WO2024031943A1 - 门店去重处理方法、装置、设备及存储介质 - Google Patents

门店去重处理方法、装置、设备及存储介质 Download PDF

Info

Publication number
WO2024031943A1
WO2024031943A1 PCT/CN2023/074805 CN2023074805W WO2024031943A1 WO 2024031943 A1 WO2024031943 A1 WO 2024031943A1 CN 2023074805 W CN2023074805 W CN 2023074805W WO 2024031943 A1 WO2024031943 A1 WO 2024031943A1
Authority
WO
WIPO (PCT)
Prior art keywords
store
target
grid area
name
grid
Prior art date
Application number
PCT/CN2023/074805
Other languages
English (en)
French (fr)
Inventor
余玮琦
佘萧寒
曾泽华
姜华
高鹏飞
万四爽
刘蓝
Original Assignee
中国银联股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 中国银联股份有限公司 filed Critical 中国银联股份有限公司
Publication of WO2024031943A1 publication Critical patent/WO2024031943A1/zh

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0201—Market modelling; Market analysis; Collecting market data
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Definitions

  • This application belongs to the field of data processing, and in particular relates to a store deduplication processing method, device, equipment and storage medium.
  • the embodiments of this application provide a store deduplication processing method, device, equipment and storage medium, which can improve the efficiency of store deduplication processing.
  • embodiments of this application provide a store deduplication processing method, which includes: obtaining the first store name and first store location information of the target store; and determining, based on the first store location information, Determine the target grid area where the target store is located; in the pre-stored stock store database, obtain the second store name and second store location information of the stock store located in the target grid area and neighbor grid area.
  • the neighbor grid area and target The grid areas are adjacent; based on the first store name, first store location information, second store name and second store location information, the target similarity between the target store and the stock stores located in the target grid area and neighbor grid area is obtained ; When the target similarity is greater than or equal to the preset deduplication similarity threshold, remove the target store as a duplicate store.
  • embodiments of the present application provide a store deduplication processing device, including: a first acquisition module, used to obtain the first store name and first store location information of the target store; a grid area determination module, used according to The first store location information determines the target grid area where the target store is located; the second acquisition module is used to obtain the second store name of the stock store located in the target grid area and the neighbor grid area in the pre-stored stock store database.
  • the neighbor grid area is adjacent to the target grid area;
  • the calculation module is used to obtain the target store based on the first store name, the first store location information, the second store name and the second store location information The target similarity with the existing stores located in the target grid area and neighbor grid area;
  • the deduplication module is used to remove the target store as a duplicate store when the target similarity is greater than or equal to the preset deduplication similarity threshold.
  • inventions of the present application provide a store deduplication processing device.
  • the device includes: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, the store deduplication processing method of the first aspect is implemented.
  • embodiments of the present application provide a computer-readable storage medium.
  • Computer program instructions are stored on the computer-readable storage medium.
  • the store deduplication processing method of the first aspect is implemented.
  • Embodiments of the present application provide a store deduplication processing method, device, equipment and storage medium, which can determine the grid area where the target store is located based on the store location information of the target store.
  • Grid areas are areas divided into areas on the map.
  • the similarity between the target store and the stock store is obtained. This similarity determines whether the newly acquired store is the same store as the existing