Nan Liu

Voting based extreme learning machine

This paper proposes an improved learning algorithm for classification which is referred to as voting based extreme learning machine. The proposed method incorporates the voting method into the popular extreme learning machine (ELM) in classification applications. Simulations on many real world classification datasets have demonstrated that this algorithm generally outperforms the original ELM algorithm as well as several recent classification algorithms.

Ensemble Based Extreme Learning Machine

Extreme learning machine (ELM) was proposed as a new class of learning algorithm for single-hidden layer feedforward neural network (SLFN). To achieve good generalization performance, ELM minimizes training error on the entire training data set, therefore it might suffer from overfitting as the learning model will approximate all training samples well. In this letter, an ensemble based ELM (EN-ELM) algorithm is proposed where ensemble learning and cross-validation are embedded into the training phase so as to alleviate the overtraining problem and enhance the predictive stability. Experimental results on several benchmark databases demonstrate that EN-ELM is robust and efficient for classification.

Ensemble of subset online sequential extreme learning machine for class imbalance and concept drift

In this paper, a computationally efficient framework, referred to as ensemble of subset online sequential extreme learning machine (ESOS-ELM), is proposed for class imbalance learning from a concept-drifting data stream. The proposed framework comprises a main ensemble representing short-term memory, an information storage module representing long-term memory and a change detection mechanism to promptly detect concept drifts. In the main ensemble of ESOS-ELM, each OS-ELM network is trained with a balanced subset of the data stream. Using ELM theory, a computationally efficient storage scheme is proposed to leverage the prior knowledge of recurring concepts. A distinctive feature of ESOS-ELM is that it can learn from new samples sequentially in both the chunk-by-chunk and one-by-one modes. ESOS-ELM can also be effectively applied to imbalanced data without concept drift. On most of the datasets used in our experiments, ESOS-ELM performs better than the state-of-the-art methods for both stationary and non-stationary environments.

Prediction of cardiac arrest in critically ill patients presenting to the emergency department using a machine learning score incorporating heart rate variability compared with the modified early warning score.

IntroductionA key aim of triage is to identify those with high risk of cardiac arrest, as they require intensive monitoring, resuscitation facilities, and early intervention. We aim to validate a novel machine learning (ML) score incorporating heart rate variability (HRV) for triage of critically ill patients presenting to the emergency department by comparing the area under the curve, sensitivity and specificity with the modified early warning score (MEWS).MethodsWe conducted a prospective observational study of critically ill patients (Patient Acuity Category Scale 1 and 2) in an emergency department of a tertiary hospital. At presentation, HRV parameters generated from a 5-minute electrocardiogram recording are incorporated with age and vital signs to generate the ML score for each patient. The patients are then followed up for outcomes of cardiac arrest or death.ResultsFrom June 2006 to June 2008 we enrolled 925 patients. The area under the receiver operating characteristic curve (AUROC) for ML scores in predicting cardiac arrest within 72 hours is 0.781, compared with 0.680 for MEWS (difference in AUROC: 0.101, 95% confidence interval: 0.006 to 0.197). As for in-hospital death, the area under the curve for ML score is 0.741, compared with 0.693 for MEWS (difference in AUROC: 0.048, 95% confidence interval: -0.023 to 0.119). A cutoff ML score ≥ 60 predicted cardiac arrest with a sensitivity of 84.1%, specificity of 72.3% and negative predictive value of 98.8%. A cutoff MEWS ≥ 3 predicted cardiac arrest with a sensitivity of 74.4%, specificity of 54.2% and negative predictive value of 97.8%.ConclusionWe found ML scores to be more accurate than the MEWS in predicting cardiac arrest within 72 hours. There is potential to develop bedside devices for risk stratification based on cardiac arrest prediction.

Patient Outcome Prediction with Heart Rate Variability and Vital Signs

The ability to predict patient outcomes is important for clinical triage, which is the process of assessing severity and assigning appropriate priority of treatment for large numbers of patients. In this study, we present an automatic prognosis system for patient outcome prediction with heart rate variability (HRV) and traditional vital signs. Support vector machine (SVM) and extreme learning machine (ELM) are employed as predictors, and SVM with linear kernel is reported to perform the best in general. In the experiments, the combination of HRV measures and vital signs is found to be more closely associated with patient outcome than either HRV or vital signs. Moreover, two new segment based methods are proposed to improve the predictive accuracy, where several sets of HRV measures are calculated from non-overlapped segments for each patient and final decision is made through the majority voting rule. The results reveal that the segment based methods are able to enhance the prediction performance significantly.

Evolutionary Extreme Learning Machine and Its Application to Image Analysis

Abstract Extreme learning machine (ELM) and evolutionary ELM (E-ELM) were proposed as a new class of learning algorithm for single-hidden layer feedforward neural network (SLFN). In order to achieve good generalization performance, E-ELM calculates the error on a subset of testing data for parameter optimization. Since E-ELMemploys extra data for validation to avoid the overfitting problem, more samples are needed for model training. In this paper, the cross-validation strategy is proposed to be embedded into the training phase so as to solve the overtraining problem. Based on this new learning structure, two extensions of E-ELM are introduced. Experimental results demonstrate that the proposed algorithms are efficient for image analysis.

Risk Scoring for Prediction of Acute Cardiac Complications from Imbalanced Clinical Data

Fast and accurate risk stratification is essential in the emergency department (ED) as it allows clinicians to identify chest pain patients who are at high risk of cardiac complications and require intensive monitoring and early intervention. In this paper, we present a novel intelligent scoring system using heart rate variability, 12-lead electrocardiogram (ECG), and vital signs where a hybrid sampling-based ensemble learning strategy is proposed to handle data imbalance. The experiments were conducted on a dataset consisting of 564 chest pain patients recruited at the ED of a tertiary hospital. The proposed ensemble-based scoring system was compared with established scoring methods such as the modified early warning score and the thrombolysis in myocardial infarction score, and showed its effectiveness in predicting acute cardiac complications within 72 h in terms of the receiver operation characteristic analysis.

Prediction of adverse cardiac events in emergency department patients with chest pain using machine learning for variable selection

BackgroundThe key aim of triage in chest pain patients is to identify those with high risk of adverse cardiac events as they require intensive monitoring and early intervention. In this study, we aim to discover the most relevant variables for risk prediction of major adverse cardiac events (MACE) using clinical signs and heart rate variability.MethodsA total of 702 chest pain patients at the Emergency Department (ED) of a tertiary hospital in Singapore were included in this study. The recruited patients were at least 30xa0years of age and who presented to the ED with a primary complaint of non-traumatic chest pain. The primary outcome was a composite of MACE such as death and cardiac arrest within 72xa0h of arrival at the ED. For each patient, eight clinical signs such as blood pressure and temperature were measured, and a 5-min ECG was recorded to derive heart rate variability parameters. A random forest-based novel method was developed to select the most relevant variables. A geometric distance-based machine learning scoring system was then implemented to derive a risk score from 0 to 100.ResultsOut of 702 patients, 29 (4.1%) met the primary outcome. We selected the 3 most relevant variables for predicting MACE, which were systolic blood pressure, the mean RR interval and the mean instantaneous heart rate. The scoring system with these 3 variables produced an area under the curve (AUC) of 0.812, and a cutoff score of 43 gave a sensitivity of 82.8% and specificity of 63.4%, while the scoring system with all the 23 variables had an AUC of 0.736, and a cutoff score of 49 gave a sensitivity of 72.4% and specificity of 63.0%. Conventional thrombolysis in myocardial infarction score and the modified early warning score achieved AUC values of 0.637 and 0.622, respectively.ConclusionsIt is observed that a few predictors outperformed the whole set of variables in predicting MACE within 72xa0h. We conclude that more predictors do not necessarily guarantee better prediction results. Furthermore, machine learning-based variable selection seems promising in discovering a few relevant and significant measures as predictors.

Feature Extraction with Genetic Algorithms Based Nonlinear Principal Component Analysis for Face Recognition

Principal component analysis (PCA) and linear discriminant analysis (LDA) are two commonly used feature extraction techniques. In this paper, a nonlinear evolutionary weighted principal component analysis (EWPCA) based on genetic algorithms is proposed. Similar to LDA, the EWPCA maximizes the ratio of between-class variations to that of within-class variations, and achieves better classification performance than that of traditional PCA. Genetic algorithms are chosen as the searching method to select optimal weights for the EWPCA. In face recognition, evolutionary facial feature obtained by performing EWPCA is used as the representation of original face images. Experimental results on ORL and combo face databases prove that EWPCA outperforms both PCA, kernel PCA and LDA

A prospective surveillance of paediatric head injuries in Singapore: a dual-centre study

Objective To study the causes of head injuries among the paediatric population in Singapore, and the association between causes and mortality, as well as the need for airway or neurosurgical intervention. Design This is a prospective observational study utilising data from the trauma surveillance system from January 2011 to March 2015. Setting Paediatric emergency departments (EDs) of KK Womens and Childrens Hospital and the National University Health System. Participants We included children aged <16u2005years presenting to the paediatric EDs with head injuries who required a CT scan, admission for monitoring of persistent symptoms, or who died from the head injury. We excluded children who presented with minor mechanisms and those whose symptoms had spontaneously resolved. Primary and secondary outcome measures Primary composite outcome was defined as death or the need for intubation or neurosurgical intervention. Secondary outcomes included length of hospital stay and type of neurosurgical intervention. Results We analysed 1049 children who met the inclusion criteria. The mean age was 6.7 (SD 5.2) years. 260 (24.8%) had a positive finding on CT. 17 (1.6%) children died, 52 (5.0%) required emergency intubation in the ED and 58 (5.5%) underwent neurosurgery. The main causes associated with severe outcomes were motor vehicle crashes (OR 7.2, 95% CI 4.3 to 12.0) and non-accidental trauma (OR 5.8, 95% CI 1.8 to 18.6). This remained statistically significant when we stratified to children aged <2u2005years and performed a multivariable analysis adjusting for age and location of injury. For motor vehicle crashes, less than half of the children were using restraints. Conclusions Motor vehicle crashes and non-accidental trauma causes are particularly associated with poor outcomes among children with paediatric head injury. Continued vigilance and compliance with injury prevention initiatives and legislature are vital.