using the People Flow Data

Size: px

Start display at page:

Download "using the People Flow Data"

Colin Burke
5 years ago
Views:

1 Analysis of Trip Chains Shapes and Regional Differences using the People Flow Data Hiroshi Ohba *, Tatsuya Kishimoto ** Abstract: In this research, we focused on the shapes of peoples trip chain in a city and regional area. We analyzed the shapes of trip chains of the people flow data of Matsuyama city region, and investigated the geometrical characteristics. We proposed 15 indices for analyzing shape of trip chain. The indices comprise maximam width of trip chain, convex hull area of trip chain, distance between centroid of trip chain and home, number of destination, calorie consumption, travelling time, CO2 emission, and etc. Totally about 20,000 shapes of trip chains are quantitatively analyzed by the indices. By PCA (principal component analysis), five principal conponents of shapes of trip chains are derived. The first principal component explains the character of long distance and wide range trip in a day. The second component explains the concentration at main destination. This component shows the characteristic that he or she stays very long at one destination. The third component explains the concentration in residence area, which shows that he or she stays at his or her resident area very long. The fourth component explains the wandering character of the trip chain at local area, and the higher value means that he or she trips circularly around his or her resident area. The last component explains healthy moving character of the trip, and the higher value means that he or she moves by non-motorized mode like bicycle or walking. We classified 20,000 trip chains of a day into five classes by cluster analysis method. The results showed that the shape of trip chain have some relation with occupations and ages. Furthermore, we calculated the average values of principal component scores in every elementary school zone, and analyzed the school zones with respect to the locality of each zone. Keywords: Person Flow Data, Trip Chain, Principal Component Analysis, Cluster Analysis ** Keio University Graduate School of Science and Technology, School of Science for Open and Environmental Systems h.ohba@a8.keio.jp ** Keio University Graduate School of Science and Technology, School of Science for Open and Environmental Systems kishimoto@sd.keio.ac.jp - 1 -

2 1. Introduction In the past, any ways of analyzing or assaying cities or lives of people have been proposed, but most of these based on the aggregate data like a population density, a use of land or a spot traffic volume. However, in the urban cities, peoples activities have a lot of variations, and for more detailed analyzing or assaying, we need to grasp the characters of urban cities and people living in there based on the information about personal activities. Because of these, in this paper, we built two hypotheses that we can grasp more clearly complicated peoples' activities by typification of them and we can grasp complex urban cities from the side of peoples activities. And from these hypotheses, we purposed to analyze people moving behaviors and regional differences by the geometrical characteristics of paths obtained from peoples moving behaviors: Trip Chains. 2. Data of peoples moving paths 2.1. Person trip survey Person trip survey investigates each target person s moving behavior for a one-day in target date and inside of a fixed research zone. This data includes gender, age, occupation and in addition to this, an origin, a destination, a purpose of moving, a transportation mode and a time length required for a destination are included. For examples of researches using data of Person trip survey, Osaragi et al. [1][2] built a space-time distribution presumption model of automobile user and train user for urban disaster prevention plans by using the person trip data and the origin-destination data. And Akiyama et al. [3] built a destination choice model considered a subjective ambiguity of moving person by the fuzzy inference. Besides these, Shimoda et al. [4] revealed the relation between the choice of transportation got from the person trip data and data of urban cities like a population-did area ratio or a population, and Taniguchi et al. [5], by using the transportation choice data and the moving distance data got from the person trip data, calculated the gasoline consumption and by comparing it and number of stations, population densities or each city, he revealed the their relations and aging People Flow Project and People Flow Data Person trip survey data have information about a destination, a departure time, an arrival time, a transportation mode and etc, but these are fragmentary information on temporal and spatial and these are not continuous information like a real peoples moving behavior. People Flow Project of Center for Spatial Information Science The University of Tokyo (CSIS) provides Spatio-temporal Interpolation Service that complements the fragmentary information about moving behavior to a minute-by-minute spatio-temporal interpolation data. And by using this service, we can get the data which minute-by-minute spatio-temporal interpolated based on Person Trip Data: People Flow Data. Figure 1 is an instance of drawing a path of a person s data got from People Flow Data. The path starts from the point no.1 and ended at the point no.9, and this locus means a one-day continuous chain of his moving behavior: Trip Chain

