CN106780262A - A kind of same bit pattern for considering urban road network constraint finds method and device - Google Patents

Abstract

The invention discloses a method and a device for discovering co-location patterns considering the constraints of urban road networks. The method includes: constructing a second-order instance adjacency table for the target area under the map projection, the second-order instance adjacency table includes all instances in the target area whose distances to their reachable distances are within a preset distance attenuation threshold and are of different types. Instance pair sets and their reachable distance values; calculate the network kernel density value of each instance under the influence of other types of instance sets different from this instance type according to the preset distance attenuation threshold and the second-order instance adjacency table; according to the network kernel density The density value is calculated to obtain the average influence of each instance set on other types of instance sets; the popularity of each candidate co-location pattern is calculated according to the average influence, and the popular co-location pattern among the candidate co-location patterns is determined according to the preset popularity threshold. Through the present invention, the accuracy of co-location pattern mining on urban facility data is improved.

Description

A Method and Device for Discovering Co-location Patterns Considering Urban Road Network Constraints

technical field

The invention relates to the field of spatial data pattern mining, in particular to a method and device for discovering co-located patterns considering the constraints of urban road networks.

Background technique

In the past few decades, the development of population, housing, infrastructure and the expansion of employment scale have made my country's urban development situation show a trend of rapid development. With the rapid development of science and technology and the increasing expansion of data resources, the follow-up "smart city" will gradually turn to integrated processing of multi-source spatial data, and urban basic service facility data covers the location and attribute information of various elements of the city, as the basis of the city. The key to the database, how to extract useful distribution rules and model feature information from it to guide the rational planning and layout of new towns has become a key issue in urban development. Most of the urban facility data exists in the form of points. The common method to solve point pattern discovery is "co-pattern mining". Wide range of applications.

Homolocation pattern mining is to mine a series of spatially interdependent feature type combinations from the actual point data set, which generally includes two steps: first, through the preset distance threshold, in the homogeneous Euclidean space Determine whether there is adjacency relationship between individuals with different characteristics, and establish a second-order entity adjacency relationship table; then, use the frequent item mining method to obtain significant homolocation patterns.

However, there are various constraints in the real urban space, and there are great limitations in using this mining framework: First, traditional methods are mostly based on Euclidean distance, which considers planar space to be homogeneous and isotropic. However, in the real urban space, many phenomena related to people occur in the traffic network, such as traffic accidents, street events, and the distribution of infrastructure. Second, the traditional method uses a one-size-fits-all distance threshold to judge the proximity of instances, and treats instances within the threshold as close and far away as the same. In fact, according to the first law of geography, things that are close are more closely related, so the impact of close instance relations on the size of the pattern popularity index is greater than that of long-distance instance relations. Under the above two premises, the use of traditional algorithms will directly lead to deviations in the final results, so that some unpopular patterns are wrongly judged as co-located patterns, or some interesting patterns are ignored.

Urban public service facilities are an important part of the urban basic database, and the study of their spatial distribution characteristics can provide an important scientific basis for urban planning and decision-making. The intensive development of the city advocates the rational allocation of facilities, and the discovery of the distribution pattern of urban facilities is the premise of realizing this goal. Traditionally, the understanding of space is homogeneous, and most of the facilities in the city are distributed in the human-induced European-style space with the road skeleton as the limiting factor. The conventional homolocation pattern mining method must not be able to solve urban problems well.

Contents of the invention

The present invention aims to solve the problems described above. The object of the present invention is to provide a method and device for discovering co-location patterns considering the constraints of urban road networks to solve the above problems.

The present invention provides a method for discovering co-location patterns considering the constraints of urban road networks, including: constructing a second-order instance adjacency table for a target area under map projection, and the second-order instance adjacency table includes all instances in the target area A set of instance pairs whose reachable distances are within the preset distance attenuation threshold and of different types and their reachable distance values; according to the preset distance attenuation threshold and the second-order instance proximity relationship table, the distance between each instance and this instance is calculated The network kernel density value under the influence of other types of instance collections of different types; calculate the average influence of each instance collection on other types of instance collections according to the network kernel density value; calculate each candidate homolocation pattern according to the average influence Popularity, according to the preset popularity threshold to determine the popular homolocation pattern among the candidate homolocation patterns.

