Combination Of Machine Learning Approaches Research Articles

Rapid natural hazard extent estimation from twitter data: investigation for hurricane impact areas

Natural hazards have occurred more frequently in the past years and pose a severe risk to human life. Their extents and, thereby, the most heavily affected areas must be estimated at the earliest to limit damages or initiate rescue services. For such estimations, a widely available data source, which is comparatively responsive to short-time changes, is needed and provided by volunteered geographic information (VGI) data. Tropical cyclones are natural hazard events that can cause enormous spatially extended damage. In this study, we introduce Machine Learning approaches such as Extremely Randomized Tree (ET) and Geographically Weighted Regression for estimating hurricane-impacted regions from VGI data. In addition to the general approximate track extent estimation, we also evaluate the possibilities of temporal estimation of track development from VGI data. Different scenarios are evaluated, and we find that the results mainly depend on the choice of the geographical splits for training and test data for the underlying regression task. Suitable splits lead to R2\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$R^2$$\\end{document} of 99% in the best cases with the ET model. The estimation results are satisfying when considering the temporal aspect and represent a use-case scenario. Such a combination of Machine Learning approaches and VGI is a simple and fast approach for early natural hazard estimation.

Use of interpretive machine learning and a crop model to investigate the impact of environment and management on soybean yield gap

Context Management and environmental conditions are the main factors influencing yield of soybean (Glycine max (L.) Merr.). Despite an increase in average soybean yield in recent years in Iran, a considerable gap remains between actual yield and potential yield. Aims The objective of this study was to identify critical climate and management factors affecting soybean yield in Iran’s major soybean production area. Methods A combination of machine learning approaches (using gradient boosted decision trees, XGBoost) and the SSM-iCrop2 simulation model was used. Critical management factors affecting soybean yield were determined through interpretive machine learning using information collected from 268 soybean fields over a 5-year period. Potential yield and water-limited potential yield at six weather stations were estimated for 30 years via the SSM-iCrop2 simulation model. Water limitation was determined by considering the ratio of water-limited yield potential to potential yield, and heat stress status was quantified as the number of days with maximum temperature >36°C during the soybean growing season. Key results The XGBoost models adequately described the observed changes in soybean yield. Root-mean-square error and Lin’s concordance correlation coefficient values of the calibrated model were 262 kg ha−1 and 0.96, respectively, which indicated that the predictor variables could describe most of the variation in soybean yield for the studied dataset. Conclusions We identified 15 climatic and management variables that affect soybean yield. A large part of the studied area is under high water stress and low heat stress. Implications Optimal planting date and improved irrigation management are the main options for reducing the yield gap in the study area.

Automated Assessment of T2-Weighted MRI to Differentiate Malignant and Benign Primary Solid Liver Lesions in Noncirrhotic Livers Using Radiomics

Distinguishing malignant from benign liver lesions based on magnetic resonance imaging (MRI) is an important but often challenging task, especially in noncirrhotic livers. We developed and externally validated a radiomics model to quantitatively assess T2-weighted MRI to distinguish the most common malignant and benign primary solid liver lesions in noncirrhotic livers. Data sets were retrospectively collected from three tertiary referral centers (A, B, and C) between 2002 and 2018. Patients with malignant (hepatocellular carcinoma and intrahepatic cholangiocarcinoma) and benign (hepatocellular adenoma and focal nodular hyperplasia) lesions were included. A radiomics model based on T2-weighted MRI was developed in data set A using a combination of machine learning approaches. The model was internally evaluated on data set A through cross-validation, externally validated on data sets B and C, and compared to visual scoring of two experienced abdominal radiologists on data set C. The overall data set included 486 patients (A: 187, B: 98, and C: 201). The radiomics model had a mean area under the curve (AUC) of 0.78 upon internal validation on data set A and a similar AUC in external validation (B: 0.74 and C: 0.76). In data set C, the two radiologists showed moderate agreement (Cohen's κ: 0.61) and achieved AUCs of 0.86 and 0.82. Our T2-weighted MRI radiomics model shows potential for distinguishing malignant from benign primary solid liver lesions. External validation indicated that the model is generalizable despite substantial MRI acquisition protocol differences. Pending further optimization and generalization, this model may aid radiologists in improving the diagnostic workup of patients with liver lesions.

