Identifying problematic classes in text classification

Real-world text classification tasks often suffer from poor class structure with many overlapping classes and blurred boundaries. Training data pooled from multiple sources tend to be inconsistent and contain erroneous labelling, leading to poor performance of standard text classifiers. The classification of health service products to specialized procurement classes is used to examine and quantify the extent of these problems. A novel method is presented to analyze the labelled data by selectively merging classes where there is not enough information for the classifier to distinguish them. Initial results show the method can identify the most problematic classes, which can be used either as a focus to improve the training data or to merge classes to increase confidence in the predicted results of the classifier.

Parallel Monte Carlo algorithms for information retrieval

In any data mining applications, automated text and text and image retrieval of information is needed. This becomes essential with the growth of the Internet and digital libraries. Our approach is based on the latent semantic indexing (LSI) and the corresponding term-by-document matrix suggested by Berry and his co-authors. Instead of using deterministic methods to find the required number of first "k" singular triplets, we propose a stochastic approach. First, we use Monte Carlo method to sample and to build much smaller size term-by-document matrix (e.g. we build k x k matrix) from where we then find the first "k" triplets using standard deterministic methods. Second, we investigate how we can reduce the problem to finding the "k"-largest eigenvalues using parallel Monte Carlo methods. We apply these methods to the initial matrix and also to the reduced one. The algorithms are running on a cluster of workstations under MPI and results of the experiments arising in textual retrieval of Web documents as well as comparison of the stochastic methods proposed are presented. (C) 2003 IMACS. Published by Elsevier Science B.V. All rights reserved.

How Twitter is studied in the medical professions: a classification of Twitter papers indexed in PubMed

Background: Since their inception, Twitter and related microblogging systems have provided a rich source of information for researchers and have attracted interest in their affordances and use. Since 2009 PubMed has included 123 journal articles on medicine and Twitter, but no overview exists as to how the field uses Twitter in research. // Objective: This paper aims to identify published work relating to Twitter indexed by PubMed, and then to classify it. This classification will provide a framework in which future researchers will be able to position their work, and to provide an understanding of the current reach of research using Twitter in medical disciplines. Limiting the study to papers indexed by PubMed ensures the work provides a reproducible benchmark. // Methods: Papers, indexed by PubMed, on Twitter and related topics were identified and reviewed. The papers were then qualitatively classified based on the paper’s title and abstract to determine their focus. The work that was Twitter focused was studied in detail to determine what data, if any, it was based on, and from this a categorization of the data set size used in the studies was developed. Using open coded content analysis additional important categories were also identified, relating to the primary methodology, domain and aspect. // Results: As of 2012, PubMed comprises more than 21 million citations from biomedical literature, and from these a corpus of 134 potentially Twitter related papers were identified, eleven of which were subsequently found not to be relevant. There were no papers prior to 2009 relating to microblogging, a term first used in 2006. Of the remaining 123 papers which mentioned Twitter, thirty were focussed on Twitter (the others referring to it tangentially). The early Twitter focussed papers introduced the topic and highlighted the potential, not carrying out any form of data analysis. The majority of published papers used analytic techniques to sort through thousands, if not millions, of individual tweets, often depending on automated tools to do so. Our analysis demonstrates that researchers are starting to use knowledge discovery methods and data mining techniques to understand vast quantities of tweets: the study of Twitter is becoming quantitative research. // Conclusions: This work is to the best of our knowledge the first overview study of medical related research based on Twitter and related microblogging. We have used five dimensions to categorise published medical related research on Twitter. This classification provides a framework within which researchers studying development and use of Twitter within medical related research, and those undertaking comparative studies of research relating to Twitter in the area of medicine and beyond, can position and ground their work.

Soft topographic map for clustering and classification of bacteria

In this work a new method for clustering and building a topographic representation of a bacteria taxonomy is presented. The method is based on the analysis of stable parts of the genome, the so-called “housekeeping genes”. The proposed method generates topographic maps of the bacteria taxonomy, where relations among different type strains can be visually inspected and verified. Two well known DNA alignement algorithms are applied to the genomic sequences. Topographic maps are optimized to represent the similarity among the sequences according to their evolutionary distances. The experimental analysis is carried out on 147 type strains of the Gammaprotebacteria class by means of the 16S rRNA housekeeping gene. Complete sequences of the gene have been retrieved from the NCBI public database. In the experimental tests the maps show clusters of homologous type strains and present some singular cases potentially due to incorrect classification or erroneous annotations in the database.

