JP5785830B2 - Asset evaluation system and resident activity prediction system - Google Patents

Description

  The present invention relates to a system and method for supporting operations for residents in local governments and administrations, for example, a system and method for outputting an effective arrangement / business hours plan for public facilities.

  When the national ID system enables various national information to be acquired in cooperation, various administrative services can be realized using such information.

  As an example of administrative services using various national information, Patent Literature 1 discloses a method for generating a public facility arrangement plan from resident information.

  Patent Document 2 discloses a method for evaluating public assets using user information.

Japanese Patent Laid-Open No. 08-110916 JP 2005-235164 A

  In the facility arrangement method described in Patent Document 1, it is possible to predict the required time from the address and age (resident information) of a resident to the facility and generate a facility arrangement plan.

  However, for example, there is a problem that the facility is located near the address even in the daytime for residents who work. In addition, it takes a considerable period from the generation of an arrangement plan to the actual construction of a facility based on the arrangement plan. Therefore, the facility is constructed based on an arrangement plan that does not reflect the inhabitants of the residents when the facility is actually constructed.

  In addition, in the facility evaluation data processing method described in Patent Document 2, the facility usage status, user satisfaction, and the like can be output as facility evaluation data. However, there was a problem that important asset value was not sought at facility evaluation. In addition, there was a problem that the facilities were not compared with many important facilities during the facility evaluation. In addition, there is a problem that the residents information cannot be used effectively when providing services for the residents.

A typical example of the invention disclosed in the present application is as follows. That is, an input unit for inputting facility information indicating a facility, resident information including part or all of the resident 's address, age, family structure, income, and asset information, resident activity information regarding resident activity, and facility And the asset evaluation information in which the information indicating the asset evaluation of the facility is associated with the storage unit, and based on the plurality of types of information of the resident information or / and the resident activity information , A clustering processing unit that performs clustering processing to obtain a cluster of residents in a multidimensional space corresponding to information , and predicts temporal changes for each population cluster generated by the clustering processing, and predicts residents as a result of the prediction Search for information indicating the property evaluation associated with facility information similar to the facility information input by the input unit from the property evaluation information and the resident trend prediction processing unit that generates information It has a search unit, the prediction resident information, based on the information indicating the asset evaluation retrieved, and a prediction effect calculation processing unit for obtaining the asset evaluation facility showing facility information entered by the input unit.

  According to another aspect of the present invention, based on resident information or / and resident activity information, the resident information based on the resident information related to the resident, the resident activity information related to the activities of the resident, and the facility information related to the facility. A community-specific clustering processing unit that performs clustering processing, a spatio-temporal clustering processing unit that clusters resident clusters in spatio-temporal based on resident clusters and resident activity information generated by resident clustering processing, and spatio-temporal of resident clusters The distance from the spatio-temporal cluster generated by the clustering process to the facility is calculated based on the location information included in the facility information, and based on the information indicating the facility usage rate included in the resident activity information and the distance, Extraction of resident activity model information showing the correlation between the distance from the space cluster to the facility and the utilization factor of the facility It has a people activity prediction processing unit.

  Further, according to another aspect of the present invention, based on resident information and / or resident activity information, a storage unit that stores resident information about resident, resident activity information about resident activity, and facility information about facility. A clustering processing unit for resident clustering, a spatiotemporal clustering processing unit for clustering resident clusters in spatio-temporal based on resident clusters and resident activity information generated by resident clustering processing, and facility information Generated by clustering by a facility extraction processing unit that extracts facilities included in a predetermined distance range from a spatiotemporal cluster generated by clustering by the spatiotemporal clustering processing unit based on the location information included, and by a clustering by the resident orientation clustering processing unit Predicts changes in the residents cluster over time, and As a result, based on the resident trend prediction processing unit that generates predicted resident information and the information indicating the utilization rate of the facility included in the predicted resident information and the resident activity information, the predicted use that calculates the number of users of the predicted facility A number-of-persons calculation processing unit, and a capacity excess / deficiency calculation processing unit for obtaining a difference between the predicted number of users of the facility and the number of accommodations of the facility extracted by the facility extraction processing unit.

  According to a typical embodiment of the present invention, by using a plurality of information including inhabitant information and activity status, planning of service for inhabitants, urban planning, prediction of effects when public investment is made, etc. It can be carried out.

It is a figure which shows an example of the local government and administrative management support method. It is a figure which shows an example of a structure of a local government and administrative management support system. It is a figure which shows an example of a resident tendency model clustering method. It is a figure which shows an example of a resident cluster. It is a figure which shows an example of the data structure of a spatiotemporal cluster. It is a figure which shows an example of inhabitant simulation. It is a figure which shows an example of the data structure of geographic information. It is a figure which shows an example of the data structure of the statistical information of resident activity information. It is an image figure which shows a past resident cluster and prediction resident information. It is a figure which shows an example of the data structure of prediction resident information. It is a figure which shows an example of inhabitant service support. It is a figure which shows an example of the data structure of a spatiotemporal cluster. It is a figure which shows an example of city planning. It is a figure which shows an example of the data structure of geographic information. It is a figure which shows an example of a public investment simulation. It is a figure which shows an example of the data structure of the data recorded on policy DB. It is a figure which shows the example of the data structure of budget allocation. It is a figure which shows an example of asset value evaluation.

  Embodiments of the present invention will be described below with reference to the drawings.

  First, the processing and system configuration example of the embodiment of the present invention will be described using FIG. 1 and FIG. The system according to the embodiment of the present invention is a system including a processing unit (210) that performs processing (111 to 116) and a storage unit (200). The storage unit (200) includes an input data recording unit (201), a generated data recording unit (202), and a program recording unit (203). The processing unit (210), the storage unit (200), and the input / output unit (220) for inputting the input data (201) and outputting the generated data (202) are connected by a communication path (230). . Each recording unit is realized by a storage device such as a magnetic disk device (Hard Disk Drive). The communication path (230) may be a network connecting other storage apparatuses. In that case, each recording unit may be a device connected to a network.

  Here, the data recorded in the input data recording unit (201) includes the resident information (101) including part or all of the resident's address, age, family structure, income, assets, etc., city hall, library, community center Usage history of public facilities such as hospital usage history, nursing care service usage history, public transportation usage history, inhabitant activity information (102) including part or all of store purchase history, and location of public facilities And geographic information including attributes such as the number of books in the library, the number of hospitalized inpatients, and the corresponding departments (internal medicine, surgery, etc.). The data recorded in the generated data (202) includes predicted resident activity information (104), predicted resident information (105), service plan (106), city plan (107), and optimal budget allocation (108). In addition, the program (203) is a resident tendency model clustering (111) program, a resident simulation (112) program, a resident trend prediction (113) program, a resident service support (114) program, a city planning (115) program, and a public investment simulation. (116) Includes a program.

  First, the processing unit (210) reads the program (203), and performs resident tendency model clustering (111) using the resident information (101) and the resident activity information (102) included in the input data (201). , A resident tendency cluster (103) is generated and stored as generated data (202). Next, by reading the program (203), the processing unit (210) uses the resident activity information (102) included in the input data (201) and the resident tendency cluster (103) included in the generated data (202). Then, a resident simulation (112) is performed to generate predicted resident activity information (104), which is stored as generated data (202). Next, by reading the program (203), the processing unit (210) uses the resident tendency cluster (103) included in the generated data (202) to perform resident trend prediction (113), and predicts resident information (105 ) Is generated and stored as generated data (202). Next, the processing unit (210) reads the program (203) and uses the predicted resident activity information (104), the resident tendency cluster (103), and the predicted resident information (105) included in the generated data (202). , Resident service support processing (114) is performed, a service plan (106) is generated, and stored as generated data (202). Next, the processing unit (210) reads the program (203) and uses the predicted resident activity information (104), the resident tendency cluster (103), and the predicted resident information (105) included in the generated data (202). Then, city planning (115) is performed, a city plan (107) is generated, and stored as generated data (202). Next, the processing unit (210) reads the program (203), performs a public investment simulation (116) using the predicted inhabitant information (105) included in the generated data (202), and optimizes budget allocation (108). ) Is generated and stored as generated data (202).

  Next, the resident tendency model clustering process (111) will be described with reference to FIG. First, residents are clustered using the resident information (301) and history statistical information (such as the frequency of use of public facilities) of the resident information and resident activity information (302), and the resident cluster (303) is obtained. For the clustering, for example, the k-means method described in “CM Bishop: Pattern recognition and machine learning, Springer Japan, 2006” can be used. Generally, the clustering result is represented as a lump on a multidimensional space equal to the number of types of information used for clustering. FIG. 4 shows an example of a resident cluster using only two types of information. Here, the vertical axis (401) represents the age, and the horizontal axis (402) represents the usage frequency of the library. The elderly cluster (411) is older, but the frequency of library use varies, and the full-time housewife-owned cluster (412) is more frequent, and the age is not as high as that of older people. Show. Next, for each generated resident cluster (303), spatio-temporal clustering (311) is performed using the activity information (302) of the resident included in the cluster to obtain a spatiotemporal cluster (304) for each resident cluster. . The spatio-temporal clustering method and the data format of the spatiotemporal cluster are described in JP 2011-008635 A. For example, the usage frequency information for each day of the library is used to cluster which day of the week the library used by the residents belonging to each resident cluster. The resident tendency cluster (300) is cluster information including the resident cluster (303) and the spatiotemporal cluster (304) obtained by the resident tendency model clustering process (111).