The above method also has the following characteristics: the construction of the second-order instance neighbor relationship table includes:

Store the reachable distance values between multiple adjacent instance pairs and their corresponding instances in a two-dimensional hash table TIns_net ₂ , and each cell unit in the table is expressed by a set of triples as follows:

TIns_net ₂ (e _x ,e _y )={<o _i ,o _j ,Rdis(o _i ,o _j )>,...},

Among them, (e _x , e _y ) are two spatial object instances, o _i , o _j are two adjacent instances, and Rdis(o _i , o _j ) is the reachable distance value between two adjacent instances;

according to

Rdis(o _i ,o _j )＝g(y)*(o _i -o _j ) _net(t) |h _t

Calculate the reachable distance value between two adjacent instances, wherein, Rdis(o _i , o _j ) is the reachable distance value between the two adjacent instances, g(y) is a binary function, if the instance o _i reaches The direction along the road of instance o _j is opposite to the traffic direction of the road, the value is 0, otherwise the value is 1, (o _i -o _j ) _net(t) is the shortest path time from instance o _i to instance o _j , h _t is the density attenuation threshold based on the path time, which is a constraint condition for finding the shortest path time, which means that the reachable distance between instances must meet the h _t threshold, otherwise instance o _i to instance o _j is unreachable.

The above-mentioned method also has the following characteristics: the network kernel density value calculated according to the preset distance attenuation threshold and the second-order instance adjacency table for each instance under the influence of other types of instance sets different from the implementation type includes:

according to

Calculate the network kernel density value of the instance o _i of type e _x under the influence of instance set O'(e _y ) of type e _y ,

in, is the network kernel density value of instance o _i of type e _x under the influence of instance set O'(e _y ) of type e _y ,

is a subset of all instances of type e _y on the region,

n _max is the maximum number of instances of a single type on the region,

n(O'(e _y )->o _i ) is the number of instance pairs reachable from the instance in O'(e _y ) to o _i ,

The value range of the calculation result of this formula is (0,1].

The above method also has the following characteristics: the average influence of each instance set on other types of instance sets is calculated according to the network kernel density value, including:

according to

Calculate the average influence of the instance set O'(e _y ) on the instance set O'(e _x ),

in, is the average influence of instance set O'(e _y ) on instance set O'(e _x ),

is the network kernel density value of instance o _i of type e _x under the influence of instance set O'(e _y ) of type e _y ,

_is a subset of the set of all instances of type ex on the region,

n(O'(e _x )) is the number of instances in the instance set O'(e _x ),

n(e _x ) is the number of all instances of type e _x in the area.

The above method also has the following features: the calculation of the popularity of each candidate co-location pattern according to the average influence, and determining the popular co-location pattern among the candidate co-location patterns according to the preset popularity threshold include:

according to

Calculate the popularity of a given candidate pattern,

Among them, PI _CP is the popularity of a given candidate pattern, and the value range is (0,1],

Tins_net _CP is an instance table composed of group instances of the candidate pattern CP, which is obtained by connecting the non-repeated instance pairs related to the type in the CP in the second-order instance adjacency relation table Tins_net ₂ through group instances,

min(.) is used to calculate the minimum value of the input set,

It is used to calculate the projection of an instance of type ex on the instance table _{Tins_net CP} ,

collection of instances set of instances the average influence of

When the calculated popularity of a candidate co-location pattern is greater than a set popularity threshold, the candidate is determined to be a popular co-location pattern.

The present invention also provides a device for discovering co-located patterns considering the constraints of urban road networks, including: a second-order instance adjacency table construction module, used to construct a second-order instance adjacency table for a target area under map projection, the second-order instance adjacency table The instance proximity relationship table contains all instance pairs in the target area whose distances and reachable distances are within the preset distance attenuation threshold and of different types and the reachable distance values between them;

The network kernel density calculation module is used to calculate the network kernel density value of each instance under the influence of other types of instance sets different from this instance type according to the preset distance attenuation threshold and the second-order instance proximity relationship table;

The average influence calculation module is used to calculate the average influence of each instance set on other types of instance sets according to the network kernel density value;

The popular co-location pattern acquisition module is configured to calculate the popularity of each candidate co-location pattern according to the average influence, and determine the popular co-location pattern among the candidate co-location patterns according to a preset popularity threshold.