Automated pre-processing strategies for species occurrence data used in biodiversity modelling

To construct Biodiversity richness maps from Environmental Niche Models (ENMs) of thousands of species is time consuming. A separate species occurrence data pre-processing phase enables the experimenter to control test AUC score variance due to species dataset size. Besides, removing duplicate occurrences and points with missing environmental data, we discuss the need for coordinate precision, wide dispersion, temporal and synonymity filters. After species data filtering, the final task of a pre-processing phase should be the automatic generation of species occurrence datasets which can then be directly ’plugged-in’ to the ENM. A software application capable of carrying out all these tasks will be a valuable time-saver particularly for large scale biodiversity studies.

Biomarker discovery and redundancy reduction towards classification using a multi-factorial MALDI-TOF MS T2DM mouse model dataset

Diabetes like many diseases and biological processes is not mono-causal. On the one hand multifactorial studies with complex experimental design are required for its comprehensive analysis. On the other hand, the data from these studies often include a substantial amount of redundancy such as proteins that are typically represented by a multitude of peptides. Coping simultaneously with both complexities (experimental and technological) makes data analysis a challenge for Bioinformatics.

An automated calibration method for non-see-through head mounted displays

Accurate calibration of a head mounted display (HMD) is essential both for research on the visual system and for realistic interaction with virtual objects. Yet, existing calibration methods are time consuming and depend on human judgements, making them error prone, and are often limited to optical see-through HMDs. Building on our existing approach to HMD calibration Gilson et al. (2008), we show here how it is possible to calibrate a non-see-through HMD. A camera is placed inside a HMD displaying an image of a regular grid, which is captured by the camera. The HMD is then removed and the camera, which remains fixed in position, is used to capture images of a tracked calibration object in multiple positions. The centroids of the markers on the calibration object are recovered and their locations re-expressed in relation to the HMD grid. This allows established camera calibration techniques to be used to recover estimates of the HMD display's intrinsic parameters (width, height, focal length) and extrinsic parameters (optic centre and orientation of the principal ray). We calibrated a HMD in this manner and report the magnitude of the errors between real image features and reprojected features. Our calibration method produces low reprojection errors without the need for error-prone human judgements.

Automated categorisation of e-journals by synonym analysis of n-grams

Automatic keyword or keyphrase extraction is concerned with assigning keyphrases to documents based on words from within the document. Previous studies have shown that in a significant number of cases author-supplied keywords are not appropriate for the document to which they are attached. This can either be because they represent what the author believes a paper is about not what it actually is, or because they include keyphrases which are more classificatory than explanatory e.g., “University of Poppleton” instead of “Knowledge Discovery in Databases”. Thus, there is a need for a system that can generate an appropriate and diverse range of keyphrases that reflect the document. This paper proposes two possible solutions that examine the synonyms of words and phrases in the document to find the underlying themes, and presents these as appropriate keyphrases. Using three different freely available thesauri, the work undertaken examines two different methods of producing keywords and compares the outcomes across multiple strands in the timeline. The primary method explores taking n-grams of the source document phrases, and examining the synonyms of these, while the secondary considers grouping outputs by their synonyms. The experiments undertaken show the primary method produces good results and that the secondary method produces both good results and potential for future work. In addition, the different qualities of the thesauri are examined and it is concluded that the more entries in a thesaurus, the better it is likely to perform. The age of the thesaurus or the size of each entry does not correlate to performance.

