1# RECORD 67d132eca470e26b7a8c41462ca1252d056a93bb2# TITLE: Spatial and Social Media Data Analytics of Housing Prices in Shenzhen, China3# YEAR: 2016.0 VENUE: PLoS ONE DOI: 10.1371/journal.pone.01645534# AUTHORS: Chao Wu; Xin-Yue Ye; F. Ren; Y. Wan; Peng-Fei Ning; Qingyun Du5# PAGES: 19 CANDIDATE TABLE PAGES (1-based): [6, 13, 14]6# ABSTRACT: Housing is among the most pressing issues in urban China and has received considerable scholarly attention. Researchers have primarily concentrated on identifying the factors that influence residential property prices and how such mechanisms function. However, few studies have examined the potential factors that influence housing prices from a big data perspective. In this article, we use a big data perspective to determine the willingness of buyers to pay for various factors. The opinions and geographical preferences of individuals for places can be represented by visit frequencies given different motivations. Check-in data from the social media platform Sina Visitor System is used in this article. Here, we use kernel density estimation (KDE) to analyse the spatial patterns of check-in spots (or places of interest, POIs) and employ the Getis-Ord Gi* method to identify the hot spots for different types of POIs in Shenzhen, China. New indexes are then proposed based on the hot-spot results as measured by check-in data to analyse the effects of these locations on housing prices. This modelling is performed using the hedonic price method (HPM) and the geographically weighted regression (GWR) method. The results show that the degree of clustering of POIs has a significant influence on housing values. Meanwhile, the GWR method has a better interpretive capacity than does the HPM because of the former method’s ability to capture spatial heterogeneity. This article integrates big social media data to expand the scope (new study content) and depth (study scale) of housing price research to an unprecedented degree.789===== PAGE 1 (score 3, numeric-paren 0, table-words 0, var-words 3) =====1011RESEARCH ARTICLE12Spatial and Social Media Data Analytics of13Housing Prices in Shenzhen, China14Chao Wu1, Xinyue Ye5, Fu Ren1,2,3, You Wan1,2,3, Pengfei Ning1, Qingyun Du1,2,3,4*151 School of Resources and Environmental Science, Wuhan University, Wuhan, China, 2 Key Laboratory of16GIS, Ministry of Education, Wuhan University, Wuhan, China, 3 Key Laboratory of Digital Mapping and Land17Information Application Engineering, National Administration of Surveying, Mapping and Geoinformation,18Wuhan University, Wuhan, China, 4 Collaborative Innovation Center of Geospatial Technology, Wuhan19University, Wuhan, China, 5 Department of Geography, Kent State University, Kent, Ohio, United States of20America21* qydu@whu.edu.cn22Abstract23Housing is among the most pressing issues in urban China and has received considerable24scholarly attention. Researchers have primarily concentrated on identifying the factors that25influence residential property prices and how such mechanisms function. However, few26studies have examined the potential factors that influence housing prices from a big data27perspective. In this article, we use a big data perspective to determine the willingness of28buyers to pay for various factors. The opinions and geographical preferences of individuals29for places can be represented by visit frequencies given different motivations. Check-in30data from the social media platform Sina Visitor System is used in this article. Here, we use31kernel density estimation (KDE) to analyse the spatial patterns of check-in spots (or places32of interest, POIs) and employ the Getis-Ord G33i method to identify the hot spots for different34types of POIs in Shenzhen, China. New indexes are then proposed based on the hot-spot35results as measured by check-in data to analyse the effects of these locations on housing36prices. This modelling is performed using the hedonic price method (HPM) and the geo-37graphically weighted regression (GWR) method. The results show that the degree of clus-38tering of POIs has a significant influence on housing values. Meanwhile, the GWR method39has a better interpretive capacity than does the HPM because of the former method’s ability40to capture spatial heterogeneity. This article integrates big social media data to expand the41scope (new study content) and depth (study scale) of housing price research to an unprece-42dented degree.43Introduction44Residential property is a multidimensional and durable commodity, and its value is determined45by a combination of characteristics categorized as structural, locational and neighbourhood46attributes [1–3]. However, these attributes do not have individual market prices. Numerous47studies have explored the relationships between housing prices and specific attributes by48PLOS ONE | DOI:10.1371/journal.pone.016455349October 26, 2016501 / 1951a1111152OPEN ACCESS53Citation: Wu C, Ye X, Ren F, Wan Y, Ning P, Du Q54(2016) Spatial and Social Media Data Analytics of55Housing Prices in Shenzhen, China. PLoS ONE 1156(10): e0164553. doi:10.1371/journal.57pone.016455358Editor: Kim-Kwang Raymond Choo, University of59Texas at San Antonio, UNITED STATES60Received: May 26, 201661Accepted: September 8, 201662Published: October 26, 201663Copyright: © 2016 Wu et al. This is an open access64article distributed under the terms of the Creative65Commons Attribution License, which permits66unrestricted use, distribution, and reproduction in67any medium, provided the original author and68source are credited.69Data Availability Statement: POI data are from70http://open.weibo.com/. We used Open API71interface to get check-in data. The relevant data are72within the paper and its Supporting Information73files. Attributes of house data are from http://sz.74fang.com/. The relevant data are within the paper75and its Supporting Information files. Housing price76data are from Shenzhen Research Centre for Digital77City Engineering. Interested readers and78researchers can contact with Shenzhen79development research center using the80284158541@qq.com. Interested readers and81researchers can cooperate with Shenzhen828384===== PAGE 6 (score 29, numeric-paren 0, table-words 3, var-words 19) =====8586or fake users [45]. We filter the check-in data using the following critera (i) the location of a87check-in is not in Shenzhen based on the geographical location; (ii) the name of the POI is not88correct based on the text discriminant; (iii) we believe that users who have only one check-in89record are not valid users and that their behaviours are accidental in a given time period(July902014 and June 2015 in this article); (iiii) and based on the third criterion, we filter valid POIs91that are signed by at least one valid user. Finally, we acquire 447,778 check-in records from92216,165 users and 22,670 POI. To ensure the privacy of personally identifiable information93(PII) [46], we do not identify the information of users. This article uses 13,268 CBF sites and941,413 GRE sites identified by researching the locations and names of POIs based on the Code95for Classification of Urban Land Use and the Planning Standards of Development Land [47].96The CBF includes commercial facilities, business facilities, entertainment and sports facilities,97and public utility outlets, and the GRE consists primarily of parks, green belts and urban98squares. POI types and aggregated information can be seen in Table 2.99Methods100Spatial hot-spot analysis method based on check-in data101In this section, we present our proposed method for using check-in data to detect hot spots.102Each check-in record includes the user ID, time, coordinates (longitude and latitude), POI ID,103Table 1. Measurement methods and housing characteristic variable signs.104Variable105Variable definition and measurement methods106Mean107Std.dev.108Expected sign109Structural Variables110Area111Square footage of the living area of house (m2)112100.48711335.689114+115Floor116Floor on which the unit is situated (floor)11718.01811810.067119+120NumBed121Number of bedrooms in unit1223.1391230.965124+125NumWash126Number of bathrooms in unit1271.7251280.643129Unknown130Fee131Property management fees (RMB)1323.8681331.069134+135RPlot136Ratio of floor area1373.9761381.404139-140RGreen141Ratio of green space area (%)1420.3451430.0729144+145Locational Variables146DCBD147Distance to CBD (central business district; km)14823.33914910.208150-151DSub152Distance to nearest metro station (km)1530.3191540.409155-156DBus157Distance to nearest bus station (km)1580.0191590.023160-161DHospital162Distance to nearest hospital (km)1630.1221640.786165Unknown166DSnursery167Distance to nearest nursery school (km)1680.5971690.464170-171DSprimary172Distance to nearest primary school (km)1730.7561740.464175-176DSmiddle177Distance to nearest middle school (km)1780.1081790.640180-181DPark182Distance to nearest park(km)1830.1131840.609185-186Neighbourhood Variables187GiZScore_G188Degree of activity in green space1890.4381902.308191-192GiZScore_C193Degree of activity in commercial centre1941.1811953.229196-197doi:10.1371/journal.pone.0164553.t001198Table 2. POI types and aggregated information.199Type200Abbreviation201Counts202Percentage203Commercial and business facilities204CBF20513,26820659%207Industrial208IND2098902103.93%211Transport facilities212TRA2131,6962147.48%215Residence communities216RES2173,56921815.74%219Green space220GRE2211,4132226.23%223Administration and public services224APS2251,8342268.09%227Total22822,670229100%230doi:10.1371/journal.pone.0164553.t002231Analytics of Housing Prices232PLOS ONE | DOI:10.1371/journal.pone.0164553233October 26, 20162346 / 19235236237===== PAGE 7 (score 5, numeric-paren 0, table-words 0, var-words 5) =====238239and name and category of the venue. We use address-coding technology to locate