3 Figure 1. Example of drawing a path from People Flow Data For examples of researches using data of People Flow Data, Shimazaki et al. [6] analized the correlation between the number of the population changed by time and the number of shops considered aging the number of shops, and Suzuki et al. [7] surveyed a start point and an end point each shopping trip, and calculated choice probabilities of shopping areas of each residence by the Huff Model, besides these, they examined a factor of the difference between the profit estimated and the actual by comparing the distance resistance parameter each area and the aging of these. Moreover, Asahara et al. [8] purpsed to grasp efficiently traffic conditions in urban cities and proposed the way of extracting the main traffic line that most moving people are using by People Flow Data of Tokyo urban area and Mixed Semi Markov-chain Model. The number of researches that are using People Flow Data is less than using Person trip survey Data, but we think People Flow Data is suitable for grasping the person s moving behavior of one-day, because it has continuous data of peoples moving behaviors. Due to this reason, in this research, we used People Flow Data of Matsuyama (a capital of Ehime prefecture in Japan) urban area in This data includes about 20,000 of peoples one-day trip data investigated on whole regions of Matsuyama City and surrounding areas

3. Indexing of moving behaviors We can get information from peoples moving behaviors which has not only the algebraical factor: distance and time but also the geometrical factor: shape, area and

First, Max Width is calculated by summing the length of two (or one) maximum length perpendiculars from the line (reference line) which is connecting two arbitrary points to other points (Figure 2).

in along his commuting route. In brief, we can infer the tripper s range of dropping in based on the reference line if we calculate Max Width. Figure 2.

4 3. Indexing of moving behaviors We can get information from peoples moving behaviors which has not only the algebraical factor: distance and time but also the geometrical factor: shape, area and centroid. In order to measure these factors, we proposed Trip Indices below. The base concepts for deciding Trip Indices are trip chain s Max Width, Area of Convex hull and Centroid. First, Max Width is calculated by summing the length of two (or one) maximum length perpendiculars from the line (reference line) which is connecting two arbitrary points to other points (Figure 2). For example, when we consider the reference line as the line connecting residence and workplace, if there are three visited points and Max Width is nearly 0, we can conjecture that the tripper drops in along his commuting route. In brief, we can infer the tripper s range of dropping in based on the reference line if we calculate Max Width. Figure 2. Diagram of Max Width Second, Area of Convex hull means the size of a convex hull generated by all of visited points. We think that we can explain the tripper s range of behavior regardless of the order of visiting or the frequency of visiting by this value (Figure 3). Figure 3. Diagram of convex hull Finally, Centroid is the center of gravity weighted by the duration of staying on each point. We think that we can measure the separation from center of tripper s life like his residence by calculating the distance between his residence and Centroid (Figure 4)

Figure 4. Diagram of Centroid Basing on these concepts, we propose 15

Indices of Main Destination When we set the place which has the longest duration of staying (ex: workplace) as Main Destination,

based on Main Destination Line as Max Width on the basis of Main Destination Line and the duration of staying there as Staying

Indices of Farthest Destination When we set the place which is the farthest place from the residence as Farthest Destination, we

5 Figure 4. Diagram of Centroid Basing on these concepts, we propose 15 Trip Indices below Indices of Main Destination When we set the place which has the longest duration of staying (ex: workplace) as Main Destination, we set the length between the residence and Main Destination (Main Destination Line) as Distance of Main Destination, Max Width based on Main Destination Line as Max Width on the basis of Main Destination Line and the duration of staying there as Staying Duration at Main Destination, and calculated these. Figure 5. Indices from Main Destination 3.2. Indices of Farthest Destination When we set the place which is the farthest place from the residence as Farthest Destination, we set the length between the residence and Farthest Destination (Farthest Destination Line) as Distance of Farthest Destination, Max Width based on Farthest Destination Line as Max Width on the basis of Farthest Destination Line and the duration of staying there as Staying Duration at Farthest Destination, and calculated these

Figure 6. Indices of Farthest Destination 3.

