A points of interest matching method using a multivariate weighting function with gradient descent optimization

Abstract

Volunteered geographic information contains abundant valuable data, which can be applied to various spatiotemporal geographical analyses. While the useful information may be distributed in different, low‐quality data sources, this issue can be solved by data integration. Generally, the primary task of integration is data matching. Unfortunately, due to the complexity and irregularities of multi‐source data, existing studies have found it difficult to efficiently establish the correspondence between different sources. Therefore, we present a multi‐stage method to match multi‐source data using points of interest. A spatial filter is constructed to obtain candidate sets for geographical entities. The weights of non‐spatial characteristics are examined by a machine learning‐related algorithm with artificially labeled random samples. A case study on Fuzhou reveals that an average of 95% of instances are accurately matched. Thus, our study provides a novel solution for researchers who are engaged in data mining and related work to accurately match multi‐source data via knowledge obtained by the idea and methods of machine learning.

Conflating point of interest (POI) data: A systematic review of matching methods

“…Wang et al, 2020). Another method is to take into account the curvature of the surface of the Earth and measure geodesic distance (McKenzie et al, 2014;Tré et al, 2013;Zhang & Yao, 2018;Toccu et al, 2019;Zhou et al, 2021;Zhao et al, 2022). As distance measures are often in real numbers, researchers also developed methods to convert real-value spatial distances to similarity measures ranging from 0 to 1.…”

“…Some studies also first standardized addresses into the same address format (e.g., using the format of door number, street name, city, postcode, country) before measuring their similarity (J. Liu et al, 2013;Morana et al, 2014;Zhou et al, 2021). A second approach is to segment an address into individual elements, such as street name, city name, and postcode, and then calculate the similarity scores over these individual elements (Y.…”

“…For example, McKenzie et al (2014) trained a binomial probit regression model to combine multiple similarity scores to determine POI matches. Zhou et al (2021) used a similar approach but trained a logistic regression model. Zhang and Yao (2018) trained a particle swarm optimization (PSO) model for tuning attribute similarity weights by optimizing an objective function for POI matching.…”

“…Other distance thresholds were also used by researchers (F. Yu et al, 2016;Zhou et al, 2021). Xing et al, (2022) adopted the k-nearest neighbors algorithm to look for the 10 closest candidate POI pairs based on their spatial locations.…”

Fine-tuned Predictive Model for Verifying POI Data

“…POI classification using machine learning approaches were studied [27][28]. Some researchers compared POI data extracted from different sources and observed the variations [29]. Machine learning methods were used to compare POI datasets extracted from different sources.…”