records for240our database and detect hot spots using the number of check-ins, which are obtained from POI241data [48]. The POI data include the POI ID, coordinates, check-in totals, user counts and cate-242gories. We measure the hot spots by identifying their relationship with different types of POIs243and exploring the effects of CBF and GRE.244Before detecting the hot-spot status of the different types of POIs, we analyse the overall245spatial patterns. The kernel density estimation (KDE) is used to analyse the spatial distribution246of all check-in POI data. KDE attempts to produce a smooth density surface of spatial point247events in geographic space [49]. The general form of a kernel density estimator is given by248f ðsÞ ¼249Xn250i¼12511252ph2 k dis253h254255256ð1Þ257where f(s) is the KDE function at location s, h is the bandwidth of the KDE, dis, is the distance258from point i to location s, and the function k (called the kernel function) represents the spatial259weight function of the ratio between dis and the bandwidth h. Briefly, the results of the KDE260rely on kernel sizes (called the search radius) and grid sizes. In this analysis, first, individual261output cells were partitioned in 100m100m areas, which can provide greater precision in the262estimation most efficiently. Because too small of a grid size can increase the computational263cost, whereas too large of a size can result in important details being ignored. Second, we select264a search radius of 1000 m based on multiple testing rounds. Thresholds of less than 1000m,265such as 600m and 800m, make the density distribution focus on a limited extent of elemental266points, and the overall characteristic are not obvious. Conversely, thresholds of greater 1000m,267such as 1200m and 1500m, make the density distribution overly smooth, and local differences268cannot be well represented. We set 1000 m as the ideal threshold because it is the longest dis-269tance that remains comfortable to travel on foot (10–15 minutes) and obtains relatively good270results in this paper.271After kernel density analysis, we rank the POIs based on the counts of check-ins and users272for all check-in spots to detect the POIs that are popular. The power-law distribution patterns273(Fig 2) show that there are far more POIs with fewer check-ins and users than POIs with many274check-ins and users. The patterns are similar to those in a previous study [33]. The results of275Fig 2 suggest that POIs with high popularity, represented by the numbers of check-ins and276users, provide more attractions to residents and are better known to people. Differences in peo-277ples’ preferences and usages for POIs reveal that the formation of hot spots and cold spots for278POIs is reasonable. We apply the ‘average nearest neighbour’ index to judge whether the spatial279distribution of points is clustered or dispersed. This process is conducted for GRE and CBF280separately. The clustering patterns of GRE and CBF are used to determine whether they are hot281spots.282Because of the discrete distribution of points and simplification of the amount of big data283[50, 51], we use grids to replace points to analyse the spatial patterns [52]. First, we divide the284study area into 500 m x 500 m spaces and then apply the grid. Testing determined that 500 m285is the optimal choice because a grid size larger than 500 m tends to mask local differences and286because a grid size smaller than 500 m can exaggerate local characteristics. We then calculate287the total number of check-in records and the number of POIs for each grid. We explore the288possible hot spots using the grids and then use Getis-Ord G289i to identify the hot spots. The290Getis-Ord G291i statistical clustering method [53, 54] is used to detect significant clusters of high292value within a city according to locations that contain many people. Getis-Ord G293i is a local sta-294tistic used to assess each feature within the context of neighboring features. This article defines295a ‘hot spot’ as a place with a high degree of activity. The statistics are calculated as follows:296Analytics of Housing Prices297PLOS ONE | DOI:10.1371/journal.pone.0164553298October 26, 20162997 / 19300301302===== PAGE 13 (score 30, numeric-paren 1, table-words 8, var-words 9) =====303304with the findings of previous studies [58–60]. The structural attributes, such as the number of305washrooms, floor height, area, property management fees and green ratio, all have effects on306housing prices and add 9.4%, 6.3%, 29.3%, and 2.3%, respectively, of extra value to the price of307housing with a 1- unit increase. The effects of location variables are similar as expected with308the exception of distance to primary schools. A 1- km increase in the distance to parks, hospi-309tals, bus stations and subway stations reduces housing prices by 11.1%, 9.4%, 4.9% and 4.9%,310respectively. Primary schools are an