  Next, FIG. 5 shows a data example of the spatiotemporal cluster. The clusters shown in the figure are two spatio-temporal clusters representing the movement of a cluster without cooperative workers, where 501 is a cluster number, 502 is a time, and 503 is a coordinate. Here, the coordinates of the spatiotemporal cluster are represented as the barycentric position coordinates of the cluster, but may be represented by a plurality of coordinates so as to represent the cluster range. From 7:10 to 7:40, cluster 1 (510-513) moving to the left from (4000,2000) to (1000,2000), and from (3000,4000) to (7000, 4000), there is a cluster 2 that moves to the right. Here, the cluster number is a number uniquely assigned to each obtained cluster.

  Next, a resident simulation (112) will be described with reference to FIG. First, a resident activity model (604) is obtained using history statistical information of the resident activity information (601), a spatio-temporal cluster (602), and geographic information (603) (610). Next, the geographic information (603) is changed by the user (611). Next, an activity simulation is performed based on the changed data using the resident activity model (604) obtained in 610 (612), and a resident activity prediction (605) is obtained. An example of specific processing is shown below. First, examples of historical statistical information of geographic information (603) and resident activity information (601) are shown in FIGS. FIG. 7 shows geographic information indicating the location of the library. Reference numeral 701 denotes a library ID, reference numeral 702 denotes coordinates, and the figure stores the positions of two libraries (710, 711). FIG. 8 shows statistical information of resident activity information indicating the frequency of use of the library. Reference numeral 801 denotes a resident cluster ID, which represents a “double working child” cluster, 802 is a library ID, and 803 is an average monthly usage count of the resident included in the resident cluster. In the activity model generation (610) in the present embodiment, a linear approximation of the number of uses to the distance to the spatiotemporal cluster is performed. That is, using the information of FIGS. 5, 7, and 8, the distance between the library 1 (710) and the closest space-time cluster 1 (510 to 512) is 1000 and the library 2 (711) is closest to the space-time cluster. Since the distance of 2 (513 to 515) is 2000, when the simultaneous linear equations are solved using the values of FIG. 8, it can be expressed as usage frequency = −0.001 × distance + 2.5. Therefore, the resident activity model regarding the frequency of library use is represented by two real values (−0.001, 2.5). Next, when the library 3 is added to the coordinates (4000, 5000) in the geographic information change (611), the distance to the nearest space-time cluster 513 to 515 is 1000 in the activity simulation (612). (5, 3, 1.5) corresponding to each column in FIG. In this way, by using resident activity information and spatio-temporal clusters, the frequency of use when a new public facility is constructed can be predicted by simulation.

  Next, the resident trend prediction (113) will be described with reference to FIGS. In the resident trend prediction, predicted resident information (105) is obtained by using a plurality of resident clusters (303) obtained in the past in the resident tendency cluster (103, 300). For example, the predicted resident information (913, 922) of the area (903) is obtained from the past resident cluster information (911, 912, 921, 922) of the resident cluster of a certain area (901, 902). In this embodiment, the resident cluster is the address coordinate information of the cluster “Kyodo Shishi” (911, 912) and the cluster “elderly” (921) in 2005 and 2010, and is included in each cluster in each year. It is indicated that the inhabitants are present at a specific place in the area (901, 902). Here, the darkness of the cluster color represents the number of residents included in the cluster. In this example, the number of “double working children” (911) in 2005 is 120, and the “double working child presence” in 2010 is shown. ”(912) is 240 people,“ elderly people ”(921) in 2005 is 400 people, and“ elderly people ”(922) in 2010 is 300 people. At this time, extrapolation is performed in the prediction of population trends. For example, the number of “double-working children” has increased by 120 in five years, and the distribution area has also increased to the right, so it is included in the “double-working children” cluster in the same region (903) in 2015 There are 360 inhabitants and the area is further increasing to the right (913). In addition, since the “elderly” is unchanged in the area and the number of people is decreasing by 100, the number of inhabitants included in the “elderly” cluster in the same area (903) in 2015 is 200, and the area is 2005, 2010. Same as the year (923). An example of the predicted inhabitant information (105) is shown in FIG. Each column represents the resident cluster ID (1000), the number of people (1001), the cluster X coordinate (1002), the cluster Y coordinate (1003), the cluster width (1004), and the cluster height (1005). Predicted inhabitant information (ID = 5, 1010) of “double working child” and predictive inhabitant information (ID = 7, 1011) of “elderly”.