The above-mentioned device also has the following characteristics: the second-order instance neighbor relationship table construction module is specifically used to store a plurality of adjacent instance pairs and the reachable distance values between the corresponding instances in a two-dimensional hash table TIns_net ₂ , each cell unit in the table, is represented by a set of triples as follows:

TIns_net ₂ (e _x ,e _y )={<o _i ,o _j ,Rdis(o _i ,o _j )>,...},

Among them, (e _x , e _y ) are two spatial object instances, o _i , o _j are two adjacent instances, and Rdis(o _i , o _j ) is the reachable distance value between two adjacent instances;

according to

Rdis(o _i ,o _j )＝g(y)*(o _i -o _j ) _net(t) |h _t

Calculating the reachable distance value between two adjacent instances, wherein Rdis(o _i , o _j ) is the reachable distance value between the two adjacent instances,

g(y) is a binary function, if the direction along the road from instance o _i to instance o _j is opposite to the direction of road traffic, the value is 0, otherwise it is 1,

(o _i -o _j ) _net(t) is the shortest path time from instance o _i to instance o _j ,

h _t is the density decay threshold based on path time, which is a constraint condition for finding the shortest path time, indicating that the reachable distance between instances must meet the h _t threshold, otherwise instance o _i to instance o _j is unreachable.

The above device also has the following characteristics: the network kernel density calculation module is specifically used for

Calculate the network kernel density value of the instance o _i of type e _x under the influence of instance set O'(e _y ) of type e _y ,

in, is a subset of all instances of type e _y on the region,

n _max is the maximum number of instances of a single type on the region,

n(O'(e _y )->o _i ) is the number of instance pairs reachable from the instance in O'(e _y ) to o _i ,

The value range of the calculation result of this formula is (0,1], which describes the influence of the instance set O'(e _y ) on the instance o _i under the network constraints;

The above-mentioned device also has the following characteristics: the average influence calculation module is specifically used for

Calculate the average influence of the instance set O'(e _y ) on the instance set O'(e _x ),

in, is the network kernel density value of instance o _i of type e _x under the influence of instance set O'(e _y ) of type e _y ,

_is a subset of the set of all instances of type ex on the region,

n(O'(e _x )) is the number of instances in the instance set O'(e _x ),

n(e _x ) is the number of all instances of type e _x in the area.

The above-mentioned device also has the following characteristics: the popular parity mode acquisition module is specifically used for

Calculate the popularity of a given candidate pattern,

Among them, PI _CP is the popularity of a given candidate pattern, and the value range is (0,1],

Tins_net _CP is an instance table composed of group instances of the candidate pattern CP, which is obtained by connecting the non-repeated instance pairs related to the type in the CP in the second-order instance adjacency relation table Tins_net ₂ through group instances,

min(.) is used to calculate the minimum value of the input set,

It is used to calculate the projection of an instance of type ex on the instance table _{Tins_net CP} ,

collection of instances set of instances the average influence of

When the calculated popularity of a candidate co-location pattern is greater than a set popularity threshold, the candidate is determined to be a popular co-location pattern.

The present invention proposes a colocation mode discovery method considering the urban road network constraints, and makes two extensions on the basis of the traditional colocation mode discovery method:

(1) Based on the fact that the interconnection of urban spatial facilities occurs in network path distances rather than Euclidean distances, the method of spatial kernel functions is placed in the network structure, and specific services that consider attribute information such as urban road traffic capacity and directions are used. The time-based accessibility index replaces the traditional two-dimensional Euclidean distance to measure the proximity between spatial facilities.

(2) Transform the traditional popularity index for judging the interestingness of the model, and add the original index calculated simply by the number of instance connections into the adjustment parameter of reachable distance weight. Compared with other existing methods in this field, the present invention respects the fact that "people's movement mainly depends on the road network in the city" and "the first law of geography", thus improving the efficiency of co-location pattern mining on urban facility data. Accuracy, and more realistic and practical value.

Description of drawings

The drawings constituting a part of the present invention are used to provide a further understanding of the present invention, and the schematic embodiments and descriptions of the present invention are used to explain the present invention, and do not constitute an improper limitation of the present invention. In the attached picture:

FIG. 1 is a flow chart of a method for discovering co-located patterns considering urban road network constraints according to Embodiment 1 of the present invention;

Fig. 2 is a schematic diagram of a method for constructing a second-order instance neighbor relationship table;

Figure 3 is a schematic diagram of an instance table composed of group instances of candidate co-location patterns;

Fig. 4 is a schematic structural diagram of an apparatus for discovering a parity pattern considering urban road network constraints according to Embodiment 2 of the present invention.

detailed description

In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments It is a part of embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention. It should be noted that, in the case of no conflict, the embodiments in the present application and the features in the embodiments can be combined arbitrarily with each other.