How social media expression can reveal personality

Personality psychology studies personality and its variation among individuals and is an essential branch of psychology. In recent years, machine learning research related to personality assessment has started to focus on the online environment and showed outstanding performance in personality assessment. However, the aspects of the personality of these prediction models measure remain unclear because few studies focus on the interpretability of personality prediction models. The objective of this study is to develop and validate a machine learning model with domain knowledge introduced to enhance accuracy and improve interpretability. Study participants were recruited via an online experiment platform. After excluding unqualified participants and downloading the Weibo posts of eligible participants, we used six psycholinguistic and mental health-related lexicons to extract textual features. Then the predictive personality model was developed using the multi-objective extra trees method based on 3,411 pairs of social media expression and personality trait scores. Subsequently, the prediction model's validity and reliability were evaluated, and each lexicon's feature importance was calculated. Finally, the interpretability of the machine learning model was discussed. The features from Culture Value Dictionary were found to be the most important predictors. The fivefold cross-validation results regarding the prediction model for personality traits ranged between 0.44 and 0.48 (p < 0.001). The correlation coefficients of five personality traits between the two "split-half" datasets data ranged from 0.84 to 0.88 (p < 0.001). Moreover, the model performed well in terms of contractual validity. By introducing domain knowledge to the development of a machine learning model, this study not only ensures the reliability and validity of the prediction model but also improves the interpretability of the machine learning method. The study helps explain aspects of personality measured by such prediction models and finds a link between personality and mental health. Our research also has positive implications regarding the combination of machine learning approaches and domain knowledge in the field of psychiatry and its applications to mental health.

Charging Coordination of Plug-in Electric Vehicles Considering Machine Learning Based State-of-Charge Prediction for Congestion Management in Distribution System

This article proposes a novel strategy for congestion mitigation in a distribution system by charging coordination strategy of plug-in electric vehicle (PEV) considering grid-to-vehicle (G2V) and vehicle-to-grid (V2G) modes. The G2V mode of a large figure of PEVs creates a congestion scenario in the distribution system. Therefore, a coordinated charging strategy has been considered in this work to mitigate distribution system congestion. Moreover, a precise estimation and prediction of PEVs state-of-charge (SOC) is necessary while formulating PEVs coordinated strategy. The study of the work is two folded. First, for the first time, a combination of machine learning approach like gradient boosting method-Bayesian optimization (GBM-BO) is considered in prediction of PEVs SOC at the finishing of trip. Second, a coordinated charging scheme is established based on particle swarm optimization (PSO) and firefly algorithm (FA) using the PEVs SOC at the finishing of trip. The charging coordination strategy is analyzed on the 38-bus radial distribution system integrated with solar powered charging-cum parking lot (SPCPL). The machine learning based prediction results reveal the significant reduction in errors between predicted and calculated values. The results further reveal the reduction in PEVs charging cost and congestion scenario while considering SPCPL in the distribution system.

Identification of decompensation episodes in chronic heart failure patients based solely on heart sounds.

Decompensation episodes in chronic heart failure patients frequently result in unplanned outpatient or emergency room visits or even hospitalizations. Early detection of these episodes in their pre-symptomatic phase would likely enable the clinicians to manage this patient cohort with the appropriate modification of medical therapy which would in turn prevent the development of more severe heart failure decompensation thus avoiding the need for heart failure-related hospitalizations. Currently, heart failure worsening is recognized by the clinicians through characteristic changes of heart failure-related symptoms and signs, including the changes in heart sounds. The latter has proven to be largely unreliable as its interpretation is highly subjective and dependent on the clinicians' skills and preferences. Previous studies have indicated that the algorithms of artificial intelligence are promising in distinguishing the heart sounds of heart failure patients from those of healthy individuals. In this manuscript, we focus on the analysis of heart sounds of chronic heart failure patients in their decompensated and recompensated phase. The data was recorded on 37 patients using two types of electronic stethoscopes. Using a combination of machine learning approaches, we obtained up to 72% classification accuracy between the two phases, which is better than the accuracy of the interpretation by cardiologists, which reached 50%. Our results demonstrate that machine learning algorithms are promising in improving early detection of heart failure decompensation episodes.

