Using Association Rules to Predict Access Points

• Main Authors: Yu-Lun Ho, 何佑倫
• Other Authors: Kuang-Hui Chi
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/32944281183254307605

碩士 === 國立雲林科技大學 === 電機工程系 === 103 === With the development of wireless technology, nowadays our surroundings are full of radio signals. When the user moves about in Wi-Fi network settings, radio link to an access point (AP) may be disrupted because the user has migrated outside coverage of the AP. In order to ensure network connectivity, the user must discovery potential next APs and then select one among them to associate with in time. However, the newly selected AP may not suit. In event that the user associates with an ill-chosen AP, poor communication can result. Therefore, it is important to decide which AP to serve the user best, so as to streamline communication activities. In view that per-user’s movement patterns are likely similar, this thesis presents a predictive handover scheme that enables a mobile user to resolve a best-fit AP in course. We use association rules to determine frequently-used APs. Association rules are of utility to acquire the relevance among items. A typical example is that consumers tend to buy bread and milk at the same time, so bread and milk are related. According to the relevance of APs, when the user is connected to an AP, it is very likely to visit another in the near future. In our architecture, a mobile handset maintains a tree of nodes indicating APs visited in the past one week. Based on the history, the mobile handset examines its current position in the tree (indexed by the identity of current AP) from which the immediate next AP with the maximum likelihood (with the highest frequency of visit) can be deduced. Once the handset detects its received signal strength from the current AP to fall below a certain threshold, a handover to the new AP shall be carried out. Our scheme is implemented over Android and exhibits performance under various movement patterns through field tests.