CN110909627B

CN110909627B - Region POI configuration visualization method and system

Info

Publication number: CN110909627B
Application number: CN201911066041.XA
Authority: CN
Inventors: 刘康; 尹凌
Original assignee: Shenzhen Institute of Advanced Technology of CAS
Current assignee: Shenzhen Institute of Advanced Technology of CAS
Priority date: 2019-11-04
Filing date: 2019-11-04
Publication date: 2022-04-26
Anticipated expiration: 2039-11-04
Also published as: CN110909627A

Abstract

The invention relates to a region POI configuration visualization method, which comprises the following steps: acquiring data of each research area and POI in a city, and preprocessing the data to count the number or the proportion of each type of POI in each research area; training all the acquired POI data in each research area and city in the city by using a word embedding method to obtain real number vector representation of each POI type from POI spatial distribution; projecting the POI type represented by a real number vector to a two-dimensional semantic space by adopting a data dimension reduction technology; and rendering the POI types in the two-dimensional semantic space according to the counted number or the counted proportion of the POIs of each type in each research area so as to visualize the POI configuration condition of the research area. The invention also relates to a system for visualizing the configuration of the regional POI. The method and the device can visually express the configuration condition of the POI in the area, are beneficial to finding the characteristics of the POI configuration in the individual area, and can be used for conveniently comparing different areas.

Description

Region POI configuration visualization method and system

Technical Field

The invention relates to a method and a system for visualizing regional POI configuration.

Background

POI (point of interest) is a specific place where people gather and perform daily activities. The POI configuration conditions in urban areas such as administrative divisions, business circles, new planning areas and the like are researched and researched, so that the functions, vitality and development status of the areas can be known and evaluated, and the method has important significance for urban planning.

Currently, research and application are based on a POI self-contained classification system, and the number or proportion of POIs in each category in each area is counted to characterize the POI configuration condition of the area (as shown in the following table).

Type 1

Type 2

Type 3

Type 4

Type 5

…

Region 1

21 are provided with

34 are provided with

2 are provided with

88 pieces of

302 are provided with

…

Region 2

7 are provided with

1 is provided with

23 are provided with

42 are provided with

52 by

…

Region N

23 are provided with

334 are provided

1 is provided with

46 are provided with

890 pieces

…

However, this method has two major problems. First, POI configuration characterized in the form of a table is not conducive to finding POI configuration characteristics of a single area and comparing between multiple areas. Secondly, the POI classification system presents a tree structure (as shown in fig. 1), neglecting semantic similarity or relevance between POI types (especially subtypes). For example, in the high-grade map POI classification system, "playgrounds" and "zoos" where parents bring children to entertain and entertain belong to the general categories of "sports leisure services" and "scenic spots", respectively; "drugstores", "convenience stores" and "barbershops" reflecting the degree of convenience of regional life belong to the broad categories of "healthcare services", "shopping services" and "life services", respectively. The POI classification system based on the tree structure easily ignores the semantic relation among types and generates information fragmentation when the POI configuration condition is represented and understood.

Disclosure of Invention

In view of the above, a method and a system for visualizing the configuration of local POIs are needed.

The invention provides a region POI configuration visualization method, which comprises the following steps: a. acquiring each research area in a city and all POI data in the city, and preprocessing the acquired POI data in each research area in the city and all POI data in the city to count the number or the proportion of each type of POI in each research area; b. training all the acquired POI data in each research area and city in the city by using a word embedding method to obtain real number vector representation of each POI type from POI spatial distribution; c. projecting the POI type represented by a real number vector to a two-dimensional semantic space by adopting a data dimension reduction technology; d. and rendering the POI types in the two-dimensional semantic space according to the counted number or the counted proportion of the POIs of each type in each research area so as to visualize the POI configuration condition of the research area.

Wherein the pretreatment comprises the following steps:

acquiring POI in each research area by using ArcMap software;

and counting the number or proportion of each type of POI in each research area.

The step b specifically comprises the following steps:

taking each POI in a city as a core POI, and acquiring a neighbor POI in a space of 500 meters;

and (3) taking all the obtained type pairs [ core POI type and neighbor POI type ] as input, and training by using Word2Vec algorithm to obtain real number vector representation of each POI type.

