Abeer El-Korany

A random forest classifier for lymph diseases

Machine learning-based classification techniques provide support for the decision-making process in many areas of health care, including diagnosis, prognosis, screening, etc. Feature selection (FS) is expected to improve classification performance, particularly in situations characterized by the high data dimensionality problem caused by relatively few training examples compared to a large number of measured features. In this paper, a random forest classifier (RFC) approach is proposed to diagnose lymph diseases. Focusing on feature selection, the first stage of the proposed system aims at constructing diverse feature selection algorithms such as genetic algorithm (GA), Principal Component Analysis (PCA), Relief-F, Fisher, Sequential Forward Floating Search (SFFS) and the Sequential Backward Floating Search (SBFS) for reducing the dimension of lymph diseases dataset. Switching from feature selection to model construction, in the second stage, the obtained feature subsets are fed into the RFC for efficient classification. It was observed that GA-RFC achieved the highest classification accuracy of 92.2%. The dimension of input feature space is reduced from eighteen to six features by using GA.

Ensemble classifiers for biomedical data: Performance evaluation

Machine Learning concept offers the biomedical research field a great support. It provides many opportunities for disease discovering and related drugs revealing. The machine learning medical applications had been evolved from the physician needs and motivated by the promising results extracted from empirical studies. Medical support systems can be provided by screening, medical images, pattern classification and microarrays gene expression analysis. Typically medical data is characterized by its huge dimensionality and relatively limited examples. Feature selection is a crucial step to improve classification performance. Recent studies in machine learning field about classification process emerged a novel strong classifier scheme called the ensemble classifier. In this paper, a study for the performance of two novel ensemble classifiers namely Random Forest (RF) and Rotation Forest (ROT) for biomedical data sets is tested with five medical datasets. Three different feature selection methods were used to extract the most relevant features in each data set. Prediction performance is evaluated using accuracy measure. It was observed that ROT achieved the highest classification accuracy in most tested cases.

Hybrid system for lymphatic diseases diagnosis

Machine-learning techniques such as decision support systems (DSS) are of great help in various fields. Medicine is one of the fields that can benefit from the application of data mining and pattern recognition techniques. The evolution of computational intelligence can improve many areas in health care including diagnosis, prognosis, screening, etc. The multiclass classification problem is important in data mining applications. Medical datasets are characterized by high dimensionality. Feature selection is considered as the main process to improve classification performance, particularly with the curse of dimensionality. This paper presents a hybrid system that combines the genetic algorithm (GA) and random forest (RF) for diagnosing lymphatic diseases. The genetic algorithm is used as a feature selection technique for reducing the dimension of the lymphatic diseases dataset and RF is used as a classifier. The performance of the proposed GA-RF system is compared with that of other feature selection algorithms combined with RF classifier such as principal component analysis (PCA), ReliefF, Fisher, sequential forward floating search (SFFS), and the sequential backward floating search (SBFS). The sensitivity and specificity were evaluated to measure the prediction performance. The experiments performed show that GA-RF achieved a high classification accuracy of 92.2%. Moreover, a subset of six features using the GA is sufficient for obtaining the classification.

Towards a suite of quality metrics for KADS-domain knowledge

In this paper, the quality metrics suite for object oriented (OO) design is partially used as an initial concept to develop metrics for statically measuring quality of KADS-domain knowledge bases. KADS-domain knowledge bases have common characteristics with OO, and other distinct features that limit the usage of some OO quality metrics. Therefore, new sets of domain knowledge quality metrics are proposed. The proposed matrices are enriched with an automatic tool used to measure quality of real world expert systems. In order to assess the effectiveness of these proposed quality metrics, these metrics were applied on a sample of real world agriculture application domains developed by CLAES (The Central Laboratory of Agriculture Expert System). Finally, complete analysis of the results obtained due to applying these metrics is presented.

Semantic Topics Modeling Approach for Community Detection

ABSTRACT Social networks play an increasingly important role in online world as it enables individuals to easily share opinions, experiences and expertise. The capability to extract latent communities based on user interest is becoming vital for a wide variety of applications. However, existing literature on community extraction has largely focused on methods based on the link structure of a given social network. Such link-based methods ignore the content of social interactions, which may be crucial for accurate and meaningful community extraction. In this paper, we present a novel approach for community extraction which naturally incorporates the content published within the social network with its semantic features. Two layer generative Restricted Boltzmann Machines model is applied for community discovery. The model assumes that users within a community communicate based on topics of mutual interest. The proposed model naturally allows users to belong to multiple communities. Through extensive experiments on the Twitter data for scientific papers, we demonstrate that the model is able to extract well-connected and topically meaningful communities.

