Chid Apte

Empirical evaluation of feature subset selection based on a real-world data set

Abstract Selecting the right set of features for classification is one of the most important problems in designing a good classifier. Decision tree induction algorithms such as C4.5 have incorporated in their learning phase an automatic feature selection strategy, while some other statistical classification algorithm require the feature subset to be selected in a preprocessing phase. It is well known that correlated and irrelevant features may degrade the performance of the C4.5 algorithm. In our study, we evaluated the influence of feature pre-selection on the prediction accuracy of C4.5 using a real-world data set. We observed that accuracy of the C4.5 classifier can be improved with an appropriate feature pre-selection phase for the learning algorithm.

Business leadership through analytics

Weak economic growth, increasing business complexity, and intensifying competition are among the motivations for the growing use of analytics across organizations of all types. In this business environment, enterprise leadership requires new tools and new insights to make better decisions. While analytic techniques, no matter how primitive, have been used to improve the effectiveness and efficiency of organizations for decades, if not centuries, capability and sophistication are now growing rapidly. Advanced analytics is proving to be a transformational technology for businesses, providing a source of competitive advantage and delivering significant, measurable outcomes. Risk management, sales-force productivity, workforce optimization, enterprise planning, and market intelligence are a few examples of core business processes that have significantly benefited from emerging analytical methodologies. As analytic tools used to transform business processes are integrated into enterprise IT (information technology) systems for use by senior business leaders, they are driving new innovations to address the size, speed, and complexity issues faced by large, global enterprises. This position paper explores emerging trends in the leadership application of analytics in global business organizations.

Data Science for Food, Energy and Water: A Workshop Report

At the 22nd ACM SIGKDD conference on Knowledge and Data Discovery (KDD), a workshop on Data Science for Food, Energy andWater (DSFEW) was held to foster an interdisciplinary community intersecting data science and societally important domains of food, energy and water. The workshop included keynotes, panel discussion, presentations and posters, and introduced the emerging area of DSFEW to ACM SIGKDD audience, and triggered interdisciplinary idea-sharing in DSFEW research. The workshop website is sites.google.com/site/2016dsfew.

Introduction to Special Issue on Large-Scale Data Mining

Large-scale data mining focuses on the main premise of all practical data mining methods, that is, the scalability and efficiency of the method. With the exponential growth of data volume, we argue that scalability often becomes the top requirement to make any data mining method relevant in practice. The goal of this special issue is to promote the theory and applications of large-scale data mining. In particular, we attracted 19 articles, which were reviewed by three anonymous referees through two rounds of reviews. Finally, we selected six articles for inclusion in this special issue. The selected articles cover various areas such as systems, algorithms, and a comparison survey. From the systems aspect, we selected two articles. “HADI: Mining Radii of Large Graphs.” by U. Kang et al. proposes an extremely scalable method for computing the diameter of the large graph using MapReduce. The authors evaluated their system on real-world graphs including one with 6-billion edges, 1/8 of a terabyte of storage. “Robust Record Linkage Blocking Using Suffix Arrays and Bloom Filters” by Timothy De Vries et al. studies the large-scale record linkage problem, which is of great practical significance in data mining and data management applications. In particular, the authors combine suffix array and bloom filters in order to achieve great scalability and accuracy. From the algorithmic aspect, we selected four articles. “Temporal Link Prediction using Matrix and Tensor Factorizations” by Daniel Dunlavy et al. studies the temporal link prediction on graphs using scalable and sparse matrix and tensor factorization. “Enhancing Clustering Quality through Landmark-based Dimensionality Reduction” by Panagis Magdalinos et al. proposes a landmark-based dimensionality reduction, which can improve the scalability of the subsequent clustering methods and the clustering quality. “Clustering Large Attributed Graphs: A Balance Between Structural and Attribute Similarities” by Hong Cheng et al. studies a special type of graph where each node in the graph is associated with certain attributes. The authors propose an efficient clustering algorithm using these graphs. Finally, “Fast Algorithms for Approximating the Singular Value Decomposition” by Aditya Krishna Menon and Charles Elkan surveys a number of fast low-rank approximation algorithms which have been used in many data mining methods and applications. The authors quantitatively compare the performance of those algorithms in various aspects. The result provides a great practical guideline for selecting the right algorithm for doing fast low-rank approximation.