The step c specifically comprises the following steps:

and (3) using a data dimension reduction algorithm to reduce the POI types represented by the 100-dimensional real number vector to a two-dimensional plane, wherein POI types with close distances still have higher semantic similarity or correlation.

The step d specifically comprises the following steps:

and rendering the POI types in different colors or sizes in a two-dimensional semantic space according to the number or the proportion of the POIs of each type in each research area so as to visualize the POI configuration condition of the research area.

The invention provides a region POI configuration visualization system, which comprises a preprocessing module, a vector representation module, a dimension reduction module and a rendering module, wherein: the preprocessing module is used for acquiring data of all POIs in each research area and each city in the city and preprocessing the acquired data of all POIs in each research area and each city in the city to count the number or the proportion of each type of POI in each research area; the vector representation module is used for training all the acquired POI data in each research area in the city and the city by using a word embedding method to obtain real number vector representation of each POI type from POI spatial distribution; the dimension reduction module is used for projecting the POI type represented by the real number vector to a two-dimensional semantic space by adopting a data dimension reduction technology; and the rendering module is used for rendering the POI types in the two-dimensional semantic space according to the counted number or the percentage of the POI types in each research area so as to visualize the POI configuration condition of the research area.

Wherein the pretreatment comprises the following steps:

acquiring POI in each research area by using ArcMap software;

The vector characterization module is specifically configured to:

The dimension reduction module is specifically configured to:

The rendering module is specifically configured to:

The method maps the POI types to the two-dimensional semantic space based on the word embedding technology and the data dimension reduction technology, visualizes the POI configuration condition of the area on the basis, and fully considers the semantic relationship among the POI types. Because semantically related or similar POI types are closer in the semantic space, the number or the proportion of the POI of each type is displayed in the semantic space, so that the mode and the characteristics of the POI configuration in the area can be more clearly found; and the method also facilitates convenient and intuitive comparison of POI configuration conditions of different areas.

Drawings

FIG. 1 is a schematic diagram of a POI classification system with a tree structure;

FIG. 2 is a flowchart of a method for area POI configuration visualization in accordance with the present invention;

FIG. 3 is a schematic diagram of a two-dimensional semantic space provided by an embodiment of the present invention;

fig. 4 is a schematic view of a POI configuration visualization effect of a lake ro subway station domain provided in an embodiment of the present invention;

fig. 5 is a schematic diagram illustrating a POI configuration visualization effect of a subway station domain of the chinese city according to an embodiment of the present invention;

FIG. 6 is a diagram of the hardware architecture of the system for area POI configuration visualization of the present invention.

Detailed Description

The present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.

Fig. 2 is a flowchart illustrating operations of the method for visualizing the configuration of local POIs according to the preferred embodiment of the present invention.

And step S1, acquiring data of each research area in the city and all POIs in the city, and preprocessing the acquired data of each research area in the city and all POIs in the city to count the number or the proportion of each type of POI in each research area.

Wherein, each research area in the city and all POI data in the city comprise: the method comprises the following steps of providing urban research area data and urban internal POI data, wherein the urban research area data are Polygon vector data of a Polygon type, and the urban internal POI data are point vector data comprising longitude and latitude coordinates and types.

Specifically, the method comprises the following steps:

in this embodiment, the intra-city research regions are 166 metro station regions (i.e., a region in a range of 700 meters around a metro station) in shenzhen city and 170 ten thousand city internal POI data in the range of shenzhen city acquired from the shengdian map open platform in 2018 and 9 months. The urban POI data comprises POI names, longitude and latitude coordinates, types and address information, wherein the types comprise 13 major types, 153 middle types and 498 minor types (shown in a table 1). This embodiment uses 498 subclasses.

TABLE 1 Gauder map POI Classification example

Class I (13 pieces)	Middle class (example)	Subclass (examples)
			Catering service	Middle dining room	Sichuan vegetable
Shopping service	Supermarket	Wallmar
			Life service	Ticket office	Airline ticket point of sale
Sports leisure service	Entertainment place	Bar
			Healthcare services	Special hospital	Hospital for orthopedics department
Accommodation service	Hotel	Five-star hotel
			Scenic spots	Scenic spots	National level scenic spot
Commercial residence	Building	Business office building
			Government agencies and social groups	Social group	Couplets of women
Science and education culture service	Media mechanism	Television station
			Transportation facility service	Port wharf	Passenger port
Financial insurance service	Bank	Bank of China
			Company enterprise	Company(s)	THE MEDICINES Co.

