Feature Extraction for Classification in the Data Mining Process

Dimensionality reduction is a very important step in the data mining process. In this paper, we consider feature extraction for classification tasks as a technique to overcome problems occurring because of “the curse of dimensionality”. Three different eigenvector-based feature extraction approaches are discussed and three different kinds of applications with respect to classification tasks are considered. The summary of obtained results concerning the accuracy of classification schemes is presented with the conclusion about the search for the most appropriate feature extraction method. The problem how to discover knowledge needed to integrate the feature extraction and classification processes is stated. A decision support system to aid in the integration of the feature extraction and classification processes is proposed. The goals and requirements set for the decision support system and its basic structure are defined. The means of knowledge acquisition needed to build up the proposed system are considered.

Entropy feature extraction of EEG signals for automatic person identification

The use of human brain electroencephalography (EEG) signals for automatic person identi cation has been investigated for a decade. It has been found that the performance of an EEG-based person identication system highly depends on what feature to be extracted from multi-channel EEG signals. Linear methods such as Power Spectral Density and Autoregressive Model have been used to extract EEG features. However these methods assumed that EEG signals are stationary. In fact, EEG signals are complex, non-linear, non-stationary, and random in nature. In addition, other factors such as brain condition or human characteristics may have impacts on the performance, however these factors have not been investigated and evaluated in previous studies. It has been found in the literature that entropy is used to measure the randomness of non-linear time series data. Entropy is also used to measure the level of chaos of braincomputer interface systems. Therefore, this thesis proposes to study the role of entropy in non-linear analysis of EEG signals to discover new features for EEG-based person identi- cation. Five dierent entropy methods including Shannon Entropy, Approximate Entropy, Sample Entropy, Spectral Entropy, and Conditional Entropy have been proposed to extract entropy features that are used to evaluate the performance of EEG-based person identication systems and the impacts of epilepsy, alcohol, age and gender characteristics on these systems. Experiments were performed on the Australian EEG and Alcoholism datasets. Experimental results have shown that, in most cases, the proposed entropy features yield very fast person identication, yet with compatible accuracy because the feature dimension is low. In real life security operation, timely response is critical. The experimental results have also shown that epilepsy, alcohol, age and gender characteristics have impacts on the EEG-based person identication systems.

Ontology-based feature modeling for construction information extraction from a building information model

A building information model (BIM) provides a rich representation of a building's design. However, there are many challenges in getting construction-specific information from a BIM, limiting the usability of BIM for construction and other downstream processes. This paper describes a novel approach that utilizes ontology-based feature modeling, automatic feature extraction based on ifcXML, and query processing to extract information relevant to construction practitioners from a given BIM. The feature ontology generically represents construction-specific information that is useful for a broad range of construction management functions. The software prototype uses the ontology to transform the designer-focused BIM into a construction-specific feature-based model (FBM). The formal query methods operate on the FBM to further help construction users to quickly extract the necessary information from a BIM. Our tests demonstrate that this approach provides a richer representation of construction-specific information compared to existing BIM tools.

Mapping semantic knowledge for unsupervised text categorisation

Text categorisation is challenging, due to the complex structure with heterogeneous, changing topics in documents. The performance of text categorisation relies on the quality of samples, effectiveness of document features, and the topic coverage of categories, depending on the employing strategies; supervised or unsupervised; single labelled or multi-labelled. Attempting to deal with these reliability issues in text categorisation, we propose an unsupervised multi-labelled text categorisation approach that maps the local knowledge in documents to global knowledge in a world ontology to optimise categorisation result. The conceptual framework of the approach consists of three modules; pattern mining for feature extraction; feature-subject mapping for categorisation; concept generalisation for optimised categorisation. The approach has been promisingly evaluated by compared with typical text categorisation methods, based on the ground truth encoded by human experts.

A data mining framework for relevance feature discovery

This thesis is a study for automatic discovery of text features for describing user information needs. It presents an innovative data-mining approach that discovers useful knowledge from both relevance and non-relevance feedback information. The proposed approach can largely reduce noises in discovered patterns and significantly improve the performance of text mining systems. This study provides a promising method for the study of Data Mining and Web Intelligence.

Enhanced n-gram extraction using relevance feature discovery

Guaranteeing the quality of extracted features that describe relevant knowledge to users or topics is a challenge because of the large number of extracted features. Most popular existing term-based feature selection methods suffer from noisy feature extraction, which is irrelevant to the user needs (noisy). One popular method is to extract phrases or n-grams to describe the relevant knowledge. However, extracted n-grams and phrases usually contain a lot of noise. This paper proposes a method for reducing the noise in n-grams. The method first extracts more specific features (terms) to remove noisy features. The method then uses an extended random set to accurately weight n-grams based on their distribution in the documents and their terms distribution in n-grams. The proposed approach not only reduces the number of extracted n-grams but also improves the performance. The experimental results on Reuters Corpus Volume 1 (RCV1) data collection and TREC topics show that the proposed method significantly outperforms the state-of-art methods underpinned by Okapi BM25, tf*idf and Rocchio.

