International Journal of Semantic Computing

I will make an argument of who will benefit from this special issue on data science and related topics.

The logical basis for information theory is the newly developed logic of partitions that is dual to the usual Boolean logic of subsets. The key concept is a "distinction" of a partition, an ordered pair of elements in distinct blocks of the partition. The logical concept of entropy based on partition logic is the normalized counting measure of the set of distinctions of a partition on a finite set — just as the usual logical notion of probability based on the Boolean logic of subsets is the normalized counting measure of the subsets (events). Thus logical entropy is a measure on the set of ordered pairs, and all the compound notions of entropy (join entropy, conditional entropy, and mutual information) arise in the usual way from the measure (e.g. the inclusion-exclusion principle) — just like the corresponding notions of probability. The usual Shannon entropy of a partition is developed by replacing the normalized count of distinctions (dits) by the average number of binary partitions (bits) necessary to make all the distinctions of the partition.

Artificial neural networks, due to their ability to find the underlying model even in complex highly nonlinear and highly coupled problems, have found significant use as prediction engines in many domains. However, in problems where the input space is of high dimensionality, there is the unsolved problem of reducing dimensionality in some optimal way such that Shannon information important to the prediction is preserved. The important Shannon information may be a subset of total information with an unknown partition, unknown coupling and linear or nonlinear in nature. Solving this problem is an important step in classes of machine learning problems and many data mining applications. This paper describes a semi-automatic algorithm that was developed over a 5-year period while solving problems with increasing dimensionality and difficulty in (a) flow prediction for a magnetically levitated artificial heart (13 dimensions), (b) simultaneous chemical identification/concentration in gas chromatography (22 detection dimensions with wavelet compressed time series of 180,000 points), and finally in (c) financial analytics portfolio prediction in credit card and sub-prime debt problems (80 to 300 dimensions of sparse data with a portfolio value of approximately US$300,000,000.00). The algorithm develops a map of input space combinations and their importance to the prediction. This information is used directly to construct the optimal neural network topology for a given error performance. Importantly, the algorithm also produces information that shows whether the space between input nodes is linear or nonlinear; an important parameter in determining the number of training points required in the reduced dimensionality of the training set. Software was developed in the MatLAB environment using the Artificial Neural Network Toolbox, Parallel and Distributed Computing toolboxes, and runs on Windows or Linux based supercomputers. Trained neural networks can be compiled and linked to server applications and run on normal servers or clusters for transaction or web based processing. In this paper, application of the algorithm to two separate financial analytics prediction problems with large dimensionality and sparse data sets are shown. The algorithm is an important development in machine learning for an important class of problems in prediction, clustering, image analysis, and data mining. In the first example application for subprime debt portfolio analysis, performance of the neural network provided a 98.4% prediction rate, compared to 33% rate using traditional linear methods. In the second example application regarding credit card debt, performance of the algorithm provided a 95% accurate prediction (in terms of match rate), and is 10% better than other methods we have compared against, primarily logistic regression.

Glaucoma is a chronic, progressive and potentially blinding condition. Predicting which patients will experience significant progression is recognized as a crucially needed development in the management of this disease. Application of the CART (Classification And Regression Trees) methodology has demonstrated that certain patterns of visual field findings may convey greater predictive information for glaucoma progression. However, the current standard classification tree method was developed for uncorrelated data. In this article a classification tree method is extended to correlated binary data. The robust Wald test statistic from generalized estimating equations (GEE) is used to measure the between-node difference while adjusting for correlation between the eyes of a patient. The proposed method is assessed through simulations conducted under a variety of model configurations and is used to analyze the perimetry and psychophysics in glaucoma (PPIG) study data. Employing an amalgamation algorithm from the result of a best-sized tree, each eye is classified to one of two prognosis categories (less likely, or more likely, to progress). Receiver operating characteristics (ROC) and area under the curve (AUC) indicate that the proposed method, applied to data from both eyes of the same patient, provides much improved prediction accuracy compared with application of standard CART method to the same PPIG data.

In this paper we demonstrate a new approach to the classification of multivariate time series streaming data by utilizing a temporal metafeature abstractions method. The technique extracts global features and metafeatures in order to capture the necessary time-lapse information in the streams of data. The features are then used to create a static, intermediate stream representation that includes all the important time-varying information, and is suitable for analysis using the standard supervised data mining techniques. The capability of the new algorithm called MineTool-TS² was demonstrated through its application to three datasets: UCSD Microgrid energy usage data, a space physics dataset and synthetic data.

The purpose of this tutorial is to introduce the powerful method of support vector machines for classifications and to advocate the usage of the statistical language R to develop SVM-based models.

We demonstrate the use of support vector machine methods to classify autism neuroimaging data collected from multiple sites.