WO2019179173A1 - Trading area determination method and device - Google Patents

Abstract

Provided in embodiments of the present description are a method and device for training a trading area determination model, a method and device for determining a trading area, and a method and device for updating trading area determination, wherein the method for training a trading area determination model comprises: acquiring location information respective to a plurality of shops in a predetermined geographical range and trading area labeling information respective to the plurality of shops; according to a CFSFDP clustering algorithm, utilizing the location information and calculating the value of a local density ρ of each shop, the value of a minimum distance δ to a shop that has a higher density, and the value of a product γ thereof; acquiring trading area determination information of each shop according to current thresholds respective to ρ, δ and γ; utilizing the trading area determination information and trading area labeling information respective to the plurality of shops, and calculating the degree of similarity between all of the trading area determination information relative to all of the trading area labeling information; and adjusting the thresholds respective to ρ, δ and γ such that the degree of similarity is increased.

Description

Method and device for determining business circle Technical field

The embodiments of the present specification relate to the field of machine learning, and more particularly, to a method and apparatus for training a business circle determination model, a method and apparatus for determining a business circle, and a method and apparatus for updating a business circle determination.

Background technique

In recent years, under the background of “new retail” and “new finance”, offline stores have been vigorously developed. Compared with online merchants, offline merchants have physical stores, and objects are clustered. They have a certain concentration in geographical location, that is, they can be divided into various business districts. Through the business circle information, you can deepen your understanding of the store: identify the offline industry market, and assist in judging the operation status of the store. The existing business circle information includes: business circle information marked by offline BD, and business circle information mainly from the results of the popular public comment business circle. The above business circle information is obtained by manual labeling. Therefore, there is a need for a more effective solution for determining a business circle.

Summary of the invention

The embodiments of the present specification aim to provide a more effective solution for determining a business circle to solve the deficiencies in the prior art.

In order to achieve the above object, an aspect of the present specification provides a method for training a business circle determination model, comprising: acquiring respective location information of a plurality of stores within a predetermined geographical range and respective business circle annotation information of the plurality of stores; according to CFSFDP a clustering algorithm, using the position information, calculating a value of a local density ρ of each store, a value of a minimum distance δ with a higher density store, and a value of a product γ; acquiring according to respective current thresholds of ρ, δ, and γ Business circle determination information of each store; calculating the similarity of all the business circle determination information with respect to all the business circle label information by using the business circle determination information and the business circle label information of the plurality of stores; and adjusting ρ The respective thresholds of δ and γ increase the similarity.

In one embodiment, in the method for training a business circle determination model, a value of a local density ρ of a shop i in the plurality of stores is ρ _i , wherein

ρ _i =∑ _j χ(d _ij -d _c ),

Wherein, when d _ij -d _c <0, χ(d _ij -d _c )=1, when d _ij -d _c ≥0, χ(d _ij -d _c )=0,

Where d _c is a radius threshold, d _ij is the distance between the store i and the store j in the plurality of stores, and i and j are natural numbers less than or equal to the total number of stores of the plurality of stores, and i ≠ j.

In one embodiment, in the method for training a business circle determination model, a value of a local density ρ of a shop i in the plurality of stores is ρ _i , wherein

Where d _c is a radius threshold, d _ij is the distance between the store i and the store j in the plurality of stores, and i and j are natural numbers less than or equal to the total number of stores of the plurality of stores, and i ≠ j.

In one embodiment, in the method of training a business circle to determine a model,

The position information of the shop i and the shop j are expressed as latitude and longitude (Lon _i , Lat _i ) and (Lon _j , Lat _j ), respectively, and the distance d _ij is calculated as follows:

Where R is the radius of the Earth.

In an embodiment, in the method for training a business circle determination model, the similarity is represented by a parameter WFS, wherein

Where i is an integer from 0 to A, j is an integer from 0 to B, A is the number of quotients in the circle, B is the number of quotients in the quotient, and N _i is the number of stores in the ith circle , N is the total number of stores of the plurality of stores,

P _ij is the accuracy rate for the i-th labeled business circle and the j-th determined business circle, and R _ij is the recall rate for the i-th labeled business circle and the j-th determined business circle, which will include a collection of labeled scattered stores Set to the 0th labeled business circle, and set the set of the determined scattered stores as the 0th determination business circle, wherein the marked scattered shop is an annotated shop that does not belong to any labeled business circle, and the determined scattered shop is A judgment shop that does not belong to any judgment business circle.

In one embodiment, in the method for training a business circle determination model, adjusting respective threshold values of ρ, δ, and γ such that the similarity improvement includes adjusting respective threshold values of ρ, δ, and γ such that the similarity The greatest degree.

Another aspect of the present specification provides a method for determining a business circle, comprising: acquiring respective location information of a plurality of stores within a predetermined geographical range; and calculating a local density ρ of each store by using the location information according to a CFSFDP clustering algorithm. a value, a value of a minimum distance δ with a higher density store, and a value of the product γ; for each of the plurality of stores by the adjusted threshold values of ρ, δ, and γ obtained by the method of training the business circle determination model Determine the business circle.