Predictive machine learning model for microvascular invasion identification in hepatocellular carcinoma based on the LI-RADS system.

This study aimed to establish a predictive model of microvascular invasion (MVI) in hepatocellular carcinoma (HCC) by contrast-enhanced computed tomography (CT), which relied on a combination of machine learning approach and imaging features covering Liver Imaging and Reporting and Data System (LI-RADS) features. The retrospective study included 279 patients with surgery who underwent preoperative enhanced CT. They were randomly allocated to training set, validation set, and test set (167 patients vs. 56 patients vs. 56 patients, respectively). Significant imaging findings for predicting MVI were identified through the Least Absolute Shrinkage and Selection Operator (LASSO) logistic regression method. Predictive models were performed by machine learning algorithm, support vector machine (SVM), in the training set and validation set, and evaluated in the test set. Further, a combined model adding clinical findings to the radiologic model was developed. Based on the LI-RADS category, subgroup analyses were conducted. We included 116 patients with MVI which were diagnosed through pathological confirmation. Six imaging features were selected about MVI prediction: four LI-RADS features (corona enhancement, enhancing capsule, non-rim aterial phase hyperehancement, tumor size) and two non-LI-RADS features (internal arteries, non-smooth tumor margin). The radiological feature with the best accuracy was corona enhancement followed by internal arteries and tumor size. The accuracies of the radiological model and combined model were 0.725-0.714 and 0.802-0.732 in the training set, validation set, and test set, respectively. In the LR-4/5 subgroup, a sensitivity of 100% and an NPV of 100% were obtained by the high-sensitivity threshold. A specificity of 100% anda PPV of 100% were acquired through the high specificity threshold in the LR-M subgroup. A combination of LI-RADS features and non-LI-RADS features and serum alpha-fetoprotein value could be applied as a preoperative biomarker for predicting MVI by the machine learning approach. Furthermore, its good performance in the subgroup by LI-RADS category may help optimize the management of HCC patients.

Machine learning to predict in-hospital mortality risk among heterogenous STEMI patients with diabetes

Abstract Funding Acknowledgements Type of funding sources: Public grant(s) – National budget only. Main funding source(s): TECHNOLOGY DEVELOPMENT FUND 1 Background Diabetes has become a major public health concern in Asia. In Malaysia, the prevalence of diabetes has escalated in adults above the age of 18, affecting 3.9 million individuals. Patients with diabetes and coronary heart disease have worse outcomes, compared with patients without diabetes who have coronary heart disease. Conventional Risk scores such as TIMI and GRACE were derived from a Western Caucasian cohort with limited data from Asian countries, despite Asia hosting 60% of the world’s population. Purpose It is important to recognize the significant features associated with in-hospital mortality risk that is population-specific in Asian diabetes patients with STEMI to achieve a reliable and effective clinical diagnosis and improved outcome. Electronic health records contain large amounts of information on patients’ medical history and are becoming invaluable research tools that could be applied to cardiovascular disease risk prediction through machine learning (ML) algorithms. With the current success of ML over conventional methods in STEMI mortality prediction, we aim to develop ML algorithms for in-hospital risk mortality in Asian patients diagnosed with DM that can be adopted for clinical predictions Methods We used registry data from the Malaysian National Cardiovascular Disease Database of 5783 patients diagnosed with DM from 2006 to 2016. Fifty parameters including demographics, cardiovascular risk, medications and clinical variables were considered. Four machine learning (ML) algorithms were constructed using a 70% registry dataset; Random Forest (RF), Support Vector Machine (SVM), Extreme Gradient Booster (XGB) and Logistic Regression (LR). Feature selections were done based on ML algorithms feature importance combined with Sequential Backward Selection (SBS). The area under the curve (AUC) was used as the performance evaluation metric. All algorithms were validated using a 30 % validation dataset and compared to the conventional TIMI risk score for STEMI. Results The best model SVM (AUC = 0.90) outperformed other ML algorithms (Figure 1) and TIMI risk score (AUC = 0.83). The best SVM model consists of 11 predictors which are Killip class, fasting blood glucose, age, systolic blood pressure, heart rate, ACE inhibitor, beta-blocker, total cholesterol, diastolic blood pressure, lower density lipoprotein, and diuretic (Figure 2). Common predictors of SVM and TIMI risk score are Killip class, age, systolic blood pressure, and heart rate. We have shown that the population-specific data mining approach for the prediction of diabetes patients’ mortality post-STEMI outperformed conventional TIMI risk score. Conclusion In the Asian multiethnic population, combination of ML approaches with features selection demonstrated promising outcomes in patients with DM that may be used for better patient prognostic than the conventional method. Abstract Figure 1: ML Best Model Performance Abstract Figure 2: Selected Predictors for ML

