Method and apparatus for mining competition relationship POIs

Patent metadata

Field	Value
Publication number	US-11580124-B2
Application number	US-202017110144-A
Country	US
Kind code	B2
Filing date	Dec 2, 2020
Priority date	Apr 22, 2020
Publication date	Feb 14, 2023
Grant date	Feb 14, 2023

Abstract

A method and apparatus for mining a competition relationship between POIs. An embodiment of the method includes: acquiring a graphlet mining result obtained by mining map retrieval data of users which encompasses attribute information of retrieved target POIs, the graphlet mining result encompassing occurrence frequencies of respective preset situations, and a preset situation comprising: conforming to attribute information of POIs represented by a corresponding preset graphlet and a preset association relationship between attribute information of at least two POIs; for a first and second POI, determining an occurrence frequency of a preset situation corresponding to a preset graphlet where attribute information of the first and second POI co-occur, and generating a relationship feature of the first and second POI; and inputting the relationship feature into a pre-trained relationship prediction model to obtain a competition relationship prediction result of the first POI and the second POI.

First claim

What is claimed is:

1. A method for mining a competition relationship between points of interest (POIs), the method comprising:
acquiring a graphlet mining result obtained by mining map retrieval data of users, the map retrieval data comprising attribute information of retrieved target POIs, the graphlet mining result comprising occurrence frequencies of respective preset situations, and a preset situation comprising: conforming to attribute information of POIs represented by a corresponding preset graphlet and a preset association relationship between attribute information of at least two POIs represented by the corresponding preset graphlet;
for a first POI and a second POI between which a competition relationship is to be determined, determining, based on the graphlet mining result, an occurrence frequency of a preset situation corresponding to a preset graphlet where attribute information of the first POI and attribute information of the second POI co-occur, and generating a relationship feature of the first POI and the second POI based on the determined occurrence frequency of the preset situation corresponding to the preset graphlet where the attribute information of the first POI and the attribute information of the second POI co-occur; and
inputting the relationship feature of the first POI and the second POI into a pre-trained relationship prediction model to obtain a competition relationship prediction result of the first POI and the second POI.

2. The method according to claim 1 , wherein the method further comprises:
acquiring the map retrieval data of the users, wherein the map retrieval data comprises time information when the users retrieve the target POIs;
constructing, based on associations between time information of different target POIs retrieved by same users, a POIs connection relationship graph representing relationships between target POIs; and
counting, on the basis of the POIs connection relationship graph, occurrence frequencies of preset situations corresponding to respective preset graphlets, to obtain the graphlet mining result.

3. The method according to claim 2 , wherein the preset graphlet comprises:
at least one first preset graphlet, the first preset graphlet representing attribute information of a pair of associated POIs and an association relationship between the attribute information of the pair of POIs and attribute information of at least one neighboring POI.

4. The method according to claim 3 , wherein the preset graphlet further comprises:
at least one second preset graphlet, the second preset graphlet representing a pair of associated POIs and an association relationship between the pair of POIs and at least one neighboring POI; and
the counting, on the basis of the POIs connection relationship graph, the occurrence frequencies of the preset situations corresponding to the respective preset graphlets, comprises:
respectively counting, based on the POIs connection relationship graph, frequencies of that the relationships between the target POIs conform to preset situations corresponding to respective second preset graphlets, to obtain counting results; and
for each first preset graphlet, respectively counting, based on attribute information of POIs in the second preset graphlets and the counting results of the frequencies of that the relationships between the target POIs respectively conform to the preset situations corresponding to the second preset graphlets, a frequency of that the attribute information of the target POIs and a relationship between the attribute information of the target POIs both conform to a preset situation corresponding to the first preset graphlet.

5. The method according to claim 4 , wherein the preset graphlet further comprises:
at least one third preset graphlet, the third preset graphlet comprising a pair of nodes and a neighboring node connected to at least one node in the pair of nodes, and node connection relationships in different third preset graphlets are different from each other;
the counting, on the basis of the POIs connection relationship graph, the occurrence frequencies of the preset situations corresponding to the respective preset graphlets, further comprises:
counting a frequency of situations that conform to a graph structure of each third preset graphlet, based on counting results of frequencies of that the attribute information of the target POIs and the relationships between the attribute information of the target POIs both conform to the preset situations corresponding to the first preset graphlets.

6. The method according to claim 1 , wherein the determining, based on the graphlet mining result, the occurrence frequency of the preset situation corresponding to the preset graphlet where the attribute information of the first POI and the attribute information of the second POI co-occur, and generating the relationship feature of the first POI and the second POI based on the determined occurrence frequency of the preset situation corresponding to the preset graphlet where the attribute information of the first POI and the attribute information of the second POI co-occur, comprises:
acquiring a first sorting list of the occurrence frequencies of respective preset situations, and a second sorting list of a preset number of hot preset situations ranked top in the first sorting list;
generating a first relationship feature of the first POI and the second POI, based on a ranking position of each preset graphlet, where the attribute information of the first POI and the attribute information of the second POI co-occur, in the first sorting list and occurrence frequencies of corresponding preset situations; and
generating a second relationship feature of the first POI and the second POI, based on a ranking position of each preset graphlet, where the attribute information of the first POI and the attribute information of the second POI co-occur, in the second sorting list and occurrence frequencies of corresponding preset situations.

7. The method according to claim 6 , wherein the method further comprises:
generating a joint attribute feature of the first POI and the second POI based on the attribute information of the first POI and the attribute information of the second POI;
the pre-trained relationship prediction model comprises: a self-attention module and a cross-attention module; and
the self-attention module processes the second relationship feature based on a preset self-attention mechanism, and the cross-attention module processes the joint attribute feature and the second relationship feature based on a preset cross-attention mechanism.

8. The method according to claim 7 , wherein the pre-trained relationship prediction model further comprises: a multi-layer perceptron; and
the multi-layer perceptron is configured to predict a competition relationship between the first POI and the second POI, based on the first relationship feature corresponding to preset graphlets, a feature output by the self-attention module, and a feature output by the cross-attention module.

9. The method according to claim 1 , wherein the method further comprises:
configuring a service resource related to the first POI based on the competition relationships between the first POI and at least two second POIs.

10. An electronic device, comprising:
at least one processor; and
a memory, communicatively connected to the at least one processor; wherein,
the memory, storing instructions executable by the at least one processor, the instructions, when executed by the at least one processor, cause the at least one processor to perform operations, the operations comprise:
acquiring a graphlet mining result obtained by mining map retrieval

Assignees

• Beijing Baidu Netcom Sci & Tech Co Ltd

Inventors

• Zhou Jingbo
• Xiong Hui

Classifications

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06F16/2465 Primary

  Query processing support for facilitating data mining operations in structured databases · CPC title

• G06F16/288 Primary

  Entity relationship models · CPC title

• G06F16/29 Primary

  Geographical information databases · CPC title