The pretreatment mainly comprises the following parts:

(1) the ArcMap software was used to obtain POIs in each study area.

(2) And counting the number or proportion of each type of POI in each research area.

And step S2, training the obtained POI data in each research area in the city and all POI data in the city by using a word embedding method to obtain real number vector representation of each POI type from POI spatial distribution. Specifically, the method comprises the following steps:

taking each POI in the city as a 'core POI', and acquiring all 'neighbor POIs' within the space range of 500 meters.

And (3) taking all the obtained type pairs [ core POI type and neighbor POI type ] as input, training by using a word embedding technology to obtain real number vector representation of each POI type, wherein the similarity between the vectors is used for measuring semantic similarity or correlation between the corresponding POI types.

Further, the present embodiment is implemented as follows:

(1) taking each POI in the city as a core POI, and acquiring neighbor POIs within a space range of 500 meters. If the total amount of POI is too large, a proportion of samples can also be randomly screened.

Specifically, a KDTree function in a Python toolkit PySal is adopted to construct a spatial index for all POI according to longitude and latitude so as to quickly search spatial neighbors.

(2) All the obtained type pairs [ core POI type and neighbor POI type ] are used as input, one of the most commonly used and famous Word embedding technology, namely Word2Vec algorithm is used for training to obtain real number vector representation of each POI type, and the similarity between the vectors represents semantic similarity or correlation between the corresponding POI types.

Specifically, the Word2Vec algorithm is implemented by tensflo, with the dimensionality of the real vector suggested to be set to 100.

The semantic similarity or correlation between POI types is calculated by the cosine similarity formula between the corresponding vectors. Table 2 shows the other POI types and corresponding cosine similarities most relevant to restaurant(s), "shopping mall(s)", and "vegetable market" (others), respectively. It can be seen that the Word2Vec algorithm can learn the semantic relationship between POI types well.

TABLE 2 semantic similarity/correlation between POI types calculated from real vector cosine similarity

Restaurant (other)		Shopping center		Vegetable market
						Sichuan vegetable	0.936	Recreation ground	0.713	Agricultural product market	0.856
Local special dish	0.903	Cinema for carrying out the method	0.710	Integrated market	0.723
						All-grass of Hunan province	0.896	Shop	0.667	Fruit market	0.656
Fast food restaurant	0.781	Brand bag store	0.656	Seafood market	0.527
						Chaozhou dish	0.764	Drech's series	0.633	Convenience store	0.491
Catering related places	0.666	Adida	0.618	Supermarket	0.447
						Chafing dish	0.642	Shopping related places	0.615	Drugstore	0.409
Convenience store	0.630	Naike	0.565	Maintenance point	0.389
						Vegetable for clearing true heat	0.624	Brand clothing store	0.563	Tobacco and wine exclusive shop	0.382
Hubei dish	0.614	Wanning	0.555	Flea market	0.366

And step S3, projecting the POI type represented by the real number vector to a two-dimensional semantic space by adopting a data dimension reduction technology.

Specifically, the method comprises the following steps: POI types closer in the space still have higher semantic similarity or relevance.

Further, the present embodiment uses one of the more commonly used data dimension reduction algorithms, i.e., the t-SNE algorithm, to reduce the POI types represented by the 100-dimensional real number vector to a two-dimensional plane, where the POI types with close distances still have higher semantic similarity or correlation, so that the two-dimensional plane is referred to as a semantic space.

Specifically, the t-SNE algorithm is realized by a TSNE function in a Python toolkit scimit-spare.

The 498 minor classes are reduced to a 2-dimensional plane with the results shown in FIG. 3. As can be seen from the box selection example, convenience stores, drug stores, supermarkets, service points, and the like, although belonging to different broad categories, all of them can reflect the degree of convenience of life, and have semantic similarity or correlation with each other, and thus are close to each other in the semantic space.