In order to better illustrate the technical solutions provided by the embodiments of the present invention, the following concepts are first described:

Shp data: short for shapefile, is an open spatial data format developed by ESRI. The spatial graphic elements and corresponding two-dimensional attributes of this format are managed through index files.

Co-location pattern: Co-location pattern C is a subset of a set of spatial features, where the number of types contained in C is called length, or order. In mining problems, there are two commonly used concepts: popular homolocation pattern and candidate homolocation pattern. They have no difference in form of expression, but have different implied meanings. Candidate homolocation patterns are a group of feature types with potential homolocation relationships, which are generally obtained from popular homolocation patterns with a lower or higher order than them. If they pass the popularity verification, they will become popular homolocation patterns. If the order of the candidate colocation pattern is equal to n, it is called a candidate n-order colocation pattern or a popular n-order colocation pattern.

Clique instance: It is a set of instances with different types that are adjacent to each other in space.

Instance table: In order to facilitate the calculation of the pattern, the types of the group instances will be arranged in a fixed order (such as the dictionary order of the types in the pattern) and stored in the form of a table.

The invention provides a method and device for discovering co-location patterns considering the constraints of urban road networks. The input is the road linear Shp data and service facility point Shp data of the target area under the same map projection O={o ₁ ,o ₂ ,...,o _n }, n is the number of facilities. Service facility data (hereinafter referred to as instances) undergoes sensitive data elimination, coordinate conversion and completeness processing to ensure that each piece of data contains facility type, X and Y coordinates, and the type set involved in these facility data is E={e ₁ , e ₂ ,…,e _m }, m is the number of types of facilities. The instances of type ex on the region are stored in the set O( _ex ), and the number of its instances is denoted as _n ( _ex ). The road network data undergoes topology inspection, coordinate transformation and completeness processing to ensure that each piece of data contains road grade, type, and direction information. In addition, the distance threshold h and popularity threshold PI_pre also need to be set. The data output is the same bit pattern that satisfies the condition.

The method and device for discovering co-location patterns considering the constraints of urban road networks according to exemplary embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

Embodiment one

Fig. 1 is a flow chart showing a method for discovering co-located patterns considering urban road network constraints according to Embodiment 1 of the present invention.

Step 101, constructing a second-order instance adjacency table for the target area under the map projection, the second-order instance adjacency table includes all instances in the target area whose distances to and reachable distances are within the preset distance attenuation threshold and are of different types reachable distance value.

On the premise of network colocation pattern mining, it is necessary to find out whether different types of instance points are reachable. If the reachable distance is within the threshold h, it means that the two instance points are adjacent. The urban road network has differences in traffic direction and traffic capacity. Therefore, the fact that the instance A is reachable to B does not mean that B is reachable from A. Under this assumption, the process of constructing the second-order instance neighbor relationship table under the network space kernel density constraint is as follows: First, transform the road network into a linear reference system based on road segments, which are the distances between two adjacent road intersections. second, divide each road segment into equal-length linear arc segments, which are called basic linear units; third, find all instances in the area that are within h and different from their type Instance, the instance proximity relationship and its reachable distance are stored in an m*m two-dimensional hash table, which is defined as a second-order instance proximity relationship table TIns_net ₂ under the constraint of network space kernel density. As shown in Figure 2, * represents a non-empty cell unit, and the triplet set of the cell unit at row x and column y is denoted as TIns_net ₂ (e _x , e _y ). Each cell unit in the instance table stores the adjacent instance pair corresponding to the second-order pattern (ex ,e _y ) and its reachable distance value, through a triple set TIns_net _{2 (ex ,e y ) =} {<o _i , o _j ,Rdis(o _i ,o _j )＞,...} expression.

Assuming that there is an instance o _i of type ex at position _x , the calculation method of the adjacent instances of o _i and their reachable distance is: take the basic linear unit closest to o _i as the generating element, and search for the shortest path length from the generating element For instance points of different types within the threshold h, pair o _i with these instance points and retain their distance values.