In one embodiment, in the method of determining a business circle, the predetermined geographic range is a predetermined city.

Another aspect of the present specification provides a method for updating a business circle determination, comprising: acquiring respective first location information of a plurality of first stores within a predetermined geographic range and a first distance between each first store; acquiring the predetermined Second position information of each of the at least one second store in the geographical range; using the first location information and the second location information, calculating a second distance between the second stores, and any one of the second stores a third distance between any one of the first stores; calculating, according to the CFSFDP clustering algorithm, the plurality of first stores and the at least one second store based on the first distance, the second distance, and the third distance The value of the local density ρ, the value of the minimum distance δ with the higher density store, and the value of the product γ; and the adjusted threshold values of ρ, δ, and γ obtained by the method of training the business circle determination model, A business circle is determined for the plurality of first stores and at least one second store.

In one embodiment, the method of updating the business circle determination is performed every predetermined time period.

In another aspect, the present disclosure provides an apparatus for training a business circle determination model, including: a first acquisition unit configured to acquire respective location information of a plurality of stores within a predetermined geographic range and respective business circle labels of the plurality of stores The first calculation unit is configured to calculate, according to the CFSFDP clustering algorithm, a value of a local density ρ of each store, a value of a minimum distance δ with a higher density store, and a value of a product γ using the position information; The second obtaining unit is configured to acquire the business circle determination information of each store according to the current threshold value of each of ρ, δ, and γ; the second calculating unit is configured to determine the information and the quotient by using the respective business circle of the plurality of stores The circle labeling information calculates a similarity of all the merchant circle determination information with respect to all of the merchant circle labeling information; and the threshold value adjusting unit is configured to adjust respective threshold values of ρ, δ, and γ such that the similarity is improved.

Another aspect of the present disclosure provides an apparatus for determining a business circle, comprising: an obtaining unit configured to acquire respective location information of a plurality of stores within a predetermined geographical range; and a calculating unit configured to use the CFSFDP clustering algorithm according to the CFSFDP clustering algorithm The position information is calculated, the value of the local density ρ of each store, the value of the minimum distance δ with the higher density store, and the value of the product γ are calculated; and the determining unit is configured to obtain by the method according to the above-described training quotation determination model The adjusted threshold values of ρ, δ, and γ are each determined for the plurality of stores.

In another aspect, the present disclosure provides an apparatus for updating a business circle determination, including: a first obtaining unit configured to acquire first location information of each of a plurality of first stores within a predetermined geographic range and between respective first stores a first distance unit, configured to acquire respective second location information of the at least one second store in the predetermined geographic range; the first calculating unit is configured to utilize the first location information and the second Position information, calculating a second distance between the respective second stores, a third distance between any one of the second stores and any one of the first stores; and a second calculating unit configured to cluster according to CFSFDP And calculating, based on the first distance, the second distance, and the third distance, a value of a local density ρ of each of the plurality of first stores and at least one second store, and a value of a minimum distance δ from a higher density store And a value of the product γ; and a determining unit configured to adjust the threshold values of ρ, δ, and γ obtained by the method of training the business circle determination model for the plurality of And at least one second store stores determination district.

By determining the business circle scheme according to the embodiment of the present specification, the business circle can be quickly and accurately determined, and the stability of the determination result can be ensured. Moreover, the embodiments of the present specification also effectively reduce the computational complexity and optimize the calculation time.

DRAWINGS

The embodiments of the present specification can be more clearly understood by describing the embodiments of the specification with reference to the accompanying drawings:

FIG. 1 shows a schematic diagram of a system 100 for determining a business circle in accordance with an embodiment of the present specification;

2 is a flow chart showing a method of training a business circle determination model according to an embodiment of the present specification;

Figure 3 schematically shows an example of a ρ _- δ profile;

Figure 4 schematically shows an example of a gamma distribution map;

FIG. 5 is a flowchart showing a method of determining a business circle according to an embodiment of the present specification;

6 shows a flow chart of a method for updating a business circle determination according to an embodiment of the present specification;

Figure 7 illustrates an apparatus 700 for training a business circle decision model in accordance with an embodiment of the present specification;

Figure 8 illustrates an apparatus 800 for determining a business circle in accordance with an embodiment of the present specification;

FIG. 9 illustrates an apparatus 900 for updating a business circle determination in accordance with an embodiment of the present specification.

detailed description

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

FIG. 1 shows a schematic diagram of a system 100 for determining a business circle in accordance with an embodiment of the present specification. As shown in FIG. 1, system 100 includes a clustering module 11, an evaluation module 12, and a threshold adjustment module 13. In the training phase, the training samples are input to the clustering module 11. The training sample includes the location information of the store and the respective business circle labeling information of the store. The clustering module 11 calculates the value of the local density ρ of each store, the value of the minimum distance δ with the higher density store, and the value of the product γ according to the CFSFDP clustering algorithm according to the position information, according to ρ, δ, and γ. The respective current thresholds are used to obtain the business circle determination information of each store. Thereafter, the clustering module 11 transmits the above-mentioned business circle determination information to the evaluation module 12. The evaluation module 12 calculates the similarity of all the business circle determination information with respect to all the business circle annotation information as the evaluation score by using the respective business circle determination information and the business circle annotation information of the plurality of stores, and the evaluation score is It is transmitted to the threshold adjustment module 13. The threshold adjustment module 13 adjusts the thresholds of the parameters ρ, δ, and γ in the clustering module 11 according to the evaluation score to increase the evaluation score, and after the plurality of adjustments, maximizes the evaluation score. After the training of the clustering module 11 reaches its optimal parameter threshold, the full amount of store information can be clustered by the distance module 11 to obtain the business circle determination result.

2 shows a flow chart of a method of training a business circle decision model in accordance with an embodiment of the present specification. As shown in FIG. 2, the method includes: acquiring, in step S21, location information of each of a plurality of stores in a predetermined geographical range and respective business circle annotation information of the plurality of stores; and in step S22, according to a CFSFDP clustering algorithm Using the position information, calculating the value of the local density ρ of each store, the value of the minimum distance δ with the higher density store, and the value of the product γ; in step S23, according to the respective threshold values of ρ, δ, and γ, Acquiring the business circle determination information of each store; in step S24, calculating the similarity of all the business circle determination information with respect to all the business circle annotation information by using the business circle determination information and the business circle annotation information of each of the plurality of stores And in step S25, the respective thresholds of ρ, δ, and γ are adjusted such that the similarity is improved.

First, in step S21, location information of each of a plurality of stores within a predetermined geographical range and business circle annotation information of each of the plurality of stores are acquired. The predetermined geographic extent may be, for example, a geographic extent including more than 100 business districts, such as a district, county, etc. of the city. For the accuracy of the parameters, the number of stores in a plurality of stores may be, for example, on the order of several thousand, for example, 3,000. Preferably, the plurality of shopping districts covered by the plurality of stores include a plurality of location relationships, for example, the business circle is adjacent to the business circle, intersects, is away from, and the like. The location information of the store may be expressed in various known forms. For example, the location information of the store may be the latitude and longitude of the store, or the location information of the store may be the city coordinates or the like. The business district labeling information of the store includes whether the store belongs to a certain business circle, and which business district the store belongs to. For example, the business circle label information of the store may be indicated by the labeled business circle field. When the field is 0, the store is a scattered store that does not belong to any business circle. When the field is a natural number, it indicates that the store belongs to the natural number. Business district.

In step S22, based on the CFSFDP clustering algorithm, the value of the local density ρ of each store, the value of the minimum distance δ with the higher density store, and the value of the product γ are calculated using the position information.

First, the calculation of the value of the parameter local density ρ in the CFSFDP algorithm used in the embodiment of the present specification will be explained.

In one embodiment, the value of the local density ρ of the shop i in the plurality of stores is ρ _i , wherein ρ _{i is} calculated by the following formula (1):

ρ _i =∑ _j χ(d _ij -d _c ) (1)

Wherein, when d _ij -d _c <time _{0, χ (d ij -d c} ) = 1, when _{d ij -d c ≥0, χ (} d ij -d c) = 0, where d _c is the radius of the threshold d _ij is the distance between the shop i and the shop j among the plurality of stores, and i and j are natural numbers smaller than or equal to the total number of stores of the plurality of stores, and i ≠ j.

In the embodiment of the present specification, the CFSFDP clustering algorithm is used to determine the business circle, that is, clustering the shop points. Since the shape of the business circle is generally fixed, d _c is set to 0.2 (ie, 200 m). With this setting, the stability of the clustering result is greatly improved. The distance d _ij between the store i and the store j may adopt different formulas according to different forms of the store location information. For example, when the position information of the shop i and the shop j are expressed as latitude and longitude (Lon _i , Lat _i ) and (Lon _j , Lat _j ), respectively, the distance d _{ij is} calculated by the following formula (2):

The d _ij calculated by the formula (2) is the distance between two points on the spherical surface, where R is the radius of the earth, and the average value is 6371 km. In another example, d _ij can be Euclidean distance, Minkowsky distance, Manhattan distance, and the like. In another example, the location information of the store i and the store j are respectively represented in three-dimensional or two-dimensional coordinates in the city coordinate system, so that, for example, the Euclidean distance between the store i and the store j can be calculated as d _ij . In the calculation, each d _ij can be calculated by, for example, the above formula (2), thereby obtaining a distance matrix.

According to the formula (1), when the distance between the shop i and the shop j is smaller than d _c (ie, 0.2 (km)), that is, when the shop j is within the radius of 200 m of the shop i, χ (d _ij -d _c The value of ") is 1, and when the shop j is outside the 200m radius of the shop i, the value of χ(d _ij -d _c ) is 0. That is to say, ρ _i here is the number of stores within a certain radius (200 m) from the shop i.

In another preferred embodiment, a distance calculation formula based on a Gaussian kernel function is introduced, and ρ _i is calculated by the following formula (3),

Where d _c is a radius threshold, d _ij is the distance between the store i and the store j in the plurality of stores, and i and j are natural numbers less than or equal to the total number of stores of the plurality of stores, and i ≠ j. Where d _c and d _ij can be obtained in the same manner as the same parameters in the above formula (1). The local density ρ _i calculated by the formula (3) represents a function of the ratio of d _ij (d _ij <d _c ) to d _c , which can be regarded as the sum of the distance gains of the shops in the range of 200 m of the shop i . The local density of the store is more accurately and accurately described by the formula (3).

Next, the calculation of the value of the parameter in the CFSFDP algorithm used in the embodiment of the present specification and the minimum distance δ of the higher density shop will be explained. Store the value [delta] i may be represented as δ _i, can (4) [delta] _i is calculated by the following equation:

From equation (4), it can be concluded that when ρ _i is the highest density shop point, δ _i is the maximum value of d _ij (where j is any store other than i), when ρ _i is not the highest density At the shop point, δ _i is equal to the minimum distance between the shop i and the higher density store.

After determining ρ _i and δ _i for each store point, the center of the quotient can be determined by plotting the ρ _- δ profile. Fig. 3 shows an example of a ρ _- δ profile. As shown in Fig. 3, the abscissa in the ρ _- δ distribution map is ρ and the ordinate is δ. As known to those skilled in the art, in the CFSFDP algorithm, the quasi-center has a higher local density value ρ and a higher high-density distance δ, and therefore, the points in the upper right portion of the graph are located in the ρ _- δ distribution map. May be the cluster center. For example, in the ρ _- δ distribution map shown in FIG. 3, points of different gradations other than black dots may be the center of the quotient circle.

In addition, after determining ρ _i and δ _{i of} each store point, the product γ _i = ρ _i * δ _i can also be calculated, and then γ _{i of} each store can be sorted, and each γ distribution map is γ _i draws the γ distribution map in descending order. Fig. 4 schematically shows an example of a gamma distribution map. As shown in FIG. 4, the ordinate is γ, and the abscissa is the serial number of the shop i sorted according to the γ _i size, wherein each serial number corresponds to one store. The larger γ, the larger the representative ρ*δ, that is, the shop corresponding to the point becomes more likely to become the center of the business circle.

Returning to Fig. 2, in step S23, the business circle determination information of each store is acquired based on the current threshold values of ρ, δ, and γ.

In the embodiment of the present specification, by combining, for example, the ρ _- δ distribution map of FIG. 3 and the γ distribution map of FIG. 4, for example, the shop that becomes the center of the business circle is determined based on the threshold values of the set ρ, δ, and γ. For example, in FIG. 3, a broken line perpendicular to the ρ axis represents a threshold of ρ, and a broken line perpendicular to the δ axis represents a threshold of δ, and in FIG. 4, a broken line perpendicular to the γ axis represents a threshold of γ. After drawing the threshold line, in FIG. 3, the point in the upper right of the intersection of the ρ threshold line and the δ threshold line is the center of the first quotient circle, and in FIG. 4, the point above the γ threshold line is the second quotient circle. center. Referring to FIG. 3 and FIG. 4, the intersection of the first business circle center determined according to FIG. 3 and the second business circle center determined according to FIG. 4 is taken, thereby obtaining the store which is finally the center of the business circle.

After acquiring the business circle center, each of the plurality of stores is clustered, that is, divided into a certain business circle. Specifically, in one embodiment, each store is classified into a business circle to which the store point closest to it and whose density is higher than it belongs. In one embodiment, each store is categorized into the business district of the nearest business district center. When the store is too far away from any store with a higher density than the store, or if the store is too far away from the center of any commercial circle, for example, if it is more than 2km, the store may be considered as a scattered store and does not belong to any business district. Or the store may be considered to be a business district with a business circle identification number of zero. Thereby, the business circle determination information of each store is obtained.

In step S24, the similarity of all the business circle determination information with respect to all the business circle label information is calculated by using the business circle determination information and the business circle label information of the plurality of stores.

The similarity is a degree indicating that the business circle determination information is similar to the business circle annotation information. The similarity can be calculated in various forms to evaluate the judgment result. For example, Precision, Recall, AUC score, log loss, Accuracy, etc. can all be used to indicate similarity.

In an embodiment of the present specification, the similarity is expressed by a WFS score, wherein the WFS is calculated by the formula (5),

Where f _{ij is} calculated by the formula (6),

Where i is an integer from 0 to A, j is an integer from 0 to B, A is the number of quotients in the circle, B is the number of quotients in the quotient, and N _i is the number of stores in the ith circle , N is the total number of stores of the plurality of stores,

P _ij is the accuracy rate for the i-th labeled business circle and the j-th determined business circle, and R _ij is the recall rate for the i-th labeled business circle and the j-th determined business circle, which will include a collection of labeled scattered stores Set to the 0th labeled business circle, and set the set of the determined scattered stores as the 0th determination business circle, wherein the marked scattered shop is an annotated shop that does not belong to any labeled business circle, and the determined scattered shop is A judgment shop that does not belong to any judgment business circle.

Specifically, P _ij can be calculated by the following formula (7), and R _ij can be calculated by the following formula (8):

Where i is an integer from 0 to A, j is an integer from 0 to B, A is the number of quotients in the labeled business circle, and B is the number of quotients in the determination of the business circle. x _ij is the number of stores in the i-th labeled business circle that are assigned to the j-th decision business circle. Wherein, it is assumed that the set including the scattered mark shop is the 0th labeled business circle (ie, i=0), and it is assumed that the set including the scattered judgment shop is the 0th decision business circle (ie, j=0), so as to pass the above Formulas (7) and (8) are counted in the scattered stores.

In step S25, the respective threshold values of ρ, δ, and γ are adjusted such that the similarity is improved.

Referring again to FIGS. 3 and 4, the at least one threshold line may be moved above and below at least one threshold line of the three threshold lines in FIGS. 3 and 4 to obtain an adjusted threshold. After the movement is performed, similarly to the above-described steps S23 and S24, the adjusted business circle determination information of each store is acquired based on the adjusted threshold values of ρ, δ, and γ, and the information and adjustment are marked by the business circle. The subsequent business circle determination information calculates the adjusted similarity. The threshold line may be moved a plurality of times in a direction in which the similarity becomes large according to the change in the similarity, so that the respective threshold values of ρ, δ, and γ are continuously adjusted, so that the similarity is continuously increased. In one embodiment, the respective thresholds of ρ, δ, and γ are adjusted such that the similarity is maximized, thereby obtaining respective thresholds of ρ, δ, and γ.

5 is a flowchart of a method for determining a business circle according to an embodiment of the present specification, including the steps of: acquiring, in step S51, respective location information of a plurality of stores within a predetermined geographic range; and in step S52, according to a CFSFDP clustering algorithm Using the position information, calculating the value of the local density ρ of each store, the value of the minimum distance δ with the higher density store, and the value of the product γ; and, in step S53, by determining the model according to the training circle described above The adjusted threshold values of ρ, δ, and γ obtained by the method determine a business circle for the plurality of stores.

In the method shown in FIG. 5, the predetermined geographic range may be a predetermined city range, that is, the business circle is determined in units of cities. The process of the step S52 is substantially the same as the step S22 of FIG. 2, and the process of the step S53 is substantially the same as the step S23 of FIG. 2, and details are not described herein again.

6 is a flowchart of a method for updating a business circle determination according to an embodiment of the present specification, including: in step S61, acquiring first location information of each of a plurality of first stores within a predetermined geographic range and each first store a first distance between the two; at step S62, acquiring second location information of each of the at least one second store within the predetermined geographic range; and in step S63, calculating the location by using the first location information and the second location information a second distance between each of the second stores, a third distance between any one of the second stores and any of the first stores; and in step S64, based on the CFSFDP clustering algorithm, based on the first distance, a second distance and a third distance, calculating a value of a local density ρ of each of the plurality of first stores and at least one second store, a value of a minimum distance δ with a higher density store, and a value of a product γ; Step S65, determining a business circle for the plurality of first stores and at least one second store by adjusting the respective thresholds of ρ, δ, and γ obtained by the method according to any one of claims 1-6.

The method shown in Figure 6 is an incremental iterative approach. As offline merchants continue to expand and the number of stores continues to expand, the direct calculation of the distance matrix will face the computational complexity of O(N ² ). Therefore, by the method shown in FIG. 6, the amount of calculation is reduced to speed up the calculation.

Specifically, first, in step S61, first location information of each of the plurality of first stores within the predetermined geographic range and a first distance between the respective first stores are acquired. The predetermined geographic extent may be, for example, a predetermined city. In one example, location information of a plurality of stores may be acquired in the initial month M ₀ , and the distance between the respective stores may be calculated as described above, thereby acquiring the distance matrix N ₀ . And in a month M ₀ M _1, add at least one store or multiple stores already have the location information has been changed at the shop. It is possible to record the newly added store or the store whose location has been changed as the second store, or x _new , and the store that is not related to the second store among the plurality of stores is the first store, or x _old . For example, when the second store is a new store, the first store to acquire all shops M ₀ months. In another example, when the shop has been changed to the second store stores position occurs, the first store is removed from the position changing all stores acquired in the rest of the store M ₀ month shop.

When M _January determination to update the business district, a first plurality of available store respective first and the respective position information of the first store from the distance matrix N ₀ store position information acquired and calculated in the M ₀ months The first distance between the first distance is the distance between the store x _old and the _old .

In step S62, second location information of each of the at least one second store within the predetermined geographic range is acquired. In the example described above, when at least one second store is added in the next month M _{1 of} M ₀ , the location information of the newly added second store is acquired. Alternatively, when the second store location information in M _January, has changed in the shops, to obtain location information after the change of the second shop.