Jmax-pruning: a facility for the information theoretic pruning of modular classification rules

The Prism family of algorithms induces modular classification rules in contrast to the Top Down Induction of Decision Trees (TDIDT) approach which induces classification rules in the intermediate form of a tree structure. Both approaches achieve a comparable classification accuracy. However in some cases Prism outperforms TDIDT. For both approaches pre-pruning facilities have been developed in order to prevent the induced classifiers from overfitting on noisy datasets, by cutting rule terms or whole rules or by truncating decision trees according to certain metrics. There have been many pre-pruning mechanisms developed for the TDIDT approach, but for the Prism family the only existing pre-pruning facility is J-pruning. J-pruning not only works on Prism algorithms but also on TDIDT. Although it has been shown that J-pruning produces good results, this work points out that J-pruning does not use its full potential. The original J-pruning facility is examined and the use of a new pre-pruning facility, called Jmax-pruning, is proposed and evaluated empirically. A possible pre-pruning facility for TDIDT based on Jmax-pruning is also discussed.

Scaling up classification rule induction through parallel processing

The fast increase in the size and number of databases demands data mining approaches that are scalable to large amounts of data. This has led to the exploration of parallel computing technologies in order to perform data mining tasks concurrently using several processors. Parallelization seems to be a natural and cost-effective way to scale up data mining technologies. One of the most important of these data mining technologies is the classification of newly recorded data. This paper surveys advances in parallelization in the field of classification rule induction.

Computationally efficient induction of classification rules with the PMCRI and J-PMCRI frameworks

In order to gain knowledge from large databases, scalable data mining technologies are needed. Data are captured on a large scale and thus databases are increasing at a fast pace. This leads to the utilisation of parallel computing technologies in order to cope with large amounts of data. In the area of classification rule induction, parallelisation of classification rules has focused on the divide and conquer approach, also known as the Top Down Induction of Decision Trees (TDIDT). An alternative approach to classification rule induction is separate and conquer which has only recently been in the focus of parallelisation. This work introduces and evaluates empirically a framework for the parallel induction of classification rules, generated by members of the Prism family of algorithms. All members of the Prism family of algorithms follow the separate and conquer approach.

Assessment of inter-examiner agreement and variability in the manual classification of auditory brainstem response

Abstract Background: The analysis of the Auditory Brainstem Response (ABR) is of fundamental importance to the investigation of the auditory system behaviour, though its interpretation has a subjective nature because of the manual process employed in its study and the clinical experience required for its analysis. When analysing the ABR, clinicians are often interested in the identification of ABR signal components referred to as Jewett waves. In particular, the detection and study of the time when these waves occur (i.e., the wave latency) is a practical tool for the diagnosis of disorders affecting the auditory system. Significant differences in inter-examiner results may lead to completely distinct clinical interpretations of the state of the auditory system. In this context, the aim of this research was to evaluate the inter-examiner agreement and variability in the manual classification of ABR. Methods: A total of 160 ABR data samples were collected, for four different stimulus intensity (80dBHL, 60dBHL, 40dBHL and 20dBHL), from 10 normal-hearing subjects (5 men and 5 women, from 20 to 52 years). Four examiners with expertise in the manual classification of ABR components participated in the study. The Bland-Altman statistical method was employed for the assessment of inter-examiner agreement and variability. The mean, standard deviation and error for the bias, which is the difference between examiners’ annotations, were estimated for each pair of examiners. Scatter plots and histograms were employed for data visualization and analysis. Results: In most comparisons the differences between examiner’s annotations were below 0.1 ms, which is clinically acceptable. In four cases, it was found a large error and standard deviation (>0.1 ms) that indicate the presence of outliers and thus, discrepancies between examiners. Conclusions: Our results quantify the inter-examiner agreement and variability of the manual analysis of ABR data, and they also allows for the determination of different patterns of manual ABR analysis.

Whole-brain functional connectivity during emotional word classification in medication-free Major Depressive Disorder: abnormal salience circuitry and relations to positive emotionality