Considering the network direction and traffic capacity, the reachable distance Rdis(o _i ,o _j ) from instance o _i to o _j can be expressed as:

Rdis(o _i ,o _j )＝g(y)*(o _i -o _j ) _net(t) |h _t (1)

In the above formula, g(y) is a binary function, if the direction along the road from o _i to o _j is opposite to the road traffic direction, the value is 0, otherwise the value is 1; (o _i -o _j ) _net(t) is the shortest path time from o _i to o _j , and h _t represents the density attenuation threshold based on the path time, which is a constraint condition for finding the shortest path time, indicating that the reachable distance of the instance must satisfy the h _t threshold, otherwise o _i to o _j is unreachable.

Step 102, calculate the network kernel density value of each instance under the influence of other types of instance sets different from this instance type according to the preset distance attenuation threshold and the second-order instance neighbor relationship table.

The spatial kernel density value can be expressed in network space as:

In the above formula, f(x) is the kernel density value at the spatial position x; h is the distance attenuation threshold, which is the highest threshold of the distance between the instance points that are adjacent to x; (xo _i ) _net is the distance between x and the adjacent instance point o The network reachable distance of _i ; n is the number of instance points whose distance from x is less than h; K represents the spatial weight function, and the geometric meaning of this function is that as the distance from position x to each instance point increases, its density value gradually changes small. Many scholars have proved that the selection of K has little influence on the pattern distribution results. In the present invention, we choose Gaussian function as the weight function. The expression formula of the Gaussian function is as follows:

In the above formula, exp(.) represents an exponential function with the natural constant e as the base. K(x) has a typical "bell-shaped" curve feature, which has three adjustable parameters a, b and c, among which a determines the peak height of the curve, b determines the position of the abscissa where the peak appears, and c determines the width of the curve. The present invention adopts a standard two-dimensional Gaussian kernel function, and a=c=1, and b=0.

In conventional co-location pattern mining, instance connections are treated without directionality. However, when road constraints are considered, instance connections need to consider directionality. Under this premise, the network kernel density value of instance o _i (type e _x ) under the influence of instance set O'(e _y ) of type e _y is defined as:

In the above formula, is a subset of all instances of type e _y on the region, n _max is the maximum number of instances of a single type on the region, n(O'(e _y )->o _i ) is O'(e _y ) The number of instance pairs reachable from the instance in o to _i . This formula is transformed from the network space kernel density model, and its value range is (0,1], which describes the influence of the set O'(e _y ) on the instance o _i under the network constraints.

In step 103, the average influence of each instance set on other types of instance sets is calculated according to the network kernel density value.

Based on formula (4), the average influence of the instance set O'(e _y ) on the instance set O'(e _x ) is calculated by the following formula:

In the above formula, n(O'(ex )) represents the number of instances in the set O'(e _x ), and _n (e _x ) represents the number of all instances of type _ex on the region. Compared with the conventional method of homotopic pattern mining, the average influence emphasizes both the interconnection between different types of instances in the candidate pattern and the degree of participation of a single type instance in the candidate pattern.

Step 104, calculate the popularity of each candidate co-location pattern according to the average influence, and determine the popular co-location pattern among the candidate co-location patterns according to a preset popularity threshold.

Based on the formula (5), the calculation formula of the popularity PI _CP of the candidate pattern based on the network kernel density is given as follows:

In the above formula, PI _CP is the popularity of a given candidate pattern, and the value range is (0,1]. As the length of the candidate pattern increases, PI _CP will decrease accordingly, and Tins_net _CP is the CP of the candidate pattern An instance table composed of group instances, which is obtained by connecting the non-repeated instance pairs related to the type in the CP in the second-order instance neighbor relation table Tins_net ₂ through group instances (as shown in Figure 3), min(. ) is used to calculate the minimum value of the input set, It is used to calculate the projection of an instance of type ex on the instance table _{Tins_net CP} , collection of instances set of instances average influence.

When the calculated popularity PI _CP for a candidate co-location pattern is greater than the set popularity threshold PI_pre, it is determined that the candidate co-location pattern is a popular co-location pattern.

Fig. 4 is a schematic diagram showing the structure of an apparatus for discovering a parity pattern considering urban road network constraints according to Embodiment 2 of the present invention.

Referring to Figure 4, the device for discovering co-location patterns considering the constraints of the urban road network includes:

The second-order instance neighbor relationship table construction module 401 is used to construct a second-order instance neighbor relationship table for the target area under the map projection. Set the set of instance pairs of different types within the distance attenuation threshold and the reachable distance value between them;

A network kernel density calculation module 402, configured to calculate the network kernel density value of each instance under the influence of other types of instance sets different from the instance type according to the preset distance attenuation threshold and the second-order instance neighbor relationship table;

The average influence calculation module 403 is used to calculate the average influence of each instance set on other types of instance sets according to the network kernel density value;

The popular co-location pattern acquisition module 404 is configured to calculate the popularity of each candidate co-location pattern according to the average influence, and determine the popular co-location pattern among the candidate co-location patterns according to a preset popularity threshold.

Among them, the second-order instance neighbor relationship table construction module 401 is specifically used to store the reachable distance values between a plurality of neighbor instance pairs and their corresponding instances in a two-dimensional hash table TIns_net ₂ , each in the table A cell unit is represented by a set of triplets as follows:

TIns_net ₂ (e _x ,e _y )={<o _i ,o _j ,Rdis(o _i ,o _j )>,...},

Among them, (e _x , e _y ) are two spatial object instances, o _i , o _j are two adjacent instances, and Rdis(o _i , o _j ) is the reachable distance value between two adjacent instances;

according to

Rdis(o _i ,o _j )＝g(y)*(o _i -o _j ) _net(t) |h _t

Calculating the reachable distance value between two adjacent instances, wherein Rdis(o _i , o _j ) is the reachable distance value between the two adjacent instances,

g(y) is a binary function, if the direction along the road from instance o _i to instance o _j is opposite to the direction of road traffic, the value is 0, otherwise it is 1,

(o _i -o _j ) _net(t) is the shortest path time from instance o _i to instance o _j ,

h _t is the density decay threshold based on path time, which is a constraint condition for finding the shortest path time, indicating that the reachable distance between instances must meet the h _t threshold, otherwise instance o _i to instance o _j is unreachable.

Wherein, the network kernel density calculation module 402 is specifically used to

Calculate the network kernel density value of the instance o _i of type e _x under the influence of instance set O'(e _y ) of type e _y ,

in, is a subset of all instances of type e _y on the region,

n _max is the maximum number of instances of a single type on the region,

n(O'(e _y )->o _i ) is the number of instance pairs reachable from the instance in O'(e _y ) to o _i ,

The value range of the calculation result of this formula is (0,1], which describes the influence of the instance set O'(e _y ) on the instance o _i under the network constraints;

Wherein, the average influence calculation module 403 is specifically used for

Calculate the average influence of the instance set O'(e _y ) on the instance set O'(e _x ),

in, is the network kernel density value of instance o _i of type e _x under the influence of instance set O'(e _y ) of type e _y ,

_is a subset of the set of all instances of type ex on the region,

n(O'(e _x )) is the number of instances in the instance set O'(e _x ),

n(e _x ) is the number of all instances of type e _x in the area.

Wherein, the popular parity pattern acquisition module 404 is specifically used to

Calculate the popularity of a given candidate pattern,

Among them, PI _CP is the popularity of a given candidate pattern, and the value range is (0,1],

Tins_net _CP is an instance table composed of group instances of the candidate pattern CP, which is obtained by connecting the non-repeated instance pairs related to the type in the CP in the second-order instance adjacency relation table Tins_net ₂ through group instances,

min(.) is used to calculate the minimum value of the input set,

It is used to calculate the projection of an instance of type ex on the instance table _{Tins_net CP} ,

collection of instances set of instances the average influence of

When the calculated popularity of a candidate co-location pattern is greater than a set popularity threshold, the candidate is determined to be a popular co-location pattern.

The present invention proposes a colocation mode discovery method considering the urban road network constraints, and makes two extensions on the basis of the traditional colocation mode discovery method:

(1) Based on the fact that the interconnection of urban spatial facilities occurs in network path distances rather than Euclidean distances, the method of spatial kernel functions is placed in the network structure, and specific services that consider attribute information such as urban road traffic capacity and directions are used. The time-based accessibility index replaces the traditional two-dimensional Euclidean distance to measure the proximity between spatial facilities.

(2) Transform the traditional popularity index for judging the interestingness of the model, and add the original index calculated simply by the number of instance connections into the adjustment parameter of reachable distance weight. Compared with other existing methods in this field, the present invention respects the fact that "people's movement mainly depends on the road network in the city" and "the first law of geography", thus improving the efficiency of co-location pattern mining on urban facility data. Accuracy, and more realistic and practical value.

The content described above can be implemented alone or combined in various ways, and these variants are all within the protection scope of the present invention.

