INTELLIGENT MOBILE DATA DISSEMINATION USING DATA MINING TECHNIQUE

Abstract

Management and dissemination of data is an importan t problem in a mobile environ-ment. Communication channels in the wireless networks are typically asymmetric in nature where the server-to-client (downlink) communication bandwidth is much higher than the client-to-server (uplink) communication bandwidth. Minimizing access latency is a challenging optimization problem in this framework. The thesis proposes a system which uses different data mining techniques along with various system parameters for optimal data dissemination at server side. The broadcast pattern in the downlink is intelligently determined to minimize user ac-cess latency. The system considers the client's behavior over a period of time and dynamically reflects those characteristics in the next broadcast sequence by mining sequential patterns from the user's requests history. As new requests come over the uplink channel which were not fulfilled by downlink channel, the mined sequential patterns gets updated and hence the broadcast order changes with time. We have modified and used incremental sequence pattern mining algorithm which does not require re-mining at every instant. With our experiments we have shown that incre-mental data dissemination using data mining outperforms static mining techniques in terms of computational cost of mining. It is also superior to frequency based data dissemination in terms of average access latency per request. With the assumption that there is some cache present at the client side we have developed cache manage - ment algorithm and shown that the average latency can be further reduced using the same mined sequential patterns. The system runs on windows xp environment with PIV 2.40 GHz processor and 256 MB RAM. The language used for code development is