珠海市是新时代粤港澳大湾区建设的重要节点城市，有着巨大的发展空间和潜力。近些年，珠海市房地产业发展十分迅猛，二手房交易量不断增加。然而，关于珠海市二手房价格问题的研究却很少，目前尚无文献研究珠海二手房价格的影响因素以及在空间分布上的差异。因此，本文对上述问题进行研究，对该市二手房市场的发展具有重要意义，为购房者、中介、政府等多方提供决策依据。

本文从二手房网站及高德地图获取珠海市2022年1月的2135条房源数据、POI数据。首先进行探索性数据分析，包括描述统计分析、空间趋势分析、空间自相关分析，分析该市房价在整体上以及各区域内的特征。然后基于相关理论并结合实际情况选取了11个变量，构建特征价格模型(Hedonic Price Model , HPM)，采用逐步回归的筛选策略选取关键变量。由于HPM忽略了空间效应，因此又基于HPM选取的关键变量建立空间计量模型。具体地，分析空间依赖性时运用了空间滞后模型(Spatial Lag Model , SLM)以及空间误差模型(Spatial Error Model , SEM)，分析空间异质性时运用地理加权回归(Geographically Weighted Regression , GWR)。结论如下：（1）由探索性数据分析知，珠海市二手房价格在东西方向上呈现自西向东增长的趋势，在南北方向上差异小；在全局空间上表现为正相关，在局部空间上大多表现为“高高集聚”“低低集聚”。（2）在不同函数形式的HPM中，对数形式的模型拟合效果最好，其回归结果表明，房龄、绿化率、容积率、物业费，港珠澳大桥珠海口岸、大型商超、大学、医院、风景名胜这9个变量对珠海市二手房价格的影响显著。（3）构建SLM、SEM并与HPM相比较，SEM具有更高的、更小的对数似然值及AIC值，拟合更好，证实了空间依赖性的存在。（4）构建GWR模型并与HPM相比较，GWR模型更优，证实了空间异质性的存在。各因素按照空间分异程度从大到小的顺序排列，依次为：医院>港珠澳大桥珠海口岸>大学>绿化率>大型商超>风景名胜>房龄>物业费>容积率。

As a pivotal city in the construction of the Guangdong-Hong Kong-Macao Greater Bay Area in the new era, Zhuhai has huge development space and potential. In recent years, the real estate market in Zhuhai has developed rapidly, and the transaction volume of second-hand houses has increased continuously. However, there is only limited research on the price of second-hand housing in Zhuhai. At present, there is no literature to study the influencing factors of second-hand housing price in Zhuhai and the difference in its spatial distribution. Therefore, this paper takes the price of second-hand housing in Zhuhai City as the research object to study its spatial distribution differences and influencing factors. They are of great significance to the development of the second-hand housing market in the city. So, they are also decision-making basis for home buyers, real estate agencies, and the government.

From second-hand housing websites and AutoNavi maps, this paper uses web crawler technology to obtain 2,135 house data and POI data in Zhuhai in January 2022. The first is exploratory data analysis, including descriptive statistical analysis, spatial trend analysis, and spatial correlation analysis. It is conducted to analyze the characteristics of housing prices as a whole and in different regions. Then, based on relevant theories and the current situation of Zhuhai City, 11 variables are selected to construct a Hedonic Price Model (HPM). A stepwise regression screening strategy is used to select the key influencing factors. Since HPM ignores spatial effects, spatial econometric models are constructed based on the key variables selected by HPM. Specifically, a Spatial Lag Model (SLM) and a Spatial Error Model (SEM) are constructed to analyze spatial dependencies, and Geographic Weighted Regression (GWR) is used to analyze spatial heterogeneity. The conclusions are as follows:

(1) According to the exploratory data analysis, the price of second-hand housing in Zhuhai shows an increasing trend from west to east, with little difference in the north-south direction. In the global space, it shows a positive correlation, while in the local space, it shows “high-high agglomeration” and “low-low agglomeration” in most cases.

(2) Among the hedonic price models of different functional forms, the logarithmic model has the best fitting effect. Universities, hospitals, and scenic spots have a significant impact on the price of second-hand housing in Zhuhai.

(3) The spatial lag model and the spatial error model are constructed and compared with the hedonic price model. The spatial error model has higher , smaller log-likelihood, AIC values, and better fitting effcet, confirming the spatial existence of dependencies.

(4) By constructing the GWR model and comparing with the hedonic price model, the GWR model is better, which confirms the existence of spatial heterogeneity. The degree of spatial differentiation of each factor is in descending order: hospitals > Zhuhai Port of Hong Kong-Zhuhai-Macao Bridge > universities > greening rate > large supermarkets > scenic spots > age of house > property fee > floor area ratio.