Those skilled in the art can understand that all or part of the steps in the above method can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium, such as a read-only memory, a magnetic disk or an optical disk, and the like. Optionally, all or part of the steps in the above embodiments can also be implemented using one or more integrated circuits. Correspondingly, each device/unit in the above embodiments can be implemented in the form of hardware, or can be implemented in the form of software function devices. The form is realized. The present invention is not limited to any specific combination of hardware and software.

It should be noted that, in this document, the terms "comprising", "comprising" or any other variation thereof are intended to cover a non-exclusive inclusion, such that an article or device comprising a series of elements includes not only those elements, but also includes none. other elements specifically listed, or also include elements inherent in the article or equipment. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the article or device comprising said element.

The above embodiments are only used to illustrate the technical solutions of the present invention rather than limit them, and the present invention is described in detail with reference to preferred embodiments. Those skilled in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and all should be covered by the claims of the present invention.

Claims (10)

1. A method for discovering a co-location pattern considering urban road network constraints, characterized in that, the method comprises: Construct a second-order instance proximity relationship table for the target area under the map projection, and the second-order instance proximity relationship table contains a set of instance pairs of different types within the target area where the distance between all instances and their reachable distances is within the preset distance attenuation threshold and their reachable distance values; Calculate the network kernel density value of each instance under the influence of other types of instance sets different from this instance type according to the preset distance attenuation threshold and the second-order instance proximity relationship table; Calculate the average influence of each instance set on other types of instance sets according to the network kernel density value; The popularity of each candidate co-location pattern is calculated according to the average influence, and the popular co-location pattern among the candidate co-location patterns is determined according to a preset popularity threshold. 2. The method of claim 1, wherein The construction of the second-order instance neighbor relationship table includes: Store the reachable distance values between multiple adjacent instance pairs and their corresponding instances in a two-dimensional hash table TIns_net ₂ , and each cell unit in the table is expressed by a set of triples as follows: TIns_net ₂ (e _x ,e _y )={<o _i ,o _j ,Rdis(o _i ,o _j )>,...}, Among them, (e _x , e _y ) are two spatial object instances, o _i , o _j are two adjacent instances, and Rdis(o _i , o _j ) is the reachable distance value between two adjacent instances; according to Rdis(o _i ,o _j )＝g(y)*(o _i -o _j ) _net(t) |h _t Calculate the reachable distance value between two adjacent instances, wherein, Rdis(o _i , o _j ) is the reachable distance value between the two adjacent instances, g(y) is a binary function, if the instance o _i reaches The direction along the road of instance o _j is opposite to the traffic direction of the road, the value is 0, otherwise the value is 1, (o _i -o _j ) _net(t) is the shortest path time from instance o _i to instance o _j , h _t is the density attenuation threshold based on the path time, which is a constraint condition for finding the shortest path time, which means that the reachable distance between instances must meet the h _t threshold, otherwise instance o _i to instance o _j is unreachable. 3. The method of claim 1, wherein, The network kernel density value calculated according to the preset distance attenuation threshold and the second-order instance neighbor relationship table under the influence of other types of instance sets different from the implementation type includes: according to Calculate the network kernel density value of the instance o _i of type e _x under the influence of instance set O'(e _y ) of type e _y , in, is the network kernel density value of instance o _i of type e _x under the influence of instance set O'(e _y ) of type e _y , is a subset of all instances of type e _y on the region, n _max is the maximum number of instances of a single type on the region, n(O'(e _y )->o _i ) is the number of instance pairs reachable from the instance in O'(e _y ) to o _i , The value range of the calculation result of this formula is (0,1]. 4. The method of claim 1 or 3, wherein, According to the network kernel density value, the average influence of each instance set on other types of instance sets includes: according to Calculate the average influence of the instance set O'(e _y ) on the instance set O'(e _x ), in, is the average influence of instance set O'(e _y ) on instance set O'(e _x ), is the network kernel density value of instance o _i of type e _x under the influence of instance set O'(e _y ) of type e _y , _is a subset of the set of all instances of type ex on the region, n(O'(e _x )) is the number of instances in the instance set O'(e _x ), n(e _x ) is the number of all instances of type e _x in the region. 