Differential Diagnosis and Molecular Stratification of Gastrointestinal Stromal Tumors on CT Images Using a Radiomics Approach

Treatment planning of gastrointestinal stromal tumors (GISTs) includes distinguishing GISTs from other intra-abdominal tumors and GISTs’ molecular analysis. The aim of this study was to evaluate radiomics for distinguishing GISTs from other intra-abdominal tumors, and in GISTs, predict the c-KIT, PDGFRA, BRAF mutational status, and mitotic index (MI). Patients diagnosed at the Erasmus MC between 2004 and 2017, with GIST or non-GIST intra-abdominal tumors and a contrast-enhanced venous-phase CT, were retrospectively included. Tumors were segmented, from which 564 image features were extracted. Prediction models were constructed using a combination of machine learning approaches. The evaluation was performed in a 100 × random-split cross-validation. Model performance was compared to that of three radiologists. One hundred twenty-five GISTs and 122 non-GISTs were included. The GIST vs. non-GIST radiomics model had a mean area under the curve (AUC) of 0.77. Three radiologists had an AUC of 0.69, 0.76, and 0.84, respectively. The radiomics model had an AUC of 0.52 for c-KIT, 0.56 for c-KIT exon 11, and 0.52 for the MI. The numbers of PDGFRA, BRAF, and other c-KIT mutations were too low for analysis. Our radiomics model was able to distinguish GISTs from non-GISTs with a performance similar to three radiologists, but less observer dependent. Therefore, it may aid in the early diagnosis of GIST, facilitating rapid referral to specialized treatment centers. As the model was not able to predict any genetic or molecular features, it cannot aid in treatment planning yet.

A multi-stage machine learning model for diagnosis of esophageal manometry

High-resolution manometry (HRM) is the primary procedure used to diagnose esophageal motility disorders. Its manual interpretation and classification, including evaluation of swallow-level outcomes and then derivation of a study-level diagnosis based on Chicago Classification (CC), may be limited by inter-rater variability and inaccuracy of an individual interpreter. We hypothesized that an automatic diagnosis platform using machine learning and artificial intelligence approaches could be developed to accurately identify esophageal motility diagnoses. Further, a multi-stage modeling framework, akin to the step-wise approach of the CC, was utilized to leverage advantages of a combination of machine learning approaches including deep-learning models and feature-based models. Models were trained and tested using a dataset comprised of 1741 patients' HRM studies with CC diagnoses assigned by expert physician raters. In the swallow-level stage, three models based on convolutional neural networks (CNNs) were developed to predict swallow type and swallow pressurization (test accuracies of 0.88 and 0.93, respectively), and integrated relaxation pressure (IRP)(regression model with test error of 4.49 mmHg). At the study-level stage, model selection from families of the expert-knowledge-based rule models, xgboost models and artificial neural network(ANN) models were conducted. A simple model-agnostic strategy of model balancing motivated by Bayesian principles was utilized, which gave rise to model averaging weighted by precision scores. The averaged (blended) models and individual models were compared and evaluated, of which the best performance on test dataset is 0.81 in top-1 prediction, 0.92 in top-2 predictions. This is the first artificial-intelligence style model to automatically predict esophageal motility (CC) diagnoses from HRM studies using raw multi-swallow data and it achieved high accuracy. Thus, this proposed modeling framework could be broadly applied to assist with HRM interpretation in a clinical setting.