such as Area of Convex hull: this explains the tripper s range

tripper s moving, Number of Destination: this is the total

of moving is changed, and Centroid Distance: this is the

6 Figure 6. Indices of Farthest Destination 3.3. Indices of entire moving behavior We calculated indices such as Area of Convex hull: this explains the tripper s range of behavior, Total Trip Distance: this is the total length of tripper s moving, Number of Destination: this is the total number of places where the transfer is changed or the purpose of moving is changed, and Centroid Distance: this is the distance between Centroid and the residence. Figure 7. Indices of entire moving behavior - 6 -

7 3.4. Indices of time and activity We calculated indices such as Duration from Trip Start to Trip End: this is hours from the start of the trip to the end of the trip, Staying Duration at the residence: the total hours that tripper is in the residence, Total Moving Duration: this is the total hours of moving from a place to another place, Calorie Consumption: this is calculated from the time of moving by bicycle or on foot, CO 2 Emission: this is calculated from the distance of moving by cars or trains. Table 1. Trip Indices Ti1 Distance of Main Destination [m] Indices of Main Destination Ti2 Max Width on the basis of Main Destination Line [m] Ti3 Staying Duration at Main Destination [min.] Ti4 Distance of Farthest Destination [m] Indices of Farthest Destination Ti5 Max Width on the basis of Farthest Destination Line [m] Ti6 Staying Duration at Farthest Destination [min.] Ti7 Area of Convex hull [m 2 ] Indices of entire moving behavior Ti8 Total Trip Distance [m] Ti9 Number of Destination Ti10 Centroid Distance [m] Ti11 Duration from Trip Start to Trip End [min.] Indices of time Ti12 Staying Duration at the residence [min.] Ti13 Total Moving Duration [min.] Indices of activity Ti14 Calorie Consumption [kcal] Ti15 CO 2 Emission [g] - 7 -

4. Typification of People s Moving Behavior using Trip Indices We integrated 15 Trip Indices into a less number of principal components (PCs) by Principal Component Analysis (PCA).

Because of this, we think we can grasp peoples moving behavior from a less parameter extracted by PCA that can integrate some parameter correlated and extract some PCs that have higher

8 4. Typification of People s Moving Behavior using Trip Indices We integrated 15 Trip Indices into a less number of principal components (PCs) by Principal Component Analysis (PCA). Most Trip Indices are not independent, for example, if Ti4 and Ti5 are increased, Ti7 is increased. Because of this, we think we can grasp peoples moving behavior from a less parameter extracted by PCA that can integrate some parameter correlated and extract some PCs that have higher interpretability. And next, we classified all the people based on PC values by TwoStep Cluster Analysis mounted on IBM SPSS Statics 20. Figure 8. Analyzing flow: typification of peoples moving behavior 4.1. Results of the principal component analysis We got the component matrix and the cumulative contribution ratio by PCA, and we interpreted each PC below (Table 2). Table 2. Results of the principal component analysis - 8 -

Reasons of these interpretations for each PC s are as follows. 1 st PC: Ti8, Ti4, Ti1, Ti5 and Ti2 have a positive weight, we can read it tending to move around wide range and long distance from this.

Ti8, negative Ti3 and negative Ti6 mean a tripper spends much time for moving. We can interpret that these tendencies appear from a long distance and a wide range moving behavior.

9 Reasons of these interpretations for each PC s are as follows. 1 st PC: Ti8, Ti4, Ti1, Ti5 and Ti2 have a positive weight, we can read it tending to move around wide range and long distance from this. In the same way, positive Ti10 and negative Ti12 mean a tripper is active mainly far from his residence; positive Ti15 means a tripper uses some transportations have environmental impact; positive Ti8, negative Ti3 and negative Ti6 mean a tripper spends much time for moving. We can interpret that these tendencies appear from a long distance and a wide range moving behavior. 2 nd PC: positive Ti3, positive Ti6 and negative Ti9 mean a tripper is tending to stay in the one place for a long time. 3 rd PC: positive Ti12, negative Ti11 and negative Ti9 mean a tripper stays in his residence without going out or goes out inside the zone of his residence for a short time. For example of a person who has a lifestyle like this, we can image a housewife or a househusband who goes out for only housework like commodity shopping and stays inside his residence for most of a day. 4 th PC: Ti7, Ti5 and Ti2 are positive value, but Ti10, Ti1 and Ti4 are negative. These mean a tripper moves around wide range and also moves to every direction based on his residence like Figure 9. 5 th PC: Ti14 is especially large positive value and this means a tendency that tripper moves by healthy ways like walking or a bicycle. Figure 9. Diagram of every-direction moving 4.2. Results of the cluster analysis We got principal component scores from the principal component analysis, and by using these scores and the cluster analysis, we classified all the people and got the results below. Table 3. Statistics of the principal component scores for each cluster - 9 -

10 Cluster-A has a very large median value of 1 st PC: +4.1, and if we consider its standard deviation, we can read that most of people belonging to Cluster-A have a positive 1 st PC value. Because of this, we interpreted Cluster-A as Type of Long Distance and Long Hours Moving. Next, Cluster-B and Cluster-C both have a larger positive 2 nd PC and a negative 3 rd PC, but in contrast, 5 th PC has a bigger margin between B and C. Because of these, we read both clusters as: they have a characteristic that belonging person has action at his main destination specializing, and also Cluster-B as: belonging person uses any transportation, Cluster-C as: he moves by healthful way like on foot or by bicycle. From these results, we interpreted Cluster-B as Type of Concentrating in Main Destination with CO2 emission and Cluster-C as Type of Concentrating in Main Destination with healthy moving. In Cluster-D, From negative 1 st PC means that the trip distance is less and both 2 nd and 3 rd PC are positive, we can read that belonging person s main destination is same as his residence and he has an action at his residence specializing. Moreover, from positive 4 th PC, we can read that the person has action near by his residence mainly. Because of these, we interpreted Cluster-D as Type of Concentrating in Residence Area. Finally, from excepting 3 rd PC of Cluster-E the all of PCs have negative value, we read that almost all the belonging people don t take a moving behavior, therefore we interpreted Cluster-E as Type of Short Time Tripping Comparing with individual attributes We compared the characteristic of each cluster by the belonging peoples individual attributes, and results from this analysis are below. From Figure 10 and Figure 11, Cluster-A includes almost no trippers that are less than 20-year-old or more than 80-year-old, and includes nobody of students(primary, junior high school)/children/kindergartners or students(high school, university). We thought reasons about these: the possibility that the aged don t/can t go out is high, and the data used for analysis that investigated on a weekday, and consequently most students can t go out a long distance. About Cluster-B, we read that its number ratio is concentrated on other worker and students/children/kindergartners from Figure 11 and Figure 12. In case of other worker, we thought the reason why the possibility of this is high: it is caused by a workers commuting behavior. And in case of students/children/kindergartners, we read the possibility of this from Figure 13 that it includes the belonging trippers who go to school by car like a pickup bus or by transportation. The reason why Cluster-C has a characteristic of healthy moving that it includes the people categorized as 15 to <20-year-old and students by the higher ratio (Figure 10-Figure 12), moreover we read from Figure 14 that students(high school, university) go to school on foot or by bicycle mainly. About Cluster-D, this is the reason why it has a characteristic of concentrating in residence area: from Figure 11, this cluster has a higher number ratio of employee of agricultural forestry industries and fisheries and students/children/kindergartners. Because of this, in case of the former there is possibility their workplace close to their house, and in case of the latter there is higher possibility that students go to school inside each elementary school zone. Cluster-E has a higher number ratio of no-occupied. In case of housewife/house-husband, most of them spend their time for housework in their own house. Accordingly, it is appropriate that they have the characteristic of this cluster. Moreover in case of employees, we can

11 understand that especially a part-timer has a similar characteristic. As stated above, we confirmed each cluster has a characteristic explained by age or occupation and these characteristics agree with our general sense of lifestyle. Figure 10. The cluster number ratio for every age group Figure 11. Thu cluster number ratio for every occupation group Figure 12. The ocupation number ratio for every cluster Figure 13. Transportation used by "students/children/kindergartners" Figure 14. Transportation used by "students"

5. Analysis of Characteristics of Areas using Trip Indices We classified areas by the cluster analysis based on the principal component scores averaged by trippers own residence areas, and we

12 5. Analysis of Characteristics of Areas using Trip Indices We classified areas by the cluster analysis based on the principal component scores averaged by trippers own residence areas, and we examined what kind of characteristics does each type have by Trip Indices and special features of each area. In this analysis, we adopted the Data of elementary school zone (2010) from Ministry of Land, Infrastructure, Transport and Tourism as the analysis unit of areas. The reason is that the unit of People Flow Data is the town and area level and it s too small for analyzing. The outline of analysis is similar to the analysis of individual, but we restricted targets that have more than 20 residents in the data because of avoiding singular values. Moreover, in the cluster analysis, we added the value of distance between the residence and Matsuyama City Station (terminal station of Matsuyama City). And we got two clusters by analyzing all of the areas, therefore we re-analyzed for each cluster for grasping the difference of areas in detail. Figure 15. Analyzing flow: typification of areas 5.1. Results of the analysis The result of analyzing all of the areas is Figure 16 and the result of re-analyzing the two clusters is Figure 17. And the statistics table is Table

Figure 16. Analysis of all areas Figure 17. Analysis of the two clusters Table 4.

And from the 2 nd cluster analysis, we can understand that the urban area is distributed two clusters clearly, but in the suburban area, some minority cluster(area-cluster-d) spreads in

Typification and characteristics of areas From these results, we examined a characteristic of each cluster: a characteristic of each area by the principal component score and Trip Indices

13 Figure 16. Analysis of all areas Figure 17. Analysis of the two clusters Table 4. Statistics of the principal component scores for each area cluster We can understand from Figure 16 that two clusters are distributed to an urban area and a suburban area. And from the 2 nd cluster analysis, we can understand that the urban area is distributed two clusters clearly, but in the suburban area, some minority cluster(area-cluster-d) spreads in Area-Cluster-C Typification and characteristics of areas From these results, we examined a characteristic of each cluster: a characteristic of each area by the principal component score and Trip Indices averaged by each area. In Area-Cluster-A, every PCs have a value nearly zero, but in these, we can read that the 1 st PC: this has the highest contribution ratio in PCs has a higher value next to Area-Cluster-C, and the 2 nd PC has a lower value next to Area-Cluster-B. We can read same results from averaged Trip Indices: from Figure 18 and Figure 21, indices of trip distance and area of convex hull are higher than the values of Area-Cluster-B; from Figure 20, the number of destinations is

numerous; from Figure 23, CO 2 Emission is higher in urban clusters; from Figure 19, the values of staying duration at the destination are shorter next to Area-Cluster-B.

14 numerous; from Figure 23, CO 2 Emission is higher in urban clusters; from Figure 19, the values of staying duration at the destination are shorter next to Area-Cluster-B. Because of these results, we understood that Area-Cluster-A has the characteristic of 1 st PC: long distance and long time trip component and doesn t have the characteristic of 2 nd PC: concentrating in Main Destination component. About Area-Cluster-B, 1 st to 4 th PC are negative and 5 th PC is positive, moreover, we can read that: from Figure 18 and Figure 21, trip distances are shorter and the area of the convex hull is smaller next to Area-Cluster-D; from Figure 19, Duration from Trip Start to Trip End and the values of staying duration at the destination are shorter; from Figure 20, the number of destinations is numerous; from Figure 22 and Figure 23, Calorie Consumption is higher and CO 2 Emission is lower. About Area-Cluster-C, 1 st to 3 rd PC are positive and 4 th, 5 th PC is negative, moreover we can read that: from Figure 18 and Figure 21, residents in this cluster area have a tendency moving a long distance and a wide range; from Figure 19, the values of staying duration at the destination are longer next to Area-Cluster-D; from Figure 23, CO 2 Emission is larger. Finally, about Area-Cluster-D, 2 nd to 5 th PC are positive and we can read that: from Figure 18 and Figure 21, trip distances are shorter than Area-Cluster-C and the area of the convex hull is the smallest; from Figure 19, staying duration in the residence area is longer; from Figure 20, the number of destinations is the least. From these results, we summarized the characteristics of each cluster of area and the comparison between urban clusters and suburban clusters (Table 5). Figure 18. Relations between area clusters and distance indices

15 Figure 19. Relations between area clusters and time indices Figure 20. Average of Number of Destination Figure 21. Average of Area of Convex hull Figure 22. Average of Calorie Consumption Figure 23. Average of CO 2 Emission

16 Table 5. Characteristics of each cluster of area Clusters Characteristics of each cluster Comparison with Urban and Suburban Urban Suburban A B C D Wider range of moving than Area-Cluster-C Larger number of destinations than Area-Cluster-B Larger CO 2 Emission than Area-Cluster-B The shortest duration at the residence area The shortest distance of moving Narrower range of moving next to Area-Cluster-D The largest number of destinations The shortest duration from trip start to trip end The smallest CO 2 Emission The longest moving distance The widest range of moving Longer duration of staying at destination next to Area-Cluster-D The largest CO 2 Emission Shorter distance of moving than Area-Cluster-C The narrowest range of moving The smallest number of destinations The longest duration of staying at the residence area Moving distance is shorter Staying duration at the residence area is shorter (Number of tripper inside the residence area is smaller) Number of destinations is larger Calorie Consumption is larger and CO 2 Emission is smaller Moving distance is longer Staying duration at the residence area is longer (Number of tripper inside the residence area is larger) Number of destinations is smaller Calorie Consumption is smaller and CO 2 Emission is larger We thought the meaning that each area cluster has against the areas. At first, Area-Cluster-B that holds the city centre around Matsuyama City has tendencies of residents for moving a short distance or a small range and Duration from Trip Start to Trip End is shorter, but the number of destinations is the biggest. We can think the reason about these is an influence of too small zones on the original People Flow Data, but if the tripper goes and returns from his residence to the destination, the number of destinations is one. Therefore, we think this influence is very little. In brief, trippers live in the city centre have the characteristic of moving that they trip to some destinations around his residence: like dropping in or wandering, besides their moving time or distance is shorter and their transportation emits less CO 2 like walking or a bicycle. Because of these, there is a high tendency that the trippers complete their live only in the city centre. Area-Cluster-A has the characteristics that the trip distance is longer than Area-Cluster-B, the area of moving is larger than Area-Cluster-C and the number of destinations is bigger than Area-Cluster-B. Because of these, the trippers have a tendency that they move to the city centre from the around of there and they drop in somewhere or wander about. Furthermore, from CO 2 Emission is large and the duration of going out in the residence area is short, we can think the trippers live in the area covered with Area-Cluster-A mainly have a lifestyle that he commutes to the city centre by a car or by public transportations and goes to his home at night. On the other hand, Area-Cluster-C has the characteristics that the trip distance and area of moving is large, besides the duration of staying at the destination is longer and CO 2 Emission is larger than Area-Cluster-D. Because of these, we can think a general suburban lifestyle that he commutes or goes out mainly to the city centre or a neighboring city area by a car or public transportations. In comparison with these, from the characteristics that the trip distance and trip range is smaller, the number of destinations is less and the duration of going out in the residence area is the longest, we can understand that the moving behavior of Area-Cluster-D; but it s suburban cluster, is similar to it of a city area. The reason of this is can be guessed from examining the characteristics of each area covered with Area-Cluster-D

17 Figure 24. Comparison between Area-Cluster-D and densely inhabited zones or depopulated areas Figure 24 shows areas belonging to Area-Cluster-D, the densely inhabited zones (2005, Ministry of Land, Infrastructure, Transport and Tourism) and the depopulated areas (same). From this, we can read that b (Hojotsuji), a (Nakadori), c (Takaokamachi), d (Nshihabumachi), e (Morimatsumachi) are in or near DID areas. In this case, we can understand the characteristic of Area-Cluster-D is similar to Area-Cluster-B that means the residents can complete their lives only in the area because the DID area has some shops and companies necessary to their lives. On the one hand, f (Oohira, Iyo City), g (Sunouchi, Touon City) are in or near depopulated areas, in this case, we can think the possibilities that a resident goes out only in his resident area especially in weekday or the transportation poor resident like an aged person can t go out and forced to stay in his resident area, ascribable to the unconvenient of transportations, the decrease of a number of shops by depopulated and so on

