks on target systems makes it difficult to distinguish between normal and attack traffic. Consequently, there is a need to detect application layer DDoS attacks from network traffic efficiently. This paper proposes a detection system coined eXtreme gradient boosting (XGB-DDoS) using a tree-based ensemble model known as XGBoost to detect application layer DDoS attacks. The Canadian institute for cybersecurity intrusion detection systems (CIC IDS) 2017 dataset consisting of both benign and malicious attacks was used in training and testing of the proposed model. The performance results of the proposed model indicate that the accuracy rate, recall, precision rate, and F1-score of XGB-DDoS are 0.999, 0.997, 0.995, and 0.996, respectively, as against those of k-nearest neighbor (KNN), support vector machine (SVM), principal component analysis (PCA) hybridized with XGBoost, and KNN with SVM. So, the XGB-DDoS detection model did better than the models that were chosen. This shows that it is good at finding application layer DDoS attacks.

Cardiorespiratory Mortality Prediction Based on Air Pollution Using Tree-Based Ensemble Models

Air pollution has a substantial negative impact on human wellbeing and health. Cardiorespiratory mortality is one of the primary effects of air pollution. In this study, we provide analysis of air pollution, cardiorespiratory mortality and the cardiorespiratory mortality is predicted based on air pollution using tree-based ensemble models. The tree-based ensemble models utilized in this study are Voting Regressor (VR), Random Forest (RF), Gradient Tree Boosting (GB), and Extreme Gradient Boosting (XGBoost). The used dataset contains data for five research locations: Shah Alam (SA), Klang (KLN), Putrajaya (PUJ), Cheras, Kuala Lumpur (CKL), and Petaling Jaya (PJ) from January 2006 to December 2016. The results show that XGBoost and VR models outperformed the rest of the models with the best evaluation metric scores in the Klang study area, XGBoost(MAE:0.005, RMSE:0.010, MAPE:0.70%) and VR (MAE:0.005, RMSE:0.011, MAPE:0.70%). The results reveal that the utilized models provided an excellent and accurate prediction of cardiorespiratory mortality based on air pollution and can follow the trend of cardiorespiratory mortality.

Fault classification and timing prediction based on shipment inspection data and maintenance reports for semiconductor manufacturing equipment

Unexpected failures of equipment in sequential and continuous manufacturing processes such as semiconductor production can result in a serious deterioration of productivity. This directly affects the interests of equipment operators and indirectly reduces the credibility of the equipment manufacturers, resulting in serious potential harm to operators and manufacturers of equipment. Hence, the demand for predictive maintenance has increased to maximize the efficiency of equipment operations by predicting faults and fault timing beyond preemptive preventive maintenance, which depends on periodic inspections while operating equipment. However, since existing studies on predictive maintenance mainly utilize data collected during the equipment operation stage to predict equipment faults and fault timing, predictive maintenance research has only been conducted from the perspective of equipment operating companies. Accordingly, we propose a machine learning method in this paper that predicts the existence of faults and the fault timing of equipment using pre-shipment inspection data from equipment manufacturers. Importantly, this can be performed by an equipment manufacturer without access to operational data. This is the first predictive maintenance study to be conducted from the perspective of an equipment manufacturing company. We used tree-based ensemble models to identify important variables for faults and proceeded to predict the existence of faults and the first fault timing of equipment. Moreover, we verified the method’s performance using real data from a leading semiconductor equipment manufacturer in South Korea. The contribution of this study can be summarized as follows. (1) We propose an initial predictive maintenance method that can be performed by equipment manufacturing companies. (2) As a result of an experiment using real data from the selected manufacturer, the proposed method yielded an accuracy of up to 94% for fault existence prediction and reduced the prediction error by more than 50% for the first fault timing compared to where a prediction model was not implemented. (3) It was demonstrated that the initial state of a piece of equipment has a great influence on future faults by deriving high performance using pre-shipment inspection data rather than operation data. (4) By using tree-based ensemble models, we identified major factors related to equipment faults.

Interpretable Soft Sensors using Extremely Randomized Trees and SHAP

Tree-based ensemble models are easy to implement and have been widely used in various fields. However, they have limitations in industrial process applications since the majority of tree-based ensemble models are prone to over-fitting. In addition, the internal structure of tree-based ensemble models is very complex and the output of the model is also difficult to explain, which makes its application in industrial soft sensors very challenging. The purpose of this work is to build accurate and interpretable soft sensors for industrial processes. First, to deal with overfitting, a robust tree-based ensemble model and extremely randomized trees are used to build accurate soft sensors. Then, to improve model interpretability, an interpretable machine learning algorithm, namely Shapely additive explanation, is used to infer the global and local contributions of each feature to the predictions. Finally, the effectiveness of the proposed algorithms is validated on real industrial fluid catalytic cracker unit data.

XML-CIMT: Explainable Machine Learning (XML) Model for Predicting Chemical-Induced Mitochondrial Toxicity

Organ toxicity caused by chemicals is a serious problem in the creation and usage of chemicals such as medications, insecticides, chemical products, and cosmetics. In recent decades, the initiation and development of chemical-induced organ damage have been related to mitochondrial dysfunction, among several adverse effects. Recently, many drugs, for example, troglitazone, have been removed from the marketplace because of significant mitochondrial toxicity. As a result, it is an urgent requirement to develop in silico models that can reliably anticipate chemical-induced mitochondrial toxicity. In this paper, we have proposed an explainable machine-learning model to classify mitochondrially toxic and non-toxic compounds. After several experiments, the Mordred feature descriptor was shortlisted to be used after feature selection. The selected features used with the CatBoost learning algorithm achieved a prediction accuracy of 85% in 10-fold cross-validation and 87.1% in independent testing. The proposed model has illustrated improved prediction accuracy when compared with the existing state-of-the-art method available in the literature. The proposed tree-based ensemble model, along with the global model explanation, will aid pharmaceutical chemists in better understanding the prediction of mitochondrial toxicity.

Toward explainable electrical load forecasting of buildings: A comparative study of tree-based ensemble methods with Shapley values

Electrical load forecasting of buildings is crucial in designing an energy operation strategy for smart city realization. Although artificial intelligence techniques have demonstrated excellent energy forecasting performance, explaining how outcomes are obtained when they are inaccurate is challenging. Explainable artificial intelligence (XAI) has recently received considerable attention in addressing this issue. This study proposes an explainable electrical load forecasting (XELF) methodology. We first preprocess data for input variable configuration and build the following tree-based ensemble models that produce outstanding results in tabular data: random forest, gradient boosting machine (GBM), extreme gradient boosting, light GBM (LightGBM), and categorical boosting. We evaluate performance in terms of the mean absolute percentage error, coefficient of variation of the root mean square error, and normalized mean absolute error. Finally, we provide the rationale for interpreting the influence of the input variables and decision-making process via Shapley additive explanations (SHAP), an XAI technique, on the best model. The experiments were conducted with an electrical load dataset from educational buildings to validate the practicality and validity of this methodology. We applied the SHAP to the LightGBM model and performed its respective analyses and visualizations for XELF.