Machine learning for characterizing growth in tourism employment in developing economies: an assessment of tourism employment in Sri Lanka

ABSTRACT Understanding the influence of tourism-linked factors on direct and indirect employment is important for tourism planning, particularly for tourism-dependent developing economies. Yet, related studies on developing countries are scant. This research considers trends of tourism growth in Sri Lanka over 1972–2018 using state-of-the-art machine learning methods: Classification and Regression Tree (CART), Boruta, hyperparameter tuning, grid search, novel robustness check strategies, and Random Forest. Our analysis confirmed that the growth in both direct and indirect tourism employment in Sri Lanka is influenced by three factors – total tourist arrivals, tourism receipts, and arrivals in the last quarter. The findings also reveal a notable seasonality impact on tourism employment, especially the growth of arrivals during the fourth quarter, for the country. Random Forest models suggest that an increase of tourist arrivals during the fourth quarter can largely compensate any detrimental impact on the growth of direct and indirect employment from a decrease in total tourist arrivals and tourism receipts. Overall, the article demonstrates that a systematic combination of machine learning approaches can provide rich insights from macro-level tourism statistics reported by tourism authorities, which in turn can guide policy formulation to boost tourism in the post-COVID-19 era.

Distinguishing pure histopathological growth patterns of colorectal liver metastases on CT using deep learning and radiomics: a pilot study

Histopathological growth patterns (HGPs) are independent prognosticators for colorectal liver metastases (CRLM). Currently, HGPs are determined postoperatively. In this study, we evaluated radiomics for preoperative prediction of HGPs on computed tomography (CT), and its robustness to segmentation and acquisition variations. Patients with pure HGPs [i.e. 100% desmoplastic (dHGP) or 100% replacement (rHGP)] and a CT-scan who were surgically treated at the Erasmus MC between 2003–2015 were included retrospectively. Each lesion was segmented by three clinicians and a convolutional neural network (CNN). A prediction model was created using 564 radiomics features and a combination of machine learning approaches by training on the clinician’s and testing on the unseen CNN segmentations. The intra-class correlation coefficient (ICC) was used to select features robust to segmentation variations; ComBat was used to harmonize for acquisition variations. Evaluation was performed through a 100 × random-split cross-validation. The study included 93 CRLM in 76 patients (48% dHGP; 52% rHGP). Despite substantial differences between the segmentations of the three clinicians and the CNN, the radiomics model had a mean area under the curve of 0.69. ICC-based feature selection or ComBat yielded no improvement. Concluding, the combination of a CNN for segmentation and radiomics for classification has potential for automatically distinguishing dHGPs from rHGP, and is robust to segmentation and acquisition variations. Pending further optimization, including extension to mixed HGPs, our model may serve as a preoperative addition to postoperative HGP assessment, enabling further exploitation of HGPs as a biomarker.

Predicting symptomatic mesenteric mass in small intestinal neuroendocrine tumors using radiomics.

Metastatic mesenteric masses of small intestinal neuroendocrine tumors (SI-NETs) are known to often cause intestinal complications. The aim of this study was to identify patients at risk to develop these complications based on routinely acquired CT scans using a standardized set of clinical criteria and radiomics. Retrospectively, CT scans of SI-NET patients with a mesenteric mass were included and systematically evaluated by five clinicians. For the radiomics approach, 1128 features were extracted from segmentations of the mesenteric mass and mesentery, after which radiomics models were created using a combination of machine learning approaches. The performances were compared to a multidisciplinary tumor board (MTB). The dataset included 68 patients (32 asymptomatic, 36 symptomatic). The clinicians had AUCs between 0.62 and 0.85 and showed poor agreement. The best radiomics model had a mean AUC of 0.77. The MTB had a sensitivity of 0.64 and specificity of 0.68. We conclude that systematic clinical evaluation of SI-NETs to predict intestinal complications had a similar performance than an expert MTB, but poor inter-observer agreement. Radiomics showed a similar performance and is objective, and thus is a promising tool to correctly identify these patients. However, further validation is needed before the transition to clinical practice.