5. The method of claim 1, wherein, The calculation of the popularity of each candidate co-location pattern according to the average influence, and determining the popular co-location pattern among the candidate co-location patterns according to the preset popularity threshold include: according to Calculate the popularity of a given candidate pattern, Among them, PI _CP is the popularity of a given candidate pattern, and the value range is (0,1], Tins_net _CP is an instance table composed of group instances of the candidate pattern CP, which is obtained by connecting the non-repeated instance pairs related to the type in the CP in the second-order instance adjacency relation table Tins_net ₂ through group instances, min(.) is used to calculate the minimum value of the input set, It is used to calculate the projection of an instance of type ex on the instance table _{Tins_net CP} , collection of instances set of instances the average influence of When the calculated popularity of a candidate co-location pattern is greater than a set popularity threshold, the candidate is determined to be a popular co-location pattern. 6. A device for discovering a co-located pattern considering urban road network constraints, characterized in that the device comprises: The second-order instance proximity relationship table construction module is used to construct a second-order instance proximity relationship table for the target area under the map projection, and the second-order instance proximity relationship table includes all instance distances and their reachable distances in the target area within a preset distance A set of instance pairs of different types within the distance decay threshold and the reachable distance between them; The network kernel density calculation module is used to calculate the network kernel density value of each instance under the influence of other types of instance sets different from this instance type according to the preset distance attenuation threshold and the second-order instance proximity relationship table; The average influence calculation module is used to calculate the average influence of each instance set on other types of instance sets according to the network kernel density value; The popular co-location pattern acquisition module is configured to calculate the popularity of each candidate co-location pattern according to the average influence, and determine the popular co-location pattern among the candidate co-location patterns according to a preset popularity threshold. 7. The apparatus of claim 6, wherein The second-order instance neighbor relationship table construction module is specifically used to store a plurality of neighbor instance pairs and the reachable distance values between the corresponding instances in a two-dimensional hash table TIns_net ₂ , and each A cell unit, represented by a set of triples as follows: TIns_net ₂ (e _x ,e _y )={<o _i ,o _j ,Rdis(o _i ,o _j )>,...}, Among them, (e _x , e _y ) are two spatial object instances, o _i , o _j are two adjacent instances, and Rdis(o _i , o _j ) is the reachable distance value between two adjacent instances; according to Rdis(o _i ,o _j )＝g(y)*(o _i -o _j ) _net(t) |h _t Calculating the reachable distance value between two adjacent instances, wherein Rdis(o _i , o _j ) is the reachable distance value between the two adjacent instances, g(y) is a binary function, if the direction along the road from instance o _i to instance o _j is opposite to the direction of road traffic, the value is 0, otherwise it is 1, (o _i -o _j ) _net(t) is the shortest path time from instance o _i to instance o _j , h _t is the density decay threshold based on path time, which is a constraint condition for finding the shortest path time, indicating that the reachable distance between instances must meet the h _t threshold, otherwise instance o _i to instance o _j is unreachable. 8. The apparatus of claim 6, wherein The network kernel density calculation module is specifically used according to Calculate the network kernel density value of the instance o _i of type e _x under the influence of instance set O'(e _y ) of type e _y , in, is a subset of all instances of type e _y on the region, n _max is the maximum number of instances of a single type on the region, n(O'(e _y )->o _i ) is the number of instance pairs reachable from the instance in O'(e _y ) to o _i , The value range of the calculation result of this formula is (0,1], which describes the influence of the instance set O'(e _y ) on the instance o _i under the network constraints. 9. The apparatus of claim 6, wherein The average influence calculation module is specifically used according to Calculate the average influence of the instance set O'(e _y ) on the instance set O'(e _x ), in, is the network kernel density value of instance o _i of type e _x under the influence of instance set O'(e _y ) of type e _y , _is a subset of the set of all instances of type ex on the region, n(O'(e _x )) is the number of instances in the instance set O'(e _x ), n(e _x ) is the number of all instances of type e _x in the region. 10. The device according to claim 6, wherein the popular parity pattern acquisition module is specifically configured to Calculate the popularity of a given candidate pattern, Among them, PI _CP is the popularity of a given candidate pattern, and the value range is (0,1], Tins_net _CP is an instance table composed of group instances of the candidate pattern CP, which is obtained by connecting the non-repeated instance pairs related to the type in the CP in the second-order instance adjacency relation table Tins_net ₂ through group instances, min(.) is used to calculate the minimum value of the input set, It is used to calculate the projection of an instance of type ex on the instance table _{Tins_net CP} , collection of instances set of instances the average influence of When the calculated popularity of a candidate co-location pattern is greater than a set popularity threshold, the candidate is determined to be a popular co-location pattern.