Improvement And Research Of Indoor Localization Algorithm Based On Machine Learning Theory

In recent years,the wireless communication technology is developing at a very amazing speed.With its driving,intelligent terminals are rapidly becoming popular which have made a great impact on location based services.Meanwhile,the localization of the most common indoor scenes in people's daily activities has gained wide attention.Indoor localization technology with high precision has shownfields of social network,emergency rescue,medical care and other fields.good application prospects in theIndoor localization algorithm based on fingerprint with the reason that it's not susceptible to the influence of signal propagation path in space and it doesn't need additional deployment of electronic devices,but maximally using the existing public network infrastructure and effectively reducing the cost of positioning,so it has attracted much attention and has become a hot topic in the research institutions.But the existing traditional methods need to harass the people and waste money.It needs to select reference points,and at the same time,need to sample many times to establish location fingerprint database.In addition,fingerprint information is vulnerable to the influence of multipath and environmental changes,resulting in regular updating of fingerprint database,which brings great challenges to the popularity of indoor localization system.Therefore,this paper proposes an indoor localization algorithm based on machine learning.And the following research work has been carried out:(1)The influence of sampling count on indoor localization results is analyzed.It is found that the more the number of sample points is,the better the location effect is,but when the number of sample points reaches a certain number,the location result tends to be stable,and basically no longer changes;(2)Do the data preprocessing to the fingerprint data.We need to filter the data before locating,so as to remove the fluctuating signal data and reduce the range of signal fluctuation,and further reduce the location error caused by random noise.At the same time,an improved access points selection algorithm is proposed to reduce the data dimension and computation time and improve the location effect;(3)A fingerprint clustering model based on fuzzy C means is proposed.A method for judging the quality of the clustering results in the localization algorithm is proposed.It is verified that the clustering model is used to divide the fingerprint database into region,which not only shortens the time of calculation,but also reduces the location error;(4)A cluster pruning algorithm is proposed.So a small number of points in the boundary parts that may misjudged by the classifier can be removed.In this paper,the support vector machine is used to distinguish the region of data.Then,the support vector regression is used to estimate the specific coordinates of the user's position,realizing the localization.In this procedure,the particle swarm optimization algorithm is used to select the parameters of the model,and an improved particle swarm optimization algorithm is proposed;(5)Finally,the simulation model is used to achieve indoor positioning and compare the effects of various methods for positioning results.At the same time,the localization algorithm is verified in real scenes and compared with other localization algorithms.The test results show that the algorithm presented in this paper has superior performance.