Susana M. Vieira

Modified binary PSO for feature selection using SVM applied to mortality prediction of septic patients

Abstract This paper proposes a modified binary particle swarm optimization (MBPSO) method for feature selection with the simultaneous optimization of SVM kernel parameter setting, applied to mortality prediction in septic patients. An enhanced version of binary particle swarm optimization, designed to cope with premature convergence of the BPSO algorithm is proposed. MBPSO control the swarm variability using the velocity and the similarity between best swarm solutions. This paper uses support vector machines in a wrapper approach, where the kernel parameters are optimized at the same time. The approach is applied to predict the outcome (survived or deceased) of patients with septic shock. Further, MBPSO is tested in several benchmark datasets and is compared with other PSO based algorithms and genetic algorithms (GA). The experimental results showed that the proposed approach can correctly select the discriminating input features and also achieve high classification accuracy, specially when compared to other PSO based algorithms. When compared to GA, MBPSO is similar in terms of accuracy, but the subset solutions have less selected features.

Decision tree search methods in fuzzy modeling and classification

This paper proposes input selection methods for fuzzy modeling, which are based on decision tree search approaches. The branching decision at each node of the tree is made based on the accuracy of the model available at the node. We propose two different approaches of decision tree search algorithms: bottom-up and top-down and four different measures for selecting the most appropriate set of inputs at every branching node (or decision node). Both decision tree approaches are tested using real-world application examples. These methods are applied to fuzzy modeling of two different classification problems and to fuzzy modeling of two dynamic processes. The models accuracy of the four different examples are compared in terms of several performance measures. Moreover, the advantages and drawbacks of using bottom-up or top-down approaches are discussed.

Two cooperative ant colonies for feature selection using fuzzy models

The available set of potential features in real-world databases is sometimes very large, and it can be necessary to find a small subset for classification purposes. One of the most important techniques in data pre-processing for classification is feature selection. Less relevant or highly correlated features decrease, in general, the classification accuracy and enlarge the complexity of the classifier. The goal is to find a reduced set of features that reveals the best classification accuracy for a classifier. Rule-based fuzzy models can be acquired from numerical data, and be used as classifiers. As rule based structures revealed to be a useful qualitative description for classification systems, this work uses fuzzy models as classifiers. This paper proposes an algorithm for feature selection based on two cooperative ant colonies, which minimizes two objectives: the number of features and the classification error. Two pheromone matrices and two different heuristics are used for these objectives. The performance of the method is compared with other features selection methods, achieving equal or better performance.

Fuzzy criteria for feature selection

The presence of less relevant or highly correlated features often decrease classification accuracy. Feature selection in which most informative variables are selected for model generation is an important step in data-driven modeling. In feature selection, one often tries to satisfy multiple criteria such as feature discriminating power, model performance or subset cardinality. Therefore, a multi-objective formulation of the feature selection problem is more appropriate. In this paper, we propose to use fuzzy criteria in feature selection by using a fuzzy decision making framework. This formulation allows for a more flexible definition of the goals in feature selection, and avoids the problem of weighting different goals is classical multi-objective optimization. The optimization problem is solved using an ant colony optimization algorithm proposed in our previous work. We illustrate the added value of the approach by applying our proposed fuzzy feature selection algorithm to eight benchmark problems.

Data mining using clinical physiology at discharge to predict ICU readmissions

Patient readmissions to intensive care units (ICUs) are associated with increased mortality, morbidity and costs. Current models for predicting ICU readmissions have moderate predictive value, and can utilize up to twelve variables that may be assessed at various points of the ICU inpatient stay. We postulate that greater predictive value can be achieved with fewer physiological variables, some of which can be assessed in the 24h before discharge. A data mining approach combining fuzzy modeling with tree search feature selection was applied to a large retrospectively collected ICU database (MIMIC II), representing data from four different ICUs at Beth Israel Deaconess Medical Center, Boston. The goal was to predict ICU readmission between 24 and 72h after ICU discharge. Fuzzy modeling combined with sequential forward selection was able to predict readmissions with an area under the receiver-operating curve (AUC) of 0.72+/-0.04, a sensitivity of 0.68+/-0.02 and a specificity of 0.73+/-0.03. Variables selected as having the highest predictive power include mean heart rate, mean temperature, mean platelets, mean non-invasive arterial blood pressure (mean), mean spO2, and mean lactic acid, during the last 24h before discharge. Collection of the six predictive variables selected is not complex in modern ICUs, and their assessment may help support the development of clinical management plans that potentially mitigate the risk of readmission.

Missing data in medical databases: Impute, delete or classify?

BACKGROUND The multiplicity of information sources for data acquisition in modern intensive care units (ICUs) makes the resulting databases particularly susceptible to missing data. Missing data can significantly affect the performance of predictive risk modeling, an important technique for developing medical guidelines. The two most commonly used strategies for managing missing data are to impute or delete values, and the former can cause bias, while the later can cause both bias and loss of statistical power. OBJECTIVES In this paper we present a new approach for managing missing data in ICU databases in order to improve overall modeling performance. METHODS We use a statistical classifier followed by fuzzy modeling to more accurately determine which missing data should be imputed and which should not. We firstly develop a simulation test bed to evaluate performance, and then translate that knowledge using exactly the same database as previously published work by [13]. RESULTS In this work, test beds resulted in datasets with missing data ranging 10-50%. Using this new approach to missing data we are able to significantly improve modeling performance parameters such as accuracy of classifications by an 11%, sensitivity by 13%, and specificity by 10%, including also area under the receiver-operator curve (AUC) improvement of up to 13%. CONCLUSIONS In this work, we improve modeling performance in a simulated test bed, and then confirm improved performance replicating previously published work by using the proposed approach for missing data classification. We offer this new method to other researchers who wish to improve predictive risk modeling performance in the ICU through advanced missing data management.

Predicting Outcomes of Septic Shock Patients Using Feature Selection Based on Soft Computing Techniques

This paper proposes the application of new knowledge based methods to a septic shock patient database. It uses wrapper methods (bottom-up tree search or ant feature selection) to reduce the number of features. Fuzzy and neural modeling are used for classification. The goal is to estimate, as accurately as possible, the outcome (survived or deceased) of these septic shock patients. Results show that the approaches presented outperform any previous solutions, specifically in terms of sensitivity.

Reducing unnecessary lab testing in the ICU with artificial intelligence

OBJECTIVES To reduce unnecessary lab testing by predicting when a proposed future lab test is likely to contribute information gain and thereby influence clinical management in patients with gastrointestinal bleeding. Recent studies have demonstrated that frequent laboratory testing does not necessarily relate to better outcomes. DESIGN Data preprocessing, feature selection, and classification were performed and an artificial intelligence tool, fuzzy modeling, was used to identify lab tests that do not contribute an information gain. There were 11 input variables in total. Ten of these were derived from bedside monitor trends heart rate, oxygen saturation, respiratory rate, temperature, blood pressure, and urine collections, as well as infusion products and transfusions. The final input variable was a previous value from one of the eight lab tests being predicted: calcium, PTT, hematocrit, fibrinogen, lactate, platelets, INR and hemoglobin. The outcome for each test was a binary framework defining whether a test result contributed information gain or not. PATIENTS Predictive modeling was applied to recognize unnecessary lab tests in a real world ICU database extract comprising 746 patients with gastrointestinal bleeding. MAIN RESULTS Classification accuracy of necessary and unnecessary lab tests of greater than 80% was achieved for all eight lab tests. Sensitivity and specificity were satisfactory for all the outcomes. An average reduction of 50% of the lab tests was obtained. This is an improvement from previously reported similar studies with average performance 37% by [1-3]. CONCLUSIONS Reducing frequent lab testing and the potential clinical and financial implications are an important issue in intensive care. In this work we present an artificial intelligence method to predict the benefit of proposed future laboratory tests. Using ICU data from 746 patients with gastrointestinal bleeding, and eleven measurements, we demonstrate high accuracy in predicting the likely information to be gained from proposed future lab testing for eight common GI related lab tests. Future work will explore applications of this approach to a range of underlying medical conditions and laboratory tests.

Cohen's kappa coefficient as a performance measure for feature selection

Measuring the performance of a given classifier is not a straightforward or easy task. Depending on the application, the overall classification rate may not be sufficient if one, or more, of the classes fail in prediction. This problem is also reflected in the feature selection process, especially when a wrapper method is used. Cohens kappa coefficient is a statistical measure of inter-rater agreement for qualitative items. It is generally thought to be a more robust measure than simple percent agreement calculation, since it takes into account the agreement occurring by chance. Considering that kappa is a more conservative measure, then its use in wrapper feature selection is suitable to test the performance of the models. This paper proposes the use of the kappa measure as an evaluation measure in a feature selection wrapper approach. In the proposed approach, fuzzy models are used to test the feature subsets and fuzzy criteria are used to formulate the feature selection problem. Results show that using the kappa measure leads to more accurate classifiers, and therefore it leads to feature subset solutions with more relevant features.

Ant Colony Optimization Applied to Feature Selection in Fuzzy Classifiers

In practice, classifiers are often build based on data or heuristic information. The number of potential features is usually large. One of the most important tasks in classification systems is to identify the most relevant features, because less relevant features can be interpreted as noise that reduces the classification accuracy, even for fuzzy classifiers which are somehow robust to noise. This paper proposes an ant colony optimization (ACO) algorithm for the feature selection problem. The goal is to find the set of features that reveals the best classification accuracy for a fuzzy classifier. The performance of the method is compared to other features selection methods based on tree search methods.