Interior Fusion Localization Method Based on Random Forest

Abstract

This paper introduces a high performance indoor location method based on random forest. Firstly, the access point with poor signal strength and unstable signal is deleted through the signal strength of the access point. Then the information gain method is used to select the access points with better positioning effect from the remaining access points to form the access points set and establish the location fingerprint database. Then, all positions in the location scene are grouped by clustering algorithm, and then a random forest model is built for each position. The location process USES the random forest model to determine the user location. The experimental results show that by constructing a high precision and high stability stochastic forest model, this paper can effectively solve the problem of limited positioning accuracy, unstable positioning effect and easy to fall into overfitting of the single-decision tree model, and improve the positioning stability and positioning accuracy. The average error of the indoor positioning method mentioned in this paper is at least 1.3718m, and the variance of the positioning result is calculated on this basis, which is 0.0173 in this paper.