Identification of osteosarcoma-associated antigen: A combination of machine learning approaches

Identification of osteosarcoma-associated antigen: A combination of machine learning approaches

The levelized cost of negative CO2 emissions from thermochemical conversion of biomass coupled with carbon capture and storage

Negative emissions technologies (NETs) that remove and store CO2 from the atmosphere will play crucial roles in mitigating greenhouse gas emissions. Bioenergy with carbon capture and storage (BECCS) that combines biomass conversion for energy production and CO2 capture and storage is a NET that has been intensively studied to meet climate goals. This study evaluates the feasibility of four thermochemical conversion technologies (e.g., hydrothermal treatment, pyrolysis, gasification, and conventional combustion) of selected feedstocks (e.g., crop residues, woody wastes, and biosolid) coupled with carbon capture and storage as NETs. A combination of machine learning approaches, life cycle assessment (LCA), and economic analysis is used to assess the environmental and economic performances of pathways comprising different combinations of technologies, feedstock properties, and reaction conditions. Levelized cost of negative CO2 emissions (LCNC), which is defined as the carbon incentives that are required for a NET to break even, is computed for each pathway to provide a clearer picture of the economic feasibility and carbon mitigation potential of evaluated NETs. Our results show that slow pyrolysis of woody wastes and crop residues constitute economically viable NETs even without carbon incentives. Other pathways, on the other hand, require carbon incentives to make projects break even. A case study in Virginia reveals the carbon sequestration and economic burdens of evaluated NETs and suggests that woody wastes and crop residue combustion coupled with CCS could effectively contribute to most of the negative carbon sequestration in Virginia (98%). In contrast, hydrothermal treatment of biosolids and slow pyrolysis of crop residues have marginal contributions due to high LCNC and the small market size of products, respectively. Use of all available waste biomasses as feedstocks would correspond to 11 million metric ton CO2 eq sequestration and $1.09 billion economic burdens (without carbon credits) annually, with an average LCNC of $100/t CO2 eq in Virginia.

An Efficient Machine Learning Techniques Based on IoT for Effective Load Forecasting

Internet of Things (IoT) networks (IoT) are computer networks which have an acute IT protection problem and in particular a computer attack detection problem. In order to solve this issue, the paper recommends the combination of machine learning approaches and concurrent data processing. The framework is developed and a new approach to the combination of the main classifiers intended for attacks on IoT networks. In which the accuracy ratio to the training time is the integral measure of efficacy, the problem classification statement is developed. We recommend the use of the data processing and multithreaded mode provided by Spark to accelerate the speed of training and testing. In addition, a technique is suggested for preprocessing data set, which results in a large reduction in the sample length. An experimental examination of the proposed approach reveals that the precision of IoT networks’ attack detection is 100 percent and the processing speed of data sets increases proportionally to the number of parallel threads.

A machine learning-based approach to decipher multi-etiology of knee osteoarthritis onset and deterioration

ObjectivesBy deploying a novel combination of machine learning approaches, we aim to investigate the contributions of each local and systemic risk factors in multi-etiology of knee osteoarthritis (KOA) to disease onset and deterioration. MethodsA machine-learning-based KOA progression prediction model is developed using the data from the National Institute of Health Osteoarthritis Biomarkers Consortium. According to Kellgren-Lawrence (KL) grade of plain radiographs at baseline, the subjects are divided into either KOA onset or deterioration study groups. The disease progression is defined as the changes in both joint space width (JSW) and WOMAC pain score. In addition to radiographic and symptomatic data, the anthropological particulars, history of the knee injury and surgery, metabolic syndrome and living habits were deployed in a multi-layer perceptron (MLP) to predict disease progression in each study group. The relative contributions of each risk factors were weighted via DeepLIFT gradient. Additionally, statistical interactions among risk factors were identified compared. ResultsOur model achieved AUC of 0.843 (95% CI 0.824, 0.862) and 0.765 (95% CI 0.756, 0.774) in prediction of KOA onset and deterioration, respectively. For KOA onset prediction, history of injury has attained the highest DeepLIFT gradient except medial joint space narrowing; while for KOA deterioration prediction, diabetes and habit of smoking obtained second and third highest gradients respectively aside from medial joint space narrowing, surpassing the impact of the injury. ConclusionWe developed a machine learning workflow which effectively dissects the risk factors’ contributions and their mutual interactions for onset and deterioration of KOA respectively.