In step S63, using the first location information and the second location information, calculating a second distance between the second stores, and a third between any of the second stores and any of the first stores distance. That is, using the first position information and second position information, calculating a third distance between the second and the distance between the _new x x x _{new x-new} and _old.

In step S64, according to the CFSFDP clustering algorithm, calculating a value of a local density ρ of each of the plurality of first stores and at least one second store based on the first distance, the second distance, and the third distance, and higher The value of the minimum distance δ of the density store and the value of the product γ. The first distance, the second distance, and the third distance together form a new distance matrix, so that the value of the local density ρ of the plurality of stores including the first store and the second store, and the higher density store can be calculated as described above The value of the minimum distance δ and the value of its product γ.

Finally, in step S65, the business circle is determined for the plurality of first stores and the at least one second store by the adjusted threshold values of ρ, δ, and γ obtained by the method of training the business circle determination model. This step is basically the same as step S23 in FIG. 2 and step S53 in FIG. 5, and details are not described herein again.

The above method of updating the business circle determination may be performed once every predetermined time period, for example, once a month, so that the determination of the business circle may be periodically updated. And the update method reduces the computational complexity of at least two orders of magnitude.

FIG. 7 shows an apparatus 700 for training a business circle determination model according to an embodiment of the present specification, including: a first acquisition unit 71 configured to acquire respective location information of a plurality of stores within a predetermined geographic range and the plurality of The respective business circle label information of the stores; the first calculating unit 72 is configured to calculate the value of the local density ρ of each store and the minimum distance δ with the higher density store by using the position information according to the CFSFDP clustering algorithm. a value and a value of the product γ; the second obtaining unit 73 is configured to acquire the business circle determination information of each store according to the current threshold of each of ρ, δ, and γ; the second calculating unit 74 is configured to utilize the plurality of The business district determination information and the business circle annotation information of each store, calculating the similarity of all the business circle determination information with respect to all the business circle annotation information; and the threshold adjustment unit 75 configured to adjust ρ, δ, and γ The respective thresholds increase the similarity.

FIG. 8 shows an apparatus 800 for determining a business circle according to an embodiment of the present specification, comprising: an obtaining unit 81 configured to acquire respective location information of a plurality of stores within a predetermined geographical range; and a calculating unit 82 configured to Calculating, according to the CFSFDP clustering algorithm, the value of the local density ρ of each store, the value of the minimum distance δ with the higher density store, and the value of the product γ; and the determining unit 83 configured to pass The adjusted threshold values of ρ, δ, and γ obtained by the method for training the business circle determination model are determined for the plurality of stores.

FIG. 9 illustrates an apparatus 900 for updating a business circle determination, including: a first obtaining unit 91 configured to acquire first location information of each of a plurality of first stores within a predetermined geographic range and between respective first stores The first obtaining unit 92 is configured to acquire second location information of each of the at least one second store in the predetermined geographic range; the first calculating unit 93 is configured to use the first location information And the second location information, calculating a second distance between the second stores, a third distance between any one of the second stores and any of the first stores; and the second calculating unit 94 is configured to: Calculating, according to the CFSFDP clustering algorithm, a value of a local density ρ of each of the plurality of first stores and at least one second store based on the first distance, the second distance, and the third distance, and a minimum value of the higher density store a value of the distance δ and a value of the product γ thereof; and a determining unit 95 configured to adjust the threshold values of ρ, δ, and γ obtained by the method of training the business circle determination model for the plurality of first And at least a second shop shop determination shopping district.

The quotient determination model according to an embodiment of the present specification can be evaluated by calculating a Sil score. The SIL score can be calculated by the following formulas (9)-(11):

a(i)=avg(d _ij ), x _i , x _j ∈c _k (10),

b(i)=avg(d _ij ), x _i ∈c _k ,x _j ∈c _p (11),

Where c _k represents the set of the kth clustering result, a(i) represents the average distance of point i to all points in the circle, and b(i) represents the average distance of point i to all points in the closest quotient circle p. Assuming that the sizes of the respective business districts are the same, the point corresponding to the s(i) value at the boundary of the two business districts is 0. Therefore, when the overall coefficient is evaluated, the proportion of stores with statistics s(i) greater than 0 is the number of valid classified stores. The percentage as the final SIL score. Through the SIL score evaluation, it was verified that the quotient circle judgment model according to the embodiment of the present specification has better judgment performance.

The method of the embodiment of the present specification only needs to obtain the geographical location information of the full amount of shops at the input end, and can determine the business circle for it without the need for manual one-to-one determination. The coverage of the business circle by the measured method can reach 92.5%, wherein Covered stores are basically isolated points or dirty data points.

The CFSFDP algorithm used in the method of the embodiment of the present specification does not need to be defined in advance, but directly obtains the circle of the business circle by the method of threshold definition.

The method of the embodiment of the present specification enhances stability from two aspects: firstly, the optimal parameter is pre-trained by using the known high accuracy rate labeling business circle information to ensure the stability of the parameter; secondly, in the case of stable parameters, Using the threshold-defined way to obtain the business circle center can ensure that the business circle finds stable results when the data is unchanged or the change is small.