A supervised learning approach to link prediction in Twitter

The growth of social networks has lately attracted both academic and industrial researchers to study the ties between people, and how the social networks evolve with time. Social networks like Facebook, Twitter and Flickr require efficient and accurate methods to recommend friends to their users in the network. Several algorithms have been developed to recommend friends or predict likelihood of future links. Two main approaches are used to utilize those features; Score-based Approaches and Machine Learning Approaches. In a previous work, a score-based method was used based on topological, node and social features to calculate similarity between users and determine the likelihood of forming future links. This work has been extended by moving to a Machine Learning Approach which treats the prediction process as a classification problem. The classifier predicts the class of each edge whether it exists or doesn’t exist. Machine Learning Approaches have the benefit of adding all similarity indices needed as the feature set fed to the classifier. While in Score-based Approach when we used multiple features with associated weights, the performance was sensitive to the values of such weights. When machine learning is applied, the learning process is performed by the classifier which is fed by eight similarity indices representing connectivity, community, interaction and trust in social network. When indices are combined, a much higher accuracy than the previous Score-based Approach is obtained and hence enhancing the prediction accuracy. In order to evaluate the correctness of the proposed model, it has been applied on a real dataset of 2.974k users on the Twitter social network. Experiments show that using both classical and ensemble classifiers outperforms baseline algorithms when applied individually.

Lymph diseases diagnosis approach based on support vector machines with different kernel functions

In this paper, a Genetic algorithm (GA) based supporting vector machine classifier (GA-SVM) is proposed for lymph diseases diagnosis. In the first stage, dimension of lymph diseases dataset that has 18 features is reduced to six features using GA. In the second stage, a support vector machine with different kernel functions including linear, Quadratic and Gaussian was utilized as a classifier. The Lymphography database was obtained from the University Medical Center, Institute of Oncology, Ljubljana, Yugoslavia. The obtained classification accuracy was very promising with regard to the other classification applications in the literature for this problem. The performance of SVM classifier with each kernel function was evaluated by using performance indices such as accuracy, sensitivity, specificity, area under curve (AUC) or (ROC), Matthews Correlation Coefficient (MCC) and F-Measure. Linear kernel function obtained highest results which verifies the efficiency of GA-linear stategy.

Integrating Semantic Features for Enhancing Arabic Named Entity Recognition

Named Entity Recognition (NER) is currently an essential research area that supports many tasks in NLP. Its goal is to find a solution to boost accurately the named entities identification. This paper presents an integrated semantic-based Machine learning (ML) model for Arabic Named Entity Recognition (ANER) problem. The basic idea of that model is to combine several linguistic features and to utilize syntactic dependencies to infer semantic relations between named entities. The proposed model focused on recognizing three types of named entities: person, organization and location. Accordingly, it combines internal features that represented linguistic features as well as external features that represent the semantic of relations between the three named entities to enhance the accuracy of recognizing them using external knowledge source such as Arabic WordNet ontology (ANW). We introduced both features to CRF classifier, which are effective for ANER. Experimental results show that this approach can achieve an overall F-measure around 87.86% and 84.72% for ANERCorp and ALTEC datasets respectively.

Ensemble-based classifiers for prostate cancer diagnosis

In this paper, we address microarray data sets dimensionality problem to achieve early and accurate diagnosis of prostate cancer without need to biopsy operation based rotation multiple classifier forest system. To evaluate the performance of presented approach, we present tests on different prostate data sets. The experimental results obtained, show that the overall accuracy offered by the employed technique is high compared with other machine learning techniques including random forest classifier, single decision trees and rough sets as well as features were reduced from 12600 features to 89 features using correlation filter method.

Enabling Semantic User Context to Enhance Twitter Location Prediction

Prediction of user interest and behavior is currently an important research area in social network analysis. Most of the current prediction frameworks rely on analyzing user’s published contents and user’s relationships. Recently the dynamic nature of user’s modelling has been introduced in the prediction frameworks. This dynamic nature would be represented by time tagged attributes such as posts or location check-ins. In this paper, we study the relationships between geo-location information published by users at different times. This geo-location information was used to model user’s interest and behavior in order to enhance prediction of user locations. Furthermore, semantic features such as topics of interest and location category were extracted from this information in order to overcome sparsity of data. Several experiments on real twitter dataset showed that the proposed context-based prediction model which applies machine learning techniques outperformed traditional probabilistic location prediction model that only rely on words extracted from tweets associated with specific locations.