White-box Machine learning approaches to identify governing equations for overall dynamics of manufacturing systems: A case study on distillation column

Dynamical equations form the basis of design for manufacturing processes and control systems; however, identifying governing equations using a mechanistic approach is tedious. Recently, Machine learning (ML) has shown promise to identify the governing dynamical equations for physical systems faster. This possibility of rapid identification of governing equations provides an exciting opportunity for advancing dynamical systems modeling. However, applicability of the ML approach in identifying governing mechanisms for the dynamics of complex systems relevant to manufacturing has not been tested. We test and compare the efficacy of two white-box ML approaches (SINDy and SymReg) for predicting dynamics and structure of dynamical equations for overall dynamics in a distillation column. Results demonstrate that a combination of ML approaches should be used to identify a full range of equations. In terms of physical law, few terms were interpretable as related to Fick’s law of diffusion and Henry’s law in SINDy, whereas SymReg identified energy balance as driving dynamics.

Development of a tailored combination of machine learning approaches to model volumetric soil water content within a mesic forest in the Pacific Northwest

To monitor the complex interdependency between soil moisture and productivity of Douglas-fir forests in the Pacific Northwest a dense low-cost network of soil moisture sensors was installed in association with 33 forest vegetation plots in a single 2260-ha watershed in western Oregon. Four sensors were installed in each plot to record volumetric soil water content and soil temperature 5 and 50 cm below the surface of mineral soil at 2-hr intervals from 2012 through 2018.Significant data gaps restricted the full utilization of the resulting dataset summarized as daily averages. A Self-Organizing Map neural network was applied to find clusters of plots with similar functional dependencies of soil temperature and soil moisture values. The results guided the subsequent assignment of the most suitable time series as input for plot-specific Distributed Time Delay Neural Networks that were trained to simulate the time series of volumetric water content in the two soil depths at 33 locations. In addition to contemporaneous inputs used in conventional neural networks, the application of Distributed Time Delay Neural Networks allows the incorporation of prior sections of the input time series data. The results show the high performance of the combined machine learning approaches that capitalize on the spatial data richness of the dense monitoring network. The correlation coefficient for the test data subsets averages 0.95 and the mean error equals 0.04 cm3/cm3. The correlation between time series at various plots was retained for the modeled data. A cross-comparison of the data revealed that 81% of the 5-cm soil moisture data and 95% of the 50-cm soil moisture data show less than 10% difference between the coefficients of determination prior to and after the gap-filling procedure. The long-term gap-filled soil moisture extremes correspond well with a high-resolution Palmer Drought Severity Index calculated for the area.

Hierarchical Intrusion Detection Using Machine Learning and Knowledge Model

Intrusion detection systems (IDS) present a critical component of network infrastructures. Machine learning models are widely used in the IDS to learn the patterns in the network data and to detect the possible attacks in the network traffic. Ensemble models combining a variety of different machine learning models proved to be efficient in this domain. On the other hand, knowledge models have been explicitly designed for the description of the attacks and used in ontology-based IDS. In this paper, we propose a hierarchical IDS based on the original symmetrical combination of machine learning approach with knowledge-based approach to support detection of existing types and severity of new types of network attacks. Multi-stage hierarchical prediction consists of the predictive models able to distinguish the normal connections from the attacks and then to predict the attack classes and concrete attack types. The knowledge model enables to navigate through the attack taxonomy and to select the appropriate model to perform a prediction on the selected level. Designed IDS was evaluated on a widely used KDD 99 dataset and compared to similar approaches.