And step S4, according to the counted number or proportion of each type of POI in each research area, rendering the POI type in a two-dimensional semantic space by using different colors or sizes so as to visualize the POI configuration condition of the research area. Specifically, the method comprises the following steps:

referring to fig. 4 and 5, in the embodiment, POI types are rendered in different colors or sizes in a two-dimensional semantic space according to the number or proportion of each type of POI in a research area, so as to visualize the POI configuration condition of the research area. The application suggests using interactive mapping or development tools like Plotly, Mapbox, ArcGIS to hierarchically and dynamically display the names of POI types during zooming.

In this embodiment, an interactive visualization tool is developed based on Mapbox, and POI configuration visualization results of 166 metro station domains in shenzhen city are shown:

http://hpcc.siat.ac.cn/liuk/POI_configuration_cn/index.html。

referring now to FIG. 6, there is shown a hardware architecture diagram of the present invention regional POI configuration visualization system 10. The system comprises: the system comprises a preprocessing module 101, a vector characterization module 102, a dimension reduction module 103 and a rendering module 104.

The preprocessing module 101 is configured to acquire data of all POIs in each research area and city in the city, and preprocess the acquired data of all POIs in each research area and city in the city to count the number or ratio of each type of POI in each research area.

Specifically, the method comprises the following steps:

TABLE 1 Gauder map POI Classification example

The pretreatment mainly comprises the following parts:

(1) the ArcMap software was used to obtain POIs in each study area.

The vector representation module 102 is configured to train, by using a word embedding method, acquired data of all POIs in each research area and city in a city to obtain a real number vector representation of each POI type from POI spatial distribution. Specifically, the method comprises the following steps:

Further, the present embodiment is implemented as follows:

The dimension reduction module 103 is configured to project the POI type represented by the real number vector to a two-dimensional semantic space by using a data dimension reduction technique.

The rendering module 104 is configured to render the POI types in the two-dimensional semantic space with different colors or sizes according to the counted number or proportion of the POIs of each type in each research area, so as to visualize the POI configuration condition of the research area. Specifically, the method comprises the following steps:

http://hpcc.siat.ac.cn/liuk/POI_configuration_cn/index.html。

although the present invention has been described with reference to the presently preferred embodiments, it will be understood by those skilled in the art that the foregoing description is illustrative only and is not intended to limit the scope of the invention, as claimed.

Claims

1. A method for visualizing configuration of POI (point of interest) in an area, the method comprising the steps of:

a. acquiring each research area in a city and all POI data in the city, and preprocessing the acquired POI data in each research area in the city and all POI data in the city to count the number or the proportion of each type of POI in each research area;

b. training all the acquired POI data in each research area and city in the city by using a word embedding method to obtain real number vector representation of each POI type from POI spatial distribution;

c. projecting the POI type represented by a real number vector to a two-dimensional semantic space by adopting a data dimension reduction technology;

d. and rendering the POI types in the two-dimensional semantic space according to the counted number or the counted proportion of the POIs of each type in each research area so as to visualize the POI configuration condition of the research area.

2. The method of claim 1, wherein the pre-processing comprises:

acquiring POI in each research area by using ArcMap software;

3. The method according to claim 2, wherein said step b specifically comprises:

4. The method according to claim 3, wherein said step c specifically comprises:

5. The method according to claim 4, wherein said step d comprises the steps of:

6. A region POI configuration visualization system is characterized by comprising a preprocessing module, a vector characterization module, a dimension reduction module and a rendering module, wherein:

the preprocessing module is used for acquiring data of all POIs in each research area and each city in the city and preprocessing the acquired data of all POIs in each research area and each city in the city to count the number or the proportion of each type of POI in each research area;

the vector representation module is used for training all the acquired POI data in each research area in the city and the city by using a word embedding method to obtain real number vector representation of each POI type from POI spatial distribution;

the dimension reduction module is used for projecting the POI type represented by the real number vector to a two-dimensional semantic space by adopting a data dimension reduction technology;

and the rendering module is used for rendering the POI types in the two-dimensional semantic space according to the counted number or the percentage of the POI types in each research area so as to visualize the POI configuration condition of the research area.

7. The system of claim 6, wherein the pre-processing comprises:

acquiring POI in each research area by using ArcMap software;

8. The system of claim 7, wherein the vector characterization module is specifically configured to:

9. The system of claim 8, wherein the dimension reduction module is specifically configured to:

10. The system of claim 9, wherein the rendering module is specifically configured to: