A New Method for Analyzing the Relationship between City and Human Behavior using Geo-Tagging Social Networking Service

Please download to get full document.

View again

of 6
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Document Description
Long-term studies on human distribution have been found numerous mathematical models on human distribution to analyze the tendency where to go. The premises of these models have been parameters on limited human parameters and weighing space
Document Share
Document Tags
Document Transcript
  1 A New Method for Analyzing the Relationship between City and Human Behaviorusing Geo-Tagging Social Networking Service Kousuke Kikuchi* 1 , Hiromu Okutsu 2 , Atsushi Enta 3 and Hitoshi Watanabe 4 1 Doctoral Student, Graduate School of Creative Science and Engineering, Waseda University, Japan 2 Master Student, Graduate School of Creative Science and Engineering, Waseda University, Japan 3 Asistant Professor, Faculty of Science and Engineering, Tokyo University of Science, Japan 4 Professor, Faculty of Science and Engineering, Waseda University, Japan Abstract Long-term studies on human distribution have been found numerous mathematical models on humandistribution to analyze the tendency where to go. The premises of these models have been parameters onlimited human parameters and weighing space parameters. However, recent information terminal changespotential human preference into actual and human behavior into purpose-oriented. In summary, researchershould think about new method of sublating human quality and quantity. In this thesis, the authors propose touse Geo-Tagging Social Networking Service to get information on Five Ws about human behavior. The mainpurpose of this thesis is establishing the methodology on the spatio-temporal distribution of human attribute’sand human activity. Secondary purpose is indicating that city is composed from human activity and attribute. Keywords : Geo-Tagging SNS; Human Behavior; City’s Function; Human Attribute; Urban Tribe*Contact Author: Kousuke Kikuchi, Doctoral Student, GraduateSchool of Creative Science and Engineering, Waseda University,3-4-1 Okubo, Shinjuku-ku, Tokyo, 169-8555 JapanTel: +81-3-5286-3276 Fax: +81-3-5286-3276E-mail: kousukekikuchi@toki.waseda.jp (The publisher will insert here: received, accepted ) 1. Introduction1.1 Research Method on Human Distribution Over the past few decades, the methods on HumanDistribution in city have been studied. To fathomhuman distribution, Nakamura et al. (1971) proposedtransition probability model 1) whose assumption is thathuman transition is evenly divided form a space toanother. And, Watanabe (1975) added human statusesinto the Nakamura’s human transition probabilitymodel using automaton model 2) . And in the UnitedKingdom, Hiller (1976) introduced space syntax 3) 4) toevaluate the ease of mobility clarifying phase structureof its city mainly based on graph theory. These studieshave aimed at the evaluation of Urban Planning orarchitectural planning in advance. However, thelimited human factors in these methods are nowadaystroubling to detect human distribution because of therecent change on human factors. 1.2 Recent Change of Human Factor The rise of information terminals is thought to bechange human behavior into purpose-oriented and theattribute of human into more versatile. The informationabout commodities, events, or hot spot can be foundmore easily than ever before with the becoming thesearching in web. In consequence, people are thoughtto change to achieve the purpose efficiently. Inmarketing, paradigm shift toward purchase model hasoccurred 5) to adapt to the information society.Also, Information has become human into moreversatile. People have several preferences latently.However, information society has formed Communityof Interest, people with same preference gather incyber space. These cohorts become significant forboth users and marketers because this concept changeshuman’s potential preference into actual 6) . 1.3 Research Methods on a City Researches on city utilizing human versatilenessare limited because the researchers force to choosethe quality of data. For example, Hayden(1995)actualized Genius Loci by forming the endemicimage of inhabitants using oral history 7) . Also, captionevaluation method is developed to understand thediversity of people’s image about townscape at theworkshop 8)9) . These attempts can be evaluated tocomplete the image of dwellers in its city using huge UIA2011 TOKYO Academic Program -388- 10074  2 human resources. However, participants of interview orworkshop are extremely partial and arbitrary comparedto the all stakeholder in its city. In Addition, suchsituation should be avoided because the queer situationat interview or workshop makes these participatorsprone to behave extraordinarily. Therefore, the methodfor all stakeholders to participate naturally is stronglyrequired. 1.4 Proposed Method The selection between quality and quantity is thelimit of analog. If we use digital media, we can sublatethe human quality and quantity. For example, digitalethnography 10) is recently proposed as the methodto analyze human being utilizing web data such ashomepage, BBS, blogs, SNS. The relationship betweenhuman and city should be analyzed using digital media.The authors propose to use geo-tagging socialnetworking service. Especially, Twitter(Fig.1) isthought to fit analysis on city because almost itscontribution are related to daily life 11) . People cancontribute with geo-information. The accumulation of human behavior will make the locality of area clearlyexist. The authors define the accumulation of humanbehavior as Urban Tribe because these behaviors’distributions are thought to be observed as tribe atmacroscopic view. 1.5 Purpose of This Thesis Primary purpose of this thesis is establishing a methodon the spatio-temporal distribution of human attribute’sand human activity and explicating its character andlimitation. Secondary purpose is suggesting a cityconsist of some Urban Tribes. Fig.2. shows the schemaof this thesis. 2. Method2.1 Data Collection from Twitter The summary of the data collection procedures isshown in Fig.3. The authors use two programs toacquire data by Twitter API 12) . First, the authors madea list of users who post in Tokyo using geotag of theirtweet. Second, the authors received all tweets on everyperson, from the user list. The authors get username,tweet-time, tweet, latitude, longitude. The nineprograms to acquire the users’ data have been operatedbecause Twitter API permits only 350 accesses perone hour. We have had 2,630 users and 199,546 tweetsfrom July 19 to December 5 in 2010. 2.2 Estimation of Human Attribute and Action The authors assumed the human behavior could beobserved "Verb" and "Object" in Twitter. According toNguyen et al.(2010), human behavior in Twitter canbe observed that “Subject”, “Verb”, “Object” 13) . Theyattempted concreting the database of human behaviorby detecting “Verb” and “Object” because almosttweets’ subjects are contributor self. Based on thispaper, we attempt the “Verb” as human activity and“Object” as human attribute in this thesis. 2.3 Analysis on Activity Because of enough data to analysis on certain area,the locations are limited for detailed analyses inRoppongi, Shibuya, Shinjuku, Ginza and Akihabarabecause of intensive tweets areas(Fig.4). Next,three-dimensional kernel density estimation isused to quantitatively estimate the spatio-temporaldistributions of activities in each area. Also, volume A   c  c  u  m  u  l   a  t   i   o  n   E  x  t  r  a c  t  i  o n  ActivityAttributeActivityAttribute ・ text ・ time ・ geotag ・ user A n a l  y s i s  time     U   r    b   a   n     T   r    i    b  e Fig.2. Schema of this thesis Fig.1. Twitter Kousuke Kikuchi -389-  3 rendering is utilized to visualize four-dimensional datain two dimension display. Brundson et al. (2007), byaugmenting the kernel density estimation into three-dimensional, have clarified the isosurface methodto analyze the distribution of crimes, which canvisualize the plane of consequent scalar in space andtime coordinates 14) . Based on Brundson, Nakaya et al.(2008) have added volume rendering, the method of visualizing on shading and transparency, and concretedmore understandably 15) . Fig.5 shows the used functionand its parameters.The following describes the method adapted at thisthesis. First, the six keywords related to daily activitywere assumed. The six keywords were eat, shopping,look, photo, rest and wait, and the similar words tokeywords were included to the keywords. Second, alldata included certain keyword had been extracted andwritten into a file by Excel VBA. Third, the Authorscalculated the evaluation by three-dimensional kerneldensity estimation, and outputted a CSV format fileby Excel VBA. Finally, Voxler had read the calculatedfile, and visualized the spatio-temporal distributionsof activity in area. Needless to say, the other keywordscan be extracted likewise because the keywords existnumerously. 