  Next, two examples of resident service support (114) for outputting a service plan (1120) will be described.

  First, an example of elderly visit plan generation will be described with reference to FIG. The distribution of the activity status of residents included in the elderly cluster (1101) in the population tendency cluster (103) obtained by the population tendency model clustering (111), for example, the frequency of use of public facilities is obtained, and the deviation from the distribution average Is calculated as biased data, and the name and address are output as a list of elderly people requiring a visit (1103) (1102). The extracted residents have an unusually low use frequency of public facilities compared to other elderly people, so it is necessary to visit and confirm the situation.

  Next, an example of child care support activity plan generation will be described with reference to FIGS. First, an activity candidate site is determined from predicted resident information (1111) obtained by resident trend prediction (113) (1112). For example, when the predicted resident information shown in FIG. 10 is obtained, the child-rearing support activities in 2015 are based on the coordinate values (1000, 5000) of the predicted resident information (1010) in which the number of people in the resident cluster “double working children” is 300 or more. Candidate site (1113). Next, in the resident activity model (1114) in the predicted resident activity information (600) obtained by the resident simulation (112) and in the resident tendency cluster (103) obtained by the resident orientation model clustering (111). The customer attracting prediction is performed using the spatiotemporal cluster (1115) of the “double working child” cluster (1116). For example, when the resident activity model (604) relating to the number of past child-rearing support activities is (−0.01, 60) and the spatiotemporal cluster is 1211-1213, the distance from the venue candidate to the spatiotemporal cluster is 1000. Therefore, when the child-rearing support activity is held around 10:30, the expected number of customers is 50.

  Next, the city planning (115) will be described with reference to FIGS. In this embodiment, a nursery placement plan is obtained. First, the resident distribution of the “double working child” cluster (ID = 5) in 2015 is extracted from the predicted resident information (1301) obtained by the resident trend prediction (113) (1302). For example, information of 913 and 1010 is obtained from the predicted inhabitant information shown in FIGS. Next, a spatiotemporal cluster corresponding to the predicted inhabitant information is extracted from the “double working child” cluster (1303) in the inhabitant tendency cluster (103) obtained by the inhabitant tendency model clustering (111) (1304). For example, the space-time cluster 2 (513 to 515) corresponding to the extracted population distribution is extracted with respect to the space-time cluster of the “double working child” cluster shown in FIG. Next, a predicted user is calculated (1306) from the extracted number of residents (1001) and the predicted resident activity information (104) and resident activity information obtained by the resident simulation (112). For example, when the nursery school utilization rate of the residents who belong to the cluster corresponding to the resident cluster ID = 5 is 0.2, the predicted number of persons at 1010 in FIG. Next, the nursery school in the vicinity of the spatiotemporal cluster extracted in 1304 is extracted from the geographical information (1307) including the number of persons in the existing nursery school (1308). An example of the nursery school geographic information (1307) is shown in FIG. Here, each column is a nursery school ID (1400), the number of accommodated persons (1401), an X coordinate (1402), and a Y coordinate (1403), and three kinds of nursery information are recorded. Assuming that the maximum value of the distance is 500, there are 1411 and 1412 nursery schools within a distance of 500 from the spatiotemporal cluster (513 to 515). Next, an excess or deficiency in the number of users is calculated from the difference between the number of users predicted in 1306 and the number of nurseries accommodated in 1308 (1309). In the case of this data example, while the predicted number of users is 72, the sum of the number of accommodated nurseries 1411 and 1412 (1401) is 50, so 22 people are insufficient. I understand. Next, a nursery school plan is generated (1301). In the case of this data example, a plan (1310) for arranging a nursery school for 22 persons at the center of gravity (5000, 4000) of the extracted spatiotemporal clusters (513 to 515) is generated.

  Here, using the predicted inhabitant information (1301) of the entire inhabitants and the spatio-temporal cluster (1303) of the entire inhabitants and using the geographical information (1307) of the disaster evacuation facilities, it is also possible to obtain a disaster evacuation facility arrangement plan. .

  Next, the public investment simulation (116) will be described with reference to FIGS. First, asset evaluation is performed on existing public facilities including other local governments and recorded in the policy DB (1500) (1501). An example of data in the policy DB is shown in FIG. Each column represents the type of public facility (1601), the resident cluster ID (1602) receiving benefits, and the effect per resident (1603). However, the resident cluster ID = 0 indicates the benefit to all resident. Accordingly, FIG. 16 shows that the nursery school has a benefit of 150 per capita for residents who have “double working children” (1611), and the library has a benefit of 1 per capita for all residents (1612). Next, the asset value data of the public facilities of the same kind is retrieved from the policy DB with respect to the public facilities to be newly started included in the budget allocation (1502) of the budget plan (1503). An example of budget allocation data is shown in FIG. Each column indicates the type of public facility (1701), the budget amount (1703), the number of people accommodated (1704), the X coordinate (1704), and the Y coordinate (1705). Since the types of public facilities are nursery school and library, 1611 and 1612 are searched in the similar policy search (1503). Next, the effect of the public facilities constructed by budget allocation is calculated (1505) from the retrieved asset value data and the predicted resident information (1504) obtained from the resident trend prediction (113). For example, when FIG. 10 is provided as predicted resident information, FIG. 16 is provided as policy DB data, and FIG. 17 is provided as budget allocation, first, the position (1000, 5000) of the nursery school 1711 to be constructed is the data 1611 related to the nursery school in the policy DB. Since it is in the position included in the cluster 1010 of the predicted resident information having the same ID as the resident cluster ID = 5, 360 people of the predicted resident information can be used, and 30 people can be used in the budget allocation data. It can be expected, and there are 4500 effects as a whole, from the effect 150 per person of the policy DB data 1611. Next, the location (8000, 3000) of the library 1712 to be constructed is expected to be used by residents with an arbitrary ID because the resident cluster ID of the data 1612 related to the library in the policy DB is 0. Since it is located in the cluster 1011, 100 people of the predicted inhabitant information can be used, and 100 out of 200 people in the budget allocation data can be expected to be used. From effect 1 to 100, there are 100 effects in total. Therefore, the effect of the budget allocation is 4500 + 100 = 4600. Next, the budget allocation is changed (1506). For example, without changing the total budget amount of 100,000,000 yen of 1711 and 1712, the number of nursery school capacity is doubled and the scale of the library is reduced (1721, 1722). The prediction effect after the change is calculated again (1505). In this case, 60 * 150 + 100 * 1 = 9100. Similarly, the budget allocation change (1506) and the budget effect calculation (1505) are repeated, and the most effective budget allocation is obtained as the optimum budget allocation (1507). For example, in the distribution of 1711, 1712 and 1721, 1722, 1721, 1722 having a high budget effect are output.

  Next, asset value evaluation (1501) will be described with reference to FIGS. First, multiple regression using the statistic of user resident information (1801) and resident activity information (1802) for each facility as explanatory variables, and the sum of residents satisfaction obtained from questionnaires (1803) as a dependent variable. Analysis is performed to obtain the correlation between the resident satisfaction and the resident information (1806). For example, the multiple regression coefficient of the multiple regression analysis result when the number of data of public facility use information in the resident activity information for a certain period (actually the number of respondents to the questionnaire) is used as an explanatory variable is the satisfaction per person in FIG. Degree (1603). Next, policy DB data (1807) is generated (1806) by assigning the attribute (1805) for each inputted facility. Here, the facility attribute in FIG. 16 is the cluster ID (1602) of the user.

200 ... storage unit 210 ... processing unit 220 ... input / output unit 230 ... communication path 101, 301, 1801 ... resident information 105, 913, 922, 1111, 1010, 1301, 1504,. Predicted inhabitant information 102, 302, 601, 1802 ... Resident activity information 104, 600 ... Predicted inhabitant activity information 303, 411, 412, 1101 ... Resident cluster 304, 602, 1115 ... Spatio-temporal cluster

Claims (11)

An input unit for inputting facility information indicating the facility;
Residents of address, age, family structure, income, and residents information, including part or all of the information assets, and the residents activity information, including the use history of the facilities of the residents, information and asset evaluation of the facility indicating the facility A storage unit for storing asset evaluation information associated with information indicating
Based on a plurality of types of information of the resident information or / and the resident activity information , a resident-oriented clustering processing unit that performs a clustering process for obtaining a mass of residents on a multidimensional space corresponding to the plurality of types of information ;
Said predicted time change for each population clusters generated by the clustering process, residents trend prediction processing unit for generating a prediction resident information as a result of the prediction,
A search processing unit for searching for information indicating an asset evaluation associated with facility information similar to the facility information input by the input unit from the asset evaluation information;
A prediction effect calculation processing unit for obtaining an asset evaluation of the facility indicated by the facility information input by the input unit based on the predicted resident information and the information indicating the searched asset evaluation; Asset valuation system. The asset evaluation system according to claim 1,
The asset evaluation system according to claim 1, wherein the information indicating the asset evaluation in the asset evaluation information includes information in which a resident cluster and a satisfaction level of a resident belonging to the resident cluster are associated with each other . The asset evaluation system according to claim 1,
The resident trend prediction processing unit predicts a temporal change of the resident cluster from the past resident information for each resident cluster. A storage unit for storing inhabitant information including part or all of inhabitant address, age, family structure, income, and asset information, inhabitant activity information on inhabitant activities, and facility information on facilities;
Based on a plurality of types of information of the resident information or / and the resident activity information, a resident-oriented clustering processing unit that performs a clustering process for obtaining a mass of residents on a multidimensional space corresponding to the plurality of types of information ;
A spatio-temporal clustering processing unit that performs clustering in spatio-temporal for each resident cluster based on the resident cluster generated by the clustering process and the resident activity information;
The distance from the spatiotemporal cluster generated by the spatiotemporal clustering process of the resident cluster to the facility is calculated based on the positional information included in the facility information, and the utilization rate of the facility included in the resident activity information is calculated. A resident activity prediction processing unit that extracts resident activity model information indicating a correlation between a distance from the space-time cluster to the facility and a utilization factor of the facility based on the information indicating and the distance; Resident activity prediction system. It is a resident activity prediction system according to claim 4,
It further has an input part for inputting information indicating the change or addition of the facility,
The resident activity prediction processing unit calculates a distance from the space-time clustering to the facility from the location information of the facility to be relocated or added, and based on the calculated distance and the resident activity model information And calculating a utilization rate for the facility whose location is changed or added. It is a resident activity prediction system according to claim 4 ,
The inhabitant information is information including part or all of the inhabitants' address, age, family structure, income, assets,
The resident activity information is information including a part of or all of a public facility use history, a hospital use history, a care service use history, a public transport use history, and a purchase history. system. It is a resident activity prediction system according to claim 4,
A resident activity prediction system comprising: a resident extraction processing unit that extracts resident corresponding to biased resident activity information from the distribution of the resident activity information for each resident cluster. It is a resident activity prediction system according to claim 4,
A resident activity prediction system characterized in that a predicted number of local activity participants is obtained based on the resident cluster, the resident cluster and the predicted activity information. It is a resident activity prediction system according to claim 4,
A resident trend prediction processing unit for generating predicted resident information indicating a prediction of temporal changes in the resident cluster generated by the resident clustering process;
A resident activity prediction system characterized in that a facility layout plan is obtained using the predicted resident information. It is a resident activity prediction system of Claim 9, Comprising :
A resident activity prediction system, wherein a facility layout plan is obtained using the spatiotemporal cluster and the predicted resident information. A storage unit for storing inhabitant information including part or all of inhabitant address, age, family structure, income, and asset information, inhabitant activity information on inhabitant activities, and facility information on facilities;
Based on a plurality of types of information of the resident information or / and the resident activity information, a resident-oriented clustering processing unit that performs a clustering process for obtaining a mass of residents on a multidimensional space corresponding to the plurality of types of information;
A spatio-temporal clustering processing unit that performs clustering in spatio-temporal for each resident cluster based on the resident cluster generated by the clustering process and the resident activity information;
A facility extraction processing unit for extracting a facility included in a predetermined distance range from a spatiotemporal cluster generated by clustering by the spatiotemporal clustering processing unit based on position information included in the facility information;
A resident trend prediction processing unit that predicts temporal changes in a resident cluster generated by clustering by the clustering processing unit for the resident tendency and generates predicted resident information as a result of the prediction;
A predicted user number calculation processing unit that calculates the number of users of the predicted facility based on the predicted resident information and information indicating the utilization rate of the facility included in the resident activity information;
  A resident activity prediction system comprising: a capacity excess / deficiency calculation processing unit that obtains a difference between the predicted number of users of the facility and the capacity of the facility extracted by the facility extraction processing unit.