18 6. Conclusion 6.1. Summary In this paper, we focused on the information that is got from peoples moving behaviors, like an algebraical information: distance or time, and a geometrical information: shape of trip chain or centroid, and we proposed Trip Indices. By using these indices, we classified peoples moving behavior and analyzed the characteristics of areas. From these results, our purpose that grasping peoples moving behaviors easily and by the method of the grasping, treating characteristics of areas or urban cities, was achieved The problem to be solved Trip Indices has some problems to be solved. For example, indices of tripper who has action only in the resident area is calculated that Distance of Farthest Destination = Distance of Main Destination = zeroes and Staying Duration at the residence = Staying Duration at Main Destination = Staying Duration at Farthest Destination. Because of these, there is possible that such this tripper s data was evaluated excessively. Therefore, we can propose the remedial measure that we calculate the trip distance estimated by transportations and the duration of moving. About principal component scores, we used these only for the cluster analysis, but we think we can use these for more detailed analyzing of a person or an area because these scores are quantitative variables. Moreover, there is possible that we can propose the way of grasping peoples moving behaviors or the characteristics of areas easily by adding more geometrical information like an angle

19 Acknowledgement This research has been held as a part of group research (No.381) on Center for Spatial Information Science The University of Tokyo (CSIS), and we received generous support from the center for using People Flow Data. I would like to thank them for a grant that made it possible to complete this research. References [1] Toshihiro OSARAGI, 2009, Spatio-temporal distribution of railroad users for disaster prevention planning, Journal of Architecture and Planning (Transactions of AIJ) Vol. 74 No. 635 pp [2] Toshihiro OSARAGI, Ren SHIMADA, 2009, Spatio-temporal distribution of population on weekdays and holidays for estimating human damages from devastating earthquake, Journal of Architecture and Planning (Transactions of AIJ) Vol. 74 No. 635 pp [3] Takamasa AKIYAMA, Masashi OKUSHIMA, Yukiko OZAWA, 2010, Destination choice model with fuzzy classifier system for transport planning, Journal of Japan Society for Fuzzy Theory and Intelligent Informatics Vol. 22 No. 1 pp [4] Kouichi SHIMODA, Mitsuyuki ASANO, Atsushi NAKANO, 1991, Relationships between city s socio - economic and their transport characteristics by analysis of the data of nationwide person - trip survey, 26 th Journal of the City Planning Institute of Japan, pp [5] Mamoru TANIGUCHI, Takeomi MURAKAWA, Tetsuo MORITA, 1991, Analysis on relationship between urban characters and usage based on personal trip data, 34 th Journal of the City Planning Institute of Japan, pp [6] Yasunobu SHIMAZAKI, Yoshihide SEKIMOTO, Ryosuke SHIBASAKI, Yuki AKIYAMA, 2009, A study on the correlation between the number of stores and the time slot population based on person flow - The comparative analysis between the time and space interpolated person trip survey data and the census data -, Journal of the City Planning Institute of Japan No.44-3 [7] Hideyuki SUZUKI, Yoshihide SEKIMOTO, 2012, Factors associated with difference between actual and planned revenue in shopping malls, 21 st Papers and proceedings of the Geographic Information Systems Association Vol.21 CD-ROM [8] Akinori ASAHARA, Yaemi TERAMOTO, Kishiko MARUYAMA, Ryosuke SHIBASAKI, 2013, Representative path extraction with spatio-temporal network analysis, IPSJ Journal (CD-ROM) Vol.54-1 pp

Daily Estimation of Passenger Flow in Large and Complicated Urban Railway Network. Shuichi Myojo. Railway Technical Research Institute, Tokyo, Japan

Daily Estimation of Passenger Flow in Large and Complicated Urban Railway Network Shuichi Myojo Abstract Railway Technical Research Institute, Tokyo, Japan Railway passenger flow data including the on-board