In addition, the method of the embodiment of the present specification introduces a distance matrix based on urban partitioning and incremental iteration, and utilizes the time series of store evolution, which effectively reduces the computational complexity, and the calculation time in the actual measurement is optimized by about 10 times.

Those of ordinary skill in the art should further appreciate that the elements and algorithm steps of the various examples described in connection with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both, in order to clearly illustrate the hardware. Interchangeability with software, the components and steps of the various examples have been generally described in terms of functionality in the above description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the solution. Different methods may be used to implement the described functionality for each particular application, but such implementation should not be considered to be beyond the scope of the application.

The steps of a method or algorithm described in connection with the embodiments disclosed herein may be implemented in hardware, in a software module in a processor orbit, or in a combination of the two. The software module can be placed in random access memory (RAM), memory, read only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or technical field. Any other form of storage medium known.

The specific embodiments of the present invention have been described in detail with reference to the preferred embodiments of the present invention. All modifications, equivalent substitutions, improvements, etc., made within the spirit and scope of the invention are intended to be included within the scope of the invention.

Claims (20)

1. A method of training a business circle decision model, comprising:

   Obtaining respective location information of a plurality of stores within a predetermined geographical range and respective business circle annotation information of the plurality of stores;

   Calculating, according to the CFSFDP clustering algorithm, the value of the local density ρ of each store, the value of the minimum distance δ with the higher density store, and the value of the product γ;

   Obtaining the business circle determination information of each store according to the current threshold values of ρ, δ, and γ;

   Calculating, by using the business circle determination information and the business circle annotation information of each of the plurality of stores, the similarity of all the business circle determination information with respect to all the business circle annotation information;

   The respective thresholds of ρ, δ, and γ are adjusted such that the similarity is improved.
2. The method of training a business circle determination model according to claim 1, wherein a value of a local density ρ of a shop i in the plurality of stores is ρ _i , wherein

   ρ _i =∑ _j χ(d _ij -d _c ),

   Wherein, when d _ij -d _c <0, χ(d _ij -d _c )=1, when d _ij -d _c ≥0, χ(d _ij -d _c )=0,

   d _c is a radius threshold, d _ij is the distance between the store i and the store j in the plurality of stores, and i and j are natural numbers less than or equal to the total number of stores of the plurality of stores, and i ≠ j.
3. The method of training a business circle determination model according to claim 1, wherein a value of a local density ρ of a shop i in the plurality of stores is ρ _i , wherein

   

   Where d _c is a radius threshold, d _ij is the distance between the store i and the store j in the plurality of stores, and i and j are natural numbers less than or equal to the total number of stores of the plurality of stores, and i≠j .
4. A method of training a business circle determination model according to claim 2 or 3, wherein

   The position information of the shop i and the shop j are expressed as latitude and longitude (Lon _i , Lat _i ) and (Lon _j , Lat _j ), respectively, and the distance d _ij is calculated as follows:

   

   Where R is the radius of the Earth.
5. The method of training a business circle determination model according to claim 1, wherein the similarity is represented by a parameter WFS, wherein

   

   

   Where i is an integer from 0 to A, j is an integer from 0 to B, A is the number of quotients in the labeled business circle, B is the number of quotients in the determination of the business circle, and N _i is the store included in the i-th labeled business circle Number, N is the total number of stores in the plurality of stores,

   P _ij is the accuracy rate for the i-th labeled business circle and the j-th determined business circle, and R _ij is the recall rate for the i-th labeled business circle and the j-th determined business circle, which will include a collection of labeled scattered stores Set to the 0th labeled business circle, and set the set of the determined scattered stores as the 0th determination business circle, wherein the marked scattered shop is an annotated shop that does not belong to any labeled business circle, and the determined scattered shop is A judgment shop that does not belong to any judgment business circle.
6. The method for training a business circle determination model according to claim 1, wherein adjusting respective threshold values of ρ, δ, and γ such that said similarity improvement comprises adjusting respective threshold values of ρ, δ, and γ such that said similarity is maximum .
7. A method of determining a business circle, including:

   Obtaining respective location information of a plurality of stores within a predetermined geographical range;

   Calculating, according to the CFSFDP clustering algorithm, the value of the local density ρ of each store, the value of the minimum distance δ with the higher density store, and the value of the product γ;

   The business circle is determined for the plurality of stores by the adjusted threshold values of ρ, δ, and γ obtained by the method according to any one of claims 1-6.
8. The method of determining a business circle according to claim 7, wherein said predetermined geographic range is a predetermined city.
9. A method for updating a business circle determination, comprising:

   Obtaining respective first location information of the plurality of first stores within the predetermined geographic range and a first distance between the respective first stores;

   Obtaining second location information of each of the at least one second store within the predetermined geographic range;

   Using the first location information and the second location information, calculating a second distance between the second stores, a third distance between any one of the second stores and any of the first stores;

   Calculating, according to the CFSFDP clustering algorithm, a value of a local density ρ of each of the plurality of first stores and at least one second store based on the first distance, the second distance, and the third distance, and a minimum value of the higher density store The value of the distance δ and the value of its product γ;

   The business circle is determined for the plurality of first stores and the at least one second store by respective adjusted thresholds of ρ, δ, and γ obtained by the method according to any one of claims 1-6.
10. The method of updating a business district decision according to claim 9, the method being performed once every predetermined time period.
11. A device for training a business circle determination model, comprising:

    The first obtaining unit is configured to acquire location information of each of the plurality of stores in the predetermined geographical range and the business circle labeling information of the plurality of stores;

    The first calculating unit is configured to calculate, according to the CFSFDP clustering algorithm, a value of a local density ρ of each store, a value of a minimum distance δ with a higher density store, and a value of a product γ by using the position information;

    a second acquiring unit, configured to acquire the business circle determination information of each store according to a current threshold value of each of ρ, δ, and γ;

    The second calculating unit is configured to calculate the similarity of all the business circle determination information with respect to all the business circle label information by using the respective business circle determination information and the business circle label information of the plurality of stores;

    The threshold adjustment unit is configured to adjust respective threshold values of ρ, δ, and γ such that the similarity is improved.
12. The apparatus for training a business circle determination model according to claim 11, wherein a value of a local density ρ of the shop i in the plurality of stores is ρ _i , wherein

    ρ _i =∑ _j χ(d _ij -d _c ),

    Wherein, when d _ij -d _c <0, χ(d _ij -d _c )=1, when d _ij -d _c ≥0, χ(d _ij -d _c )=0,

    d _c is a radius threshold, d _ij is the distance between the store i and the store j in the plurality of stores, and i and j are natural numbers less than or equal to the total number of stores of the plurality of stores, and i ≠ j.
13. The apparatus for training a business circle determination model according to claim 11, wherein a value of a local density ρ of the shop i in the plurality of stores is ρ _i , wherein

    

    Where d _c is a radius threshold, d _ij is the distance between the store i and the store j in the plurality of stores, and i and j are natural numbers less than or equal to the total number of stores of the plurality of stores, and i≠j .
14. The apparatus for training a business circle determination model according to claim 12 or 13, wherein

    The position information of the shop i and the shop j are expressed as latitude and longitude (Lon _i , Lat _i ) and (Lon _j , Lat _j ), respectively, and the distance d _ij is calculated as follows:

    

    Where R is the radius of the Earth.
15. The apparatus for training a business circle determination model according to claim 11, wherein the similarity is represented by a parameter WFS, wherein

    

    

    Where i is an integer from 0 to A, j is an integer from 0 to B, A is the number of quotients in the labeled business circle, B is the number of quotients in the determination of the business circle, and N _i is the store included in the i-th labeled business circle Number, N is the total number of stores in the plurality of stores,

    P _ij is the accuracy rate for the i-th labeled business circle and the j-th determined business circle, and R _ij is the recall rate for the i-th labeled business circle and the j-th determined business circle, which will include a collection of labeled scattered stores Set to the 0th labeled business circle, and set the set of the determined scattered stores as the 0th determination business circle, wherein the marked scattered shop is an annotated shop that does not belong to any labeled business circle, and the determined scattered shop is A judgment shop that does not belong to any judgment business circle.
16. The apparatus for training a business circle determination model according to claim 11, wherein said threshold adjustment unit is further configured to adjust respective threshold values of ρ, δ, and γ such that said similarity is maximum.
17. A device for determining a business circle, comprising:

    An obtaining unit configured to acquire location information of each of a plurality of stores within a predetermined geographical range;

    a calculating unit configured to calculate, according to the CFSFDP clustering algorithm, a value of a local density ρ of each store, a value of a minimum distance δ with a higher density store, and a value of a product γ thereof;

    The determining unit is configured to determine the business circle for the plurality of stores by the adjusted threshold values of ρ, δ, and γ obtained by the method according to any one of claims 1-6.
18. The apparatus for determining a business district according to claim 17, wherein said predetermined geographic range is a predetermined city.
19. A device for updating a business circle determination, comprising:

    a first acquiring unit, configured to acquire first location information of each of the plurality of first stores in a predetermined geographic range and a first distance between the respective first stores;

    a second acquiring unit, configured to acquire second location information of each of the at least one second store in the predetermined geographic range;

    The first calculating unit is configured to calculate, by using the first location information and the second location information, a second distance between the second stores, any one of the second stores, and any one of the first stores The third distance between

    a second calculating unit, configured to calculate, according to the CFSFDP clustering algorithm, a value of a local density ρ of each of the plurality of first stores and at least one second store based on the first distance, the second distance, and the third distance The value of the minimum distance δ from the higher density store and the value of the product γ;

    The determining unit is configured to determine, by the respective adjusted thresholds of ρ, δ, and γ obtained by the method according to any one of claims 1-6, for the plurality of first stores and at least one second store Business circle.
20. The apparatus for updating a business district decision according to claim 19, the apparatus being implemented once every predetermined time period.

Images