MODIFICATION OF LOG-NORMAL PREDICTION MODEL FOR HSPA NETWORK USING ADAPTIVE NEURO-FUZZY INFERENCE SYSTEM
The transmission of radio signals over a channel for proper path-loss prediction is a core aspect of planning in wireless communication. Some conventional path-loss prediction models such as Log-normal, Okumura-Hata and COST 231 models are not appropriate for predicting the path-loss values due to differences in frequencies of operation which, therefore, need adaptation before employing. This paper, therefore, modifies the Log-normal prediction model for High Speed Packet Access (HSPA) using Adaptive Neuro-fuzzy Inference System (ANFIS). The modification is carried out by measuring the Received Signal Strength (RSS) using drive test at Ayetoro area of Lagos, Nigeria on (Longitude 3.19647E and Latitude 6.59167 N). The drive test equipment consists of a computer system integrated with Test Equipment for Mobile System (TEMS) software, Ericson TEMS phone and Global Positioning System (GPS). Suitability of the conventional models is determined using Base Station (BS) parameters of the network after which the modification of Log-normal prediction model is carried out by obtaining the path-loss exponent. The path-loss exponent is used to determine the deviation for proper modification. The modified model is further enhanced using ANFIS model which is developed by training five layer ANFIS architecture for adaptation. The models are evaluated using path-loss values and Root Mean Square Error (RMSE) to determine the performances. The results obtained show that ANFIS, COST 231, and modified Log-normal models give the lowest RMSE values with their path-loss values closest to the measured values. Therefore, these models are suitable for predicting the HSPA signal in this area and can be used for future planning of wireless network.