Local POI matching based on KNN and LightGBM Method

High-quality POI data is of great significance to both businesses and individuals. For businesses, POI data is advantageous for management to analyze the competitive landscape and make decisions. For individuals, the accurate location of unknown places will significantly improve the travel experience. Impressive POI matching methods assist to improve the quality of POI data. This paper proposes a novel POI matching approach, which firstly utilzes the KNN algorithm to construct local candidate sets for the problem of high global matching complexity, which significantly improves the optimization efficiency. In addition, our approach leveraging TF-IDF and BERT to re-encode the original features to improve the feature representation. Finally, we build a binary classification model based on LightGBM to improve the matching performance. The experimental results demonstrate that the approach proposed in this paper has achieved effective improvement compared to the DNN and SVM.