Simulation Research Of KNN-Hybrid Algorithm Based On Improved MapReduce Model

Location fingerprint(LF) position technology has overcome some disadvantages of wireless range location, such as high requirement of clock synchronization and significant impact of measurement. This thesis proposes a novel LF algorithm and realizes application based on nowadays rapid developing cloud computing technology in accordance with some characteristics of LF. The main innovative points are as follow:(1) NN-KNN algorithm is proposed; advantages other than traditional method is analyzed based on some particular circumstance.Simulation further proverb its better performance of position especially high accurate position. This method utilizes algorithm accuracy and utilizes MapReduce model of cloud computing to accelerate computing process. Final position results are calculated by weighted average.(2) A more general and flexible KNN-Hybrid algorithm is proposed based on NN-KNN algorithm, and a more adaptive MapReduce model is used. Different form NN-KNN algorithm, KNN-Hybrid increases more data of K parameter in order to enhance the robustness of different environment. A data filter is added in traditional basic MapReduce model, called improved MapReduce model, which fits the process of KNN-Hybrid algorithm and can reduce some, unnecessary data.This thesis proposed positioning model and algorithm are based on traditional positioning technology and current popular cloud computing theory. Theoretical analyze and simulation experiments are made as follows: 1. Positioning mode is proposed and explains of positioning flow are made. 2. Validity of NN-KNN algorithm is analyzed. Its position performance is verified by simulation. Results prove its obvious better accuracy and reliability. 3. KNN-Hybrid algorithm is verified by simulation, which indicates its better performance CEP rate of 2m than NN-KNN algorithm. Data size of KNN-Hybrid algorithm of improved MapReduce model is demonstrated by simulation. Simulation results show that improved MapReduce model can reduce as much as KNN-Hybrid algorithm.