2.4 Analysis on Attribute The space distribution could be analyzed only bykernel density estimation because of less data toanalyze on certain city and space-time coordinates.Following describes the method adapted at this thesis.The four keywords that are thought to be related toboth human preference and urban factor of humanbehavior, are extracted; Running, Walking, Strolling,Cycling. Next, all data, included certain keyword, has been extracted and written into CSV format files, by Excel VBA. Next, these data have been read on Fig.3. Summary of data collection Fig.4. All geo-tagged tweets in Tokyo Fig.5. The functions and parameters used in this thesis Kousuke Kikuchi the ArcGIS to calculate the kernel density estimationand write KML format files. Needless to say, theother keywords can be extracted likewise because thekeywords exist numerously. 3. Results and Discussions3.1 Evaluation on Activity In this thesis, the authors visualized the distribution of the six areas and the six keywords with daily routine.The following describes commonly the observed or thecharacteristic result.First, the locality of certain area can be shown tocompare the neighboring city or adjacent area bythese distributions. Fig.6 is shown the distributionof Shibuya. The neighboring cities of Shibuya areHarajuku, Omotesando and Ebisu. Although almostall distributions are extent in Shibuya, other citieshave several characters. Ebisu distribution is mainly -390-  4 consisting of “Eat” and Omotesando distribution isconsisting of “Rest” and “Shopping.” Fig.7 is shownthe distribution of Akihabara. Same tendency is shownin this distribution. Fig.8 is shown the distributionof Ikebukuro. In Ikebukuro, although there is noneighboring city, the distribution of activities isdifferent in east and west of Ikebukuro station. Theseexamples can infer that the comparison betweenother cities or inner areas can be the way to detect thelocality.Second, by spatio-temporal distributions, transition Kousuke Kikuchi of people can be detected. Fig.9 is shown the “Picture” distribution of Ginza. This figure shows that the density transition starts from Yurakucho at 14:30 to Ginza andends from Ginza at 20:00 to Yurakucho at 22:30.Third, except Roppongi, “Picture” distribution isthought to be related to “Shopping” distribution.Fig.10, 11 are shown the “Picture” and "Shopping"distribution of Shinjuku and Roppongi. Fig.10 wasshown that the “Picture” distribution includes the“Shopping” distribution. However in Roppongi“Shopping” distribution is extent in “Picture” Fig.6. Distribution in Shibuya Fig.7. Distribution in Akihabara Fig.8. Distribution in Ikebukuro Fig.9. Spatio-temporal distribution of "Photo" in Ginza -391-  5 Kousuke Kikuchi distribution. Although complexes buildings and artmuseums are thought to cause this area into low-density, this thesis did not clarify the reason of theexception.In conclusion, the possibility of detecting the localitycan be inferred. Also, the spatio-temporal transitionof human could be visualized. And, the existenceof Urban Tribe is indicated because of the differentactivities between each area. 3.2 Evaluation on Attribute The evaluation was visualized by kernel densityestimation because little data for analyzing on eacharea. Following and Fig.12 are shown the resultdistribution of four keywords.These four keywords are related to the locomotionof people. These places shown in the Figure.12are assumed to rest because posting on moving isimpossible. The place overlapped several attributes isAsakusa, Meguro and the environs of Imperial Palace.And the places with high density of these attributes areDen-en-Chofu and Komazawa Park. There areas arethought to be good at resting.But, for little data of analyzing the distribution of human attribute, the crude place to rest only can be extracted. Further verification on this analysis should be done. 3.3 Discussion on This Methodology This methodology still remains strong arbitrariness.The selections of words determined to the keywordusing human power are not preferable to analysisand not suitable for analyzing huge data. Also, The  presumption that human attribute can be defined as the “Object” is not strictly preferable. The situation can bethought that the contributor posts without the keywordsthough he think thing related to the keywords. Themethod on analyzing human self automatically isstrongly required. Indeed, Natural language processingshould be utilized to concrete the mathematical model Fig.10. Distribution of "Photo" and "Shopping" in Shinjuku Fig.11. Distribution of "Photo" and "Shopping" in Roppongi Fig.12. Distribution of four attributes -392-
Similar documents
View more...
Search Related
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks