Ahmed Fawzi Otoom

A dynamic replication strategy based on categorization for Data Grid

Data Replication is copying the data from a certain location to another location. Replication is used in Data Grid to have two or more copies of the same data at different locations. In this paper, a Category-based dynamic replication strategy (CDRS) is proposed. The strategy takes into account that the replicas exist on a node belong to different categories. Each of these categories is given a value that determines its importance for the node. When the nodes storage is full, the node starts to store only the replicas that belong to the category with the highest value. The results of the simulations show that the new proposed strategy achieved better performance than Plain Caching and Fast Spread strategies in terms of total transit time and total bandwidth consumption.

Towards author identification of Arabic text articles

We target the problem of identifying the author of an Arabic text article. Our main aim is to develop an intelligent system that is capable of classifying a new article into one of seven classes that belong to seven different authors. For this purpose, we propose a novel dataset consisting of 12 features and 456 instances belonging to the 7 authors. In addition, we combine the proposed feature set with strong classification algorithms to assist in distinguishing between the different authors. Our results show that the proposed dataset has proved successful with a classification performance accuracy of 82% with the hold-out test.

Simplified features for email authorship identification

We present an investigation analysis approach for mining anonymous email content. The core idea behind our approach is concentrated on collecting various effective features from previous emails for all the possible suspects. The extracted features are then used with several machine learning algorithms to extract a unique writing style for each suspect. A sophisticated comparison between the investigated anonymous email and the suspects writing styles is employed to extract evidence of the possible email sender. Extensive experimental results on a real data sets show the improved performance of the proposed method with very limited number of features.

Randomized geographic-based routing with nearly guaranteed delivery for three-dimensional ad hoc network

Several routing algorithms have been proposed for efficient routing in mobile ad hoc networks, most of them consider mobile nodes embedded in two-dimensional environments. However, in reality, these networks are embedded in three-dimensional environments. Usually, two-dimensional routing algorithms have several assumptions that are not valid for three-dimensional spaces. In this article, we propose four different randomized geographic-based routing algorithms that have the following properties: (1) nearly guaranteed delivery rate, by using randomize route to overcome local minimum problems; (2) low overhead, by extracting a virtual backbone of the network and then conducting the routing algorithms over the extracted backbone to decrease the search space; (3) low path dilation, by hybridizing the new algorithms with progress-based routing which have very low path dilation; and (4) works in three-dimensional environment. The first algorithm 3DRanDom chooses the next neighbor randomly from a dominating set of the network (extracted locally). The second algorithm 3DRanDomProb extracts a dominating set and sends to one of the resulted neighbors randomly with more probability for the nodes closer to the destination. The third algorithm G_3DRanDomProb tries to progress as much as possible to the destination, if the progress is not possible, the algorithm switches to 3DRanDomProb. The fourth algorithm G_3DRanDomProb_G uses progress-based routing as much as possible, then it switches to 3DRanDomProb until it overcomes the local minimum problem and then goes back to progress-based routing. We show experimentally that these hybrid randomized routing algorithms on three-dimensional mobile ad hoc networks can achieve nearly guaranteed delivery while discovering routes significantly closer in length to the shortest path and with low overhead.

An intelligent system for author attribution based on a hybrid feature set

Authorship analysis is a long explored area in the computational research. Recently, there has been growing interest in developing intelligent systems that are capable of authorship identification. Inspired by recent works, we address the problem of author attribution of Arabic text. This area, in specific, has not been targeted in the literature except for few studies. However, it is a challenging problem as there are linguistic complexities associated with the Arabic language including elongation and inflection challenges. For this purpose, we propose a novel hybrid feature set consisting of: lexical, syntactic, structural and content-specific features for 456 instances belonging to seven different Arabic authors. For validation, we run extensive experiments with different intelligent classifiers and show the strength of the proposed feature set. Our results show that the proposed feature set has proved successful with a classification performance accuracy of 88% with the hold-out test and 82% with the cross-validation test.

A genetic algorithms-based hybrid recommender system of matrix factorization and neighborhood-based techniques

Abstract Recommender system (RS) is the current applications’ main choice to guide the customers in choosing their favorite items. A collaborative filtering (CF) RSs use either the neighbourhood or the latent factor models to recommend items for the active user (customer). The matrix factorization method are widely used in the latent factor model to find the high-expected rated items and hence highly favoured items by the active user. Navgaran’ et al. build a genetic-based matrix-factorization RS to make recommendation for the active user. In this project, we build a novel genetic-based CF RS that hybridizes both the neighbourhood and the latent factor models to predict items for the active user. The main difference between our RS and Navgaran’ et al. RS is that, we only consider the users and items that are related to the active user rather than considering the whole users and items. Using three different datasets: MovieLens, FilmTrust, and CiaoDVD, we experimentally show that our model improved Navgaran’ et al. RS in term of MAE, precision, and recall. We show that our model is at least 15.2 times faster and has at least 87% less MAE value than Navgaran’ et al. RS. In addition, it has better recall and precision values in MovieLens and CiaoDVD datasets.

Severity prediction of software bugs

We target the problem of identifying the severity of a bug report. Our main aim is to develop an intelligent system that is capable of predicting the severity of a newly submitted bug report through a bug tracking system. For this purpose, we build a dataset consisting of 59 features characterizing 163 instances that belong to two classes: severe and non-severe. We combine the proposed feature set with strong classification algorithms to assist in predicting the severity of bugs. Moreover, the proposed algorithms are integrated within a boosting algorithm for an enhanced performance. Our results show that the proposed technique has proved successful with a classification performance accuracy of more than 76% with the AdaBoost algorithm and cross validation test. Moreover, boosting has been effective in enhancing the performance of its base classifiers with improvements of up to 4.9%.

Local Search Algorithms for Solving the Combinatorial Optimization and Constraint Satisfaction Problems

Local search is a metaheuristic for solving computationally hard optimization problems. In the past three decades, local search has grown from a simple heuristic idea into a mature field of research in combinatorial optimization that is attracting ever-increasing attention. It is still the method of choice for NP-hard problems as it provides a robust approach for obtaining high-quality solutions to problems of a realistic size in reasonable time. Optimization problems such as the shortest path, the traveling salesman, pin packing, and the Knapsack problems. Local search techniques have been successful in solving large and tight constraint satisfaction problems. Local search algorithms turn out to be effective in solving many constraint satisfaction problems. This chapter gives an introduction to the local search algorithms, the optimization and the constraint satisfaction problems, and the local search methods used to solve them.

Spectral Graph and Minimal Spanning Tree for 3D Polygonal Meshes Fingerprinting

In this paper, the authors present a robust three-dimensional fingerprint algorithm for verification, indexing, and identification. The core idea behind our technique is to apply the eigen-decomposition to the mesh Laplacian matrix, and then compute minimum spanning trees MST of several concentrations of the mesh shape structure. The fixed size hash vector of a 3D mesh is defined in terms of the MST values and number of the initial patches. The extensive experimental results on several 3D meshes prove the uniqueness of the extracted hash vectors and the robustness of the proposed technique against the most common attacks including distortion-less attacks, compression, noise, smoothing, scaling, rotation as well as mixtures of these attacks.