Centroid training to achieve effective text classification

Traditional text classification technology based on machine learning and data mining techniques has made a big progress. However, it is still a big problem on how to draw an exact decision boundary between relevant and irrelevant objects in binary classification due to much uncertainty produced in the process of the traditional algorithms. The proposed model CTTC (Centroid Training for Text Classification) aims to build an uncertainty boundary to absorb as many indeterminate objects as possible so as to elevate the certainty of the relevant and irrelevant groups through the centroid clustering and training process. The clustering starts from the two training subsets labelled as relevant or irrelevant respectively to create two principal centroid vectors by which all the training samples are further separated into three groups: POS, NEG and BND, with all the indeterminate objects absorbed into the uncertain decision boundary BND. Two pairs of centroid vectors are proposed to be trained and optimized through the subsequent iterative multi-learning process, all of which are proposed to collaboratively help predict the polarities of the incoming objects thereafter. For the assessment of the proposed model, F1 and Accuracy have been chosen as the key evaluation measures. We stress the F1 measure because it can display the overall performance improvement of the final classifier better than Accuracy. A large number of experiments have been completed using the proposed model on the Reuters Corpus Volume 1 (RCV1) which is important standard dataset in the field. The experiment results show that the proposed model has significantly improved the binary text classification performance in both F1 and Accuracy compared with three other influential baseline models.

Nested dictionary learning for hierarchical organization of imagery and text

A tree-based dictionary learning model is developed for joint analysis of imagery and associated text. The dictionary learning may be applied directly to the imagery from patches, or to general feature vectors extracted from patches or superpixels (using any existing method for image feature extraction). Each image is associated with a path through the tree (from root to a leaf), and each of the multiple patches in a given image is associated with one node in that path. Nodes near the tree root are shared between multiple paths, representing image characteristics that are common among different types of images. Moving toward the leaves, nodes become specialized, representing details in image classes. If available, words (text) are also jointly modeled, with a path-dependent probability over words. The tree structure is inferred via a nested Dirichlet process, and a retrospective stick-breaking sampler is used to infer the tree depth and width.

A statistical framework for natural feature representation

This paper presents a robust stochastic framework for the incorporation of visual observations into conventional estimation, data fusion, navigation and control algorithms. The representation combines Isomap, a non-linear dimensionality reduction algorithm, with expectation maximization, a statistical learning scheme. The joint probability distribution of this representation is computed offline based on existing training data. The training phase of the algorithm results in a nonlinear and non-Gaussian likelihood model of natural features conditioned on the underlying visual states. This generative model can be used online to instantiate likelihoods corresponding to observed visual features in real-time. The instantiated likelihoods are expressed as a Gaussian mixture model and are conveniently integrated within existing non-linear filtering algorithms. Example applications based on real visual data from heterogenous, unstructured environments demonstrate the versatility of the generative models.

Wide baseline correspondence extraction beyond local features

Robust, affine covariant, feature extractors provide a means to extract correspondences between images captured by widely separated cameras. Advances in wide baseline correspondence extraction require looking beyond the robust feature extraction and matching approach. This study examines new techniques of extracting correspondences that take advantage of information contained in affine feature matches. Methods of improving the accuracy of a set of putative matches, eliminating incorrect matches and extracting large numbers of additional correspondences are explored. It is assumed that knowledge of the camera geometry is not available and not immediately recoverable. The new techniques are evaluated by means of an epipolar geometry estimation task. It is shown that these methods enable the computation of camera geometry in many cases where existing feature extractors cannot produce sufficient numbers of accurate correspondences.

The use of temporal speech and lip information for multi-modal speaker identification via multi-stream HMMs

Investigates the use of temporal lip information, in conjunction with speech information, for robust, text-dependent speaker identification. We propose that significant speaker-dependent information can be obtained from moving lips, enabling speaker recognition systems to be highly robust in the presence of noise. The fusion structure for the audio and visual information is based around the use of multi-stream hidden Markov models (MSHMM), with audio and visual features forming two independent data streams. Recent work with multi-modal MSHMMs has been performed successfully for the task of speech recognition. The use of temporal lip information for speaker identification has been performed previously (T.J. Wark et al., 1998), however this has been restricted to output fusion via single-stream HMMs. We present an extension to this previous work, and show that a MSHMM is a valid structure for multi-modal speaker identification

Spatio temporal feature evaluation for action recognition

Spatio-Temporal interest points are the most popular feature representation in the field of action recognition. A variety of methods have been proposed to detect and describe local patches in video with several techniques reporting state of the art performance for action recognition. However, the reported results are obtained under different experimental settings with different datasets, making it difficult to compare the various approaches. As a result of this, we seek to comprehensively evaluate state of the art spatio- temporal features under a common evaluation framework with popular benchmark datasets (KTH, Weizmann) and more challenging datasets such as Hollywood2. The purpose of this work is to provide guidance for researchers, when selecting features for different applications with different environmental conditions. In this work we evaluate four popular descriptors (HOG, HOF, HOG/HOF, HOG3D) using a popular bag of visual features representation, and Support Vector Machines (SVM)for classification. Moreover, we provide an in-depth analysis of local feature descriptors and optimize the codebook sizes for different datasets with different descriptors. In this paper, we demonstrate that motion based features offer better performance than those that rely solely on spatial information, while features that combine both types of data are more consistent across a variety of conditions, but typically require a larger codebook for optimal performance.

Product feature taxonomy learning based on user reviews

In recent years, the Web 2.0 has provided considerable facilities for people to create, share and exchange information and ideas. Upon this, the user generated content, such as reviews, has exploded. Such data provide a rich source to exploit in order to identify the information associated with specific reviewed items. Opinion mining has been widely used to identify the significant features of items (e.g., cameras) based upon user reviews. Feature extraction is the most critical step to identify useful information from texts. Most existing approaches only find individual features about a product without revealing the structural relationships between the features which usually exist. In this paper, we propose an approach to extract features and feature relationships, represented as a tree structure called feature taxonomy, based on frequent patterns and associations between patterns derived from user reviews. The generated feature taxonomy profiles the product at multiple levels and provides more detailed information about the product. Our experiment results based on some popularly used review datasets show that our proposed approach is able to capture the product features and relations effectively.