exception because of the policy of school district housing311in China. The effect of school location on housing prices is determined by distance as well as312school quality.313Geographicallyweighted regression results. The results of the local regression analysis314(GWR) are summarized in Table 5. As previously mentioned, this method evaluates spatial315variations in the relationships between the variables and housing prices. In this model, the R2316value increases from 0.724 (HPM) to 0.9399 (GWR), thereby showing that the GWR method317has a better interpretive capability than the HPM. The GWR can explain 93.99% of the varia-318tion according to the R2 value. We also measure the spatial auto-correlation (Moran’s I319index = 0.041, z-score = 3.86) by calculating Moran’s I index for the residuals of the GWR to320ensure that the validity of the model for the residuals is random. Additionally, the AICc of the321GWR is -74703.46207, and the bandwidth equals 0.337, which are less than that of the HPM322(AICc = -36082.16912). The difference in the AICc of greater than 3 suggests that the GWR is323more representative of reality [56]. Table 5 shows that all the variables are significant at 1% or3245% with the exception of DPark (p-value = 0.855). Additionally, the outstanding achievement325of the GWR method is its ability to visualize local parameters. Fig 7 presents the GiZscore_C326and GiZscore_G. Because of the discrete distribution of points, we use the inverse distance327weighted (IDW) interpolation method to obtain a better visual effect. The output cell size of328the IDW processes is set to 100 m, and a natural breaks (Jenks) classification is applied. Con-329sidering that the coefficientshave positive and negative values, we set a zero value as a class330limit to distinguish the positive and negative effects.331Table 4. Hedonic price method (HPM) parameter estimate summary.332Explanatory variable333Estimated coefficient334t-ratio335p-value336Constant3370.06033819.388339.000***340DCBD341-0.326342-55.863343.000***344Fee3450.29334643.604347.000***348DPark349-0.111350-27.376351.000***352NumWash3530.09435421.743355.000***356GiZScore_C3570.08335817.104359.000***360DHospital361-0.094362-24.942363.000***364SPrimary3650.08036621.966367.000***368Floor3690.06337020.077371.000***372GiZScore_G3730.13437424.466375.000***376DBus377-0.049378-14.273379.000***380DSub381-0.049382-11.417383.000***384Area3850.0313866.907387.000***388RGreen3890.0233906.854391.000***392R23930.724394*** Significant at the 1% level.395doi:10.1371/journal.pone.0164553.t004396Analytics of Housing Prices397PLOS ONE | DOI:10.1371/journal.pone.0164553398October 26, 201639913 / 19400401402===== PAGE 14 (score 25, numeric-paren 1, table-words 3, var-words 9) =====403404The regression coefficientfor GiZscore_C is greater than zero because major concentrations405are observedin Futian, Longhua, central and northeast Nanshan, the southern part of Bao’an406and Guangming and central Dapeng. The CBF is more strongly clustered, which is associated407with more check-ins, and the housing prices are higher in these locations, which represent the408most prosperous areas of each district and have more complete facilities and more workplaces.409These areas provide convenient amenities to housing buyers, such as employment, cafes, shop-410ping and entertainment. Therefore, the housing prices are high. However, Fig 7(a) presents411unexpectedresults for Longgang, Luohu and Yantian, with lower housing prices despite the412greater clustering (with more check-ins) of the commercial and business facilities. GiZscore_C413Table 5. Geographically weighted regression (GWR) parameter estimate summary.414Explanatory Variable415Minimum416Lower quartile417Median418Upper quartile419Maximum420p-value421Constant422-12.8186423-1.38137424-0.26754250.0461442610.28509427.000***428Floor429-0.352914300.0312464310.0401694320.0566284330.339561434.000***435NumBed436-0.66773437-0.021494380.0408874390.1280114400.493299441.000***442NumWash443-0.64065444-0.04324450.0225234460.068474470.358014448.000***449Area450-0.68338451-0.017174520.0743584530.1426334541.335995455.000***456RPlot457-10.9418458-0.5452459-0.10744600.1456954611.900368462.000***463RGreen464-3.84742465-0.03124660.0897824670.3996574681.858314469.000***470Fee471-2.681124720.0159454730.254594740.6215514754.052185476.000***477DSub478-12.302479-0.40319480-0.020854810.56317448215.22164483.000***484DBus485-4.17797486-0.35394487-0.030284880.50998348923.5197490.004**491DSnursery492-18.9376493-0.1124494-0.032784950.1581634964.604339497.000***498DSprimary499-12.4242500-0.132585010.0176525020.2611095032.282185504.000***505DSmiddle506-3.55686507-0.410955080.0173845090.24963551014.10076511.000***512DHospital513-4.868514-0.62254515-0.128275160.3006555175.447187518.000***519DPark520-2.27624521-0.073675220.0129925230.32867352411.93626525.855526DCBD527-16.3229528-0.74499529-0.097055300.4309075311.94271532.000***533GiZScore_G534-11.9198535-2.41877536-0.12525370.1793575384.508242539.000***540GiZScore_C541-9.165235420.0155675430.4633025442.22986454511.68491546.000***547R25480.9399549Bandwidth5500.337551***, ** Significant at the 1% and 5% levels, respectively.552doi:10.1371/journal.pone.0164553.t005553Fig 7. Spatial variation of the local parameters of the GWR. (a) and (b) are GiZscore_C and GiZscore_G, respectively.554doi:10.1371/journal.pone.0164553.g007555Analytics of Housing Prices556PLOS ONE | DOI:10.1371/journal.pone.0164553557October 26, 201655814 / 19559560561===== PAGE 15 (score 12, numeric-paren 0, table-words 2, var-words 8) =====562563has a negative effect on housing prices, and because of the development of Futian and Nan-564shan, the position of Luohu has been declining. Luohu is an old urban district that is currently565facing the problem of an old-age demographic bump. Senior citizens tend to choose housing in566locations with a good environment characterized by peace and quiet. The clustering of resi-567dents can cause noise and traffic jams, which may affect the residents’ daily lives and rest. Yan-568tian has relatively poor traffic conditions and no subway system. Thus, the clustering of CBF in569Yantian may result in worse traffic conditions. Similarly, GiZscore_C has a negative effect on570housing prices in Yantian. Longgang is a relatively poorly developed area in Shenzhen and has571comparatively low housing prices. Housing buyers in Longgang pay more attention to trans-572portation and the convenience of getting to work. Thus, it is not difficultto imagine that GiZs-573core_C would have a negative effect on housing prices in Longgang.574The GWR model reveals that the relationship between GiZscore_G and housing prices is575not stationary across the study area, as illustrated in Fig 7(b). Most major areas show a positive576effect of GiZscore_C on housing prices, with the exception of the south of Bao’an, the west577and centre of Longgang, and most of Yantian. Yantian was the first ‘national biome’ in South578China, and the air quality is good, which attracts many people. However, Yantian’s traffic con-579ditions are relative poor, and it is located in a fault zone in Shenzhen. Therefore, it is likely that580people prefer to visit this scenic spot for recreational purposes rather than live there.581Conclusions582This article suggests a framework for applying a GWR model and a big data perspective for583evaluating the influence of urban hot spots on housing prices in Shenzhen, China. This work584contributes to the field of housing price research by identifying POI hot spots through the anal-585ysis of check-in data and by exploring the influence of a POI’s degree of activity (from the586perspective of popular perception) on housing prices. Despite the widespread use of crowd-587sourced geographic data (CSGD) [49, 61–64], few studies have used these data to examine588housing prices, which are closely tied to public opinion and sentiments. We apply GWR589method to test the hypothesis that hot spots (identified using check-in data) have significant590effects on housing values. Our results are consistent with expectations. In this article, we intro-591duce the use of check-in data, which represent a novel data source in the field of housing price592research. Additionally, we take spatial heterogeneity into consideration and explore the factors593that influence housing prices for different locations. Finally, we chose Shenzhen, China as our594case study because housing prices in Shenzhen have skyrocketed in recent years, which is an595area of continued concern [61]. This article integrates big social media data and expands the596scope (new research content) and depth (study scale) of housing price research to an unprece-597dented degree.598This article applies the Getis-Ord G599i technique to detect hot-spot areas of CBF and GRE.600We then explore the relationship between housing prices and hot spots using hedonic price601modelling and geographical regression modelling. The R2 of the HPM is 0.724, which indicated602that the HPM can be used to analyse the effects of hot spots from a global perspective. CBF and603GRE hot spots both increase housing prices. The R2 of the GWR is 0.9399, which indicates a604stronger interpretive capability compared with the HPM. The results of the GWR reveal that605the effects of influential factors vary over space, as represented by different coefficients. Finally,606we select typical and special areas for our interpretation.607Most of the data values require innovative methods of analysis to identify their potential608uses [65]. Although this article provides a new perspective for the application of big social609media data to the study of housing prices, it still has several limitations.610Analytics of Housing Prices611PLOS ONE | DOI:10.1371/journal.pone.0164553612October 26, 201661315 / 19614