Pattern Recognition and Image Analysis

International Workshop on Pattern Recognition for Crime Prevention, Security and Surveillance.- Image Enhancement Optimization for Hand-Luggage Screening at Airports.- Parameter Optimization for Image Segmentation Algorithms: A Systematic Approach.- Fingerprint Image Enhancement Using STFT Analysis.- Symmetric Hash Functions for Fingerprint Minutiae.- A Digital Rights Management Approach for Gray-Level Images.- Millimetre-Wave Personnel Scanners for Automated Weapon Detection.- A Thermal Hand Vein Pattern Verification System.- Illumination Tolerant Face Recognition Using Phase-Only Support Vector Machines in the Frequency Domain.- Regional and Online Learnable Fields.- Spatial Feature Based Recognition of Human Dynamics in Video Sequences.- Using Behavior Knowledge Space and Temporal Information for Detecting Intrusions in Computer Networks.- Biometrics.- View Independent Video-Based Face Recognition Using Posterior Probability in Kernel Fisher Discriminant Space.- Attention Based Facial Symmetry Detection.- An Efficient Iris Segmentation Method for Recognition.- Multi-scale Palmprint Recognition Using Registration Information and 2D Gabor Feature.- Effects of JPEG and JPEG2000 Compression on Face Recognition.- 3D Action Modeling and Reconstruction for 2D Human Body Tracking.- A Non-parametric Dimensionality Reduction Technique Using Gradient Descent of Misclassification Rate.- On the Automatic 2D Retinal Vessel Extraction.- Modeling Phase Spectra Using Gaussian Mixture Models for Human Face Identification.- Belief Theory Applied to Facial Expressions Classification.- Face Recognition Using Uncorrelated, Weighted Linear Discriminant Analysis.- Face Recognition Using Heteroscedastic Weighted Kernel Discriminant Analysis.- Class-Specific Discriminant Non-negative Matrix Factorization for Frontal Face Verification.- Partial Relevance in Interactive Facial Image Retrieval.- An Integration of Biometrics and Mobile Computing for Personal Identification.- Eyes Segmentation Applied to Gaze Direction and Vigilance Estimation.- Bilinear Discriminant Analysis for Face Recognition.- Adaptive Object Recognition Using Context-Aware Genetic Algorithm Under Dynamic Environment.- A Multi-scale and Multi-pose Face Detection System.- Conditionally Dependent Classifier Fusion Using AND Rule for Improved Biometric Verification.- Measurement of Face Recognizability for Visual Surveillance.- A Fingerprint Authentication Mobile Phone Based on Sweep Sensor.- A Robust and Efficient Algorithm for Eye Detection on Gray Intensity Face.- Silhouette Spatio-temporal Spectrum (SStS) for Gait-Based Human Recognition.- Adaptive Estimation of Human Posture Using a Component-Based Model.- Fusion of Locally Linear Embedding and Principal Component Analysis for Face Recognition (FLLEPCA).- Proposal of Novel Histogram Features for Face Detection.- Feature Selection Based on KPCA, SVM and GSFS for Face Recognition.- Eigen and Fisher-Fourier Spectra for Shift Invariant Pose-Tolerant Face Recognition.- Image Processing.- Q-Gram Statistics Descriptor in 3D Shape Classification.- A New Inpainting Method for Highlights Elimination by Colour Morphology.- Clustering of Objects in 3D Electron Tomography Reconstructions of Protein Solutions Based on Shape Measurements.- Improving Tracking by Handling Occlusions.- Image Reconstruction with Polar Zernike Moments.- Texture Exemplars for Defect Detection on Random Textures.- Semantic-Based Cross-Media Image Retrieval.- Texture Image Retrieval: A Feature-Based Correspondence Method in Fourier Spectrum.- Surface Reconstruction from Stereo Data Using Three-Dimensional Markov Random Field Model.- Unsupervised Markovian Segmentation on Graphics Hardware.- Texture Detection for Image Analysis.- Evaluation of the Quality of Ultrasound Image Compression by Fusion of Criteria with a Genetic Algorithm.- 3D Model Retrieval Based on Adaptive Views Clustering.- Colour Texture Segmentation Using Modelling Approach.- Human-Centered Object-Based Image Retrieval.- Multi-scale Midline Extraction Using Creaseness.- Automatic Indexing of News Videos Through Text Classification Techniques.- Weighted Adaptive Neighborhood Hypergraph Partitioning for Image Segmentation.- Parallel-Sequential Texture Analysis.- Region Growing with Automatic Seeding for Semantic Video Object Segmentation.- Object Coding for Real Time Image Processing Applications.- Designing a Fast Convolution Under the LIP Paradigm Applied to Edge Detection.- Local Feature Saliency for Texture Representation.- A Segmentation Algorithm for Rock Fracture Detection.- ELIS: An Efficient Leaf Image Retrieval System.- Mosaicing and Restoration from Blurred Image Sequence Taken with Moving Camera.- Finding People in Video Streams by Statistical Modeling.- Camera Motion Estimation by Image Feature Analysis.- Shape Retrieval by Principal Components Descriptor.- Automatic Monitoring of Forbidden Areas to Prevent Illegal Accesses.- Dynamic Time Warping of Cyclic Strings for Shape Matching.- Meeting the Application Requirements of Intelligent Video Surveillance Systems in Moving Object Detection.- Classification Using Scale and Rotation Tolerant Shape Signatures from Convex Hulls.- On the Filter Combination for Efficient Image Preprocessing Under Uneven Illumination.- Image Merging Based on Perceptual Information.- An Automated Video Annotation System.- Tracking by Cluster Analysis of Feature Points and Multiple Particle Filters.- Medical Imaging.- A Benchmark for Indoor/Outdoor Scene Classification.- Spinal Deformity Detection Employing Back Propagation on Neural Network.- Bone Segmentation in Metacarpophalangeal MR Data.- Lung Field Segmentation in Digital Postero-Anterior Chest Radiographs.- Relationship Between the Stroma Edge and Skin-Air Boundary for Generating a Dependency Approach to Skin-Line Estimation in Screening Mammograms.- Segmentation of Erythema from Skin Photographs for Assisted Diagnosis in Allergology.- Learning Histopathological Microscopy.- An Adaptive Rule Based Automatic Lung Nodule Detection System.- Experiments with SVM and Stratified Sampling with an Imbalanced Problem: Detection of Intestinal Contractions.- Multiple Particle Tracking for Live Cell Imaging with Green Fluorescent Protein (GFP) Tagged Videos.