Academic & Scientific Articles

Permanent URI for this communityhttp://dl.cerist.dz/handle/CERIST/3

Browse

Has files: YesSubject: search.filters.subject.clustering

Search Results

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    A Clustering application for the Web Usage Mining.
    (Pushpa Publishing House, 2013-07) Kouici, Salima; Khelladi, Abdelkader
    The web usage mining constitutes a new branch of the web mining. It allows the study of the behavior of both users and potential customers via their site navigation. The mainly used source for the web usage mining is the servers log files. A log file contains an important mass of data, including user’s information (username, used software, etc.) and all the queries he has made on the web site (requested files, the number of bytes transferred, time spent on each page, the page of entry to the site, etc.). In this work, we shall outline an application made on this type of data, which is based on a clustering method, namely K-means. This application allows the definition of homogeneous groups constituting users’ profiles so that to anticipate the needs and with a view of communication adapted to each segment of users. In this application, we have recorded some technical problems. These problems concern the data cleaning (removing queries of images and multimedia files associated with web pages, removing queries from search engines, etc.) and the setting up of visitor sessions, knowing that a session is a sequence of pages viewed by the same user.
  • Thumbnail Image
    Item
    Self-Repairing Clusters for Time-Efficient and Scalable Actor-Fault-Tolerance in Wireless Sensor and Actor Networks
    (Springer, 2011) Amirouche, Loucif; Djenouri, Djamel; Badache, Nadjib
    A new solution for fault-tolerance in wireless sensor and actor networks (WSAN) is proposed. The solution deals with fault-tolerance of actors, contrary to most of the literature that only considers sensors. It considers real-time communication, and ensures the execution of tasks with low latency despite fault occurrence. A simplified MAMS (multiple-actor multiple-sensor) model is used, where sensed events are duplicated only to a limited number of actors. This is different from the basic MAMS model and semi-passive coordination (SPC), which use data dissemination to all actors for every event. Although it provides high level of fault- tolerance, this large dissemination is costly in terms of power consumption and communication overhead. The proposed solution relies on the construction of self-repairing clusters amongst actors, on which the simplified MAMS is applied. This clustering enables actors to rapidly replace one another whenever some actor breaks down, and eliminates the need of consensus protocol execution upon fault detection, as required by the current approaches to decide which actor should replace the faulty node. The extensive simulation study carried out with TOSSIM in different scenarios shows that the proposed protocol reduces the latency of replacing faulty actors compared to current protocols like SPC. The reduction of the overall delay for executing actions reaches 59%, with very close fault-tolerance (action execution success rate). The difference for this metric does not exceed 8% in the worst case. Scenarios of different network sizes confirm the results and demonstrate the protocol’s scalability.
  • Thumbnail Image
    Item
    Cluster-Based Fast Broadcast in Duty-Cycled Wireless Sensor Networks
    (IEEE, 2012-08) Khiati, Mustapha; Djenouri, Djamel
    This paper proposes a cluster-based broadcast protocol to disseminate delay-sensitive information throughout a wireless sensor network (WSN). The protocol considers the use of duty-cycling at the MAC layer, which is essential to reduce energy dissipation. LEACH's energy-efficiency approach is used for cluster construction. The proposed protocol adds new common static and dynamic broadcast periods to support and accelerate broadcasting. The dynamic periods are scheduled following the past arrivals of messages, and using a Markov-chain model. To our knowledge, this work is the first that proposes the use of clustering to reduce broadcast latency. The clustering mechanism allows for simultaneous local broadcasts at several clusters in the WSN, and it also ensures scalability with the increase of the network size. The protocol has been simulated, numerically analyzed, and compared with LEACH. The results show clear improvement over LEACH with regard to the latency.