Major Depressive Disorder (MDD) has been associated with biased processing and abnormal regulation of negative and positive information, which may result from compromised coordinated activity of prefrontal and subcortical brain regions involved in evaluating emotional information. We tested whether patients with MDD show distributed changes in functional connectivity with a set of independently derived brain networks that have shown high correspondence with different task demands, including stimulus salience and emotional processing. We further explored if connectivity during emotional word processing related to the tendency to engage in positive or negative emotional states. In this study, 25 medication-free MDD patients without current or past comorbidity and matched controls (n=25) performed an emotional word-evaluation task during functional MRI. Using a dual regression approach, individual spatial connectivity maps representing each subject’s connectivity with each standard network were used to evaluate between-group differences and effects of positive and negative emotionality (extraversion and neuroticism, respectively, as measured with the NEO-FFI). Results showed decreased functional connectivity of the medial prefrontal cortex, ventrolateral prefrontal cortex, and ventral striatum with the fronto-opercular salience network in MDD patients compared to controls. In patients, abnormal connectivity was related to extraversion, but not neuroticism. These results confirm the hypothesis of a relative (para)limbic-cortical decoupling that may explain dysregulated affect in MDD. As connectivity of these regions with the salience network was related to extraversion, but not to general depression severity or negative emotionality, dysfunction of this network may be responsible for the failure to sustain engagement in rewarding behavior.

Analysis of full-waveform LiDAR data for classification of an orange orchard scene

Full-waveform laser scanning data acquired with a Riegl LMS-Q560 instrument were used to classify an orange orchard into orange trees, grass and ground using waveform parameters alone. Gaussian decomposition was performed on this data capture from the National Airborne Field Experiment in November 2006 using a custom peak-detection procedure and a trust-region-reflective algorithm for fitting Gauss functions. Calibration was carried out using waveforms returned from a road surface, and the backscattering coefficient c was derived for every waveform peak. The processed data were then analysed according to the number of returns detected within each waveform and classified into three classes based on pulse width and c. For single-peak waveforms the scatterplot of c versus pulse width was used to distinguish between ground, grass and orange trees. In the case of multiple returns, the relationship between first (or first plus middle) and last return c values was used to separate ground from other targets. Refinement of this classification, and further sub-classification into grass and orange trees was performed using the c versus pulse width scatterplots of last returns. In all cases the separation was carried out using a decision tree with empirical relationships between the waveform parameters. Ground points were successfully separated from orange tree points. The most difficult class to separate and verify was grass, but those points in general corresponded well with the grass areas identified in the aerial photography. The overall accuracy reached 91%, using photography and relative elevation as ground truth. The overall accuracy for two classes, orange tree and combined class of grass and ground, yielded 95%. Finally, the backscattering coefficient c of single-peak waveforms was also used to derive reflectance values of the three classes. The reflectance of the orange tree class (0.31) and ground class (0.60) are consistent with published values at the wavelength of the Riegl scanner (1550 nm). The grass class reflectance (0.46) falls in between the other two classes as might be expected, as this class has a mixture of the contributions of both vegetation and ground reflectance properties.

J-measure based hybrid pruning for complexity reduction in classification rules

Prism is a modular classification rule generation method based on the ‘separate and conquer’ approach that is alternative to the rule induction approach using decision trees also known as ‘divide and conquer’. Prism often achieves a similar level of classification accuracy compared with decision trees, but tends to produce a more compact noise tolerant set of classification rules. As with other classification rule generation methods, a principle problem arising with Prism is that of overfitting due to over-specialised rules. In addition, over-specialised rules increase the associated computational complexity. These problems can be solved by pruning methods. For the Prism method, two pruning algorithms have been introduced recently for reducing overfitting of classification rules - J-pruning and Jmax-pruning. Both algorithms are based on the J-measure, an information theoretic means for quantifying the theoretical information content of a rule. Jmax-pruning attempts to exploit the J-measure to its full potential because J-pruning does not actually achieve this and may even lead to underfitting. A series of experiments have proved that Jmax-pruning may outperform J-pruning in reducing overfitting. However, Jmax-pruning is computationally relatively expensive and may also lead to underfitting. This paper reviews the Prism method and the two existing pruning algorithms above. It also proposes a novel pruning algorithm called Jmid-pruning. The latter is based on the J-measure and it reduces overfitting to a similar level as the other two algorithms but is better in avoiding underfitting and unnecessary computational effort. The authors conduct an experimental study on the performance of the Jmid-pruning algorithm in terms of classification accuracy and computational efficiency. The algorithm is also evaluated comparatively with the J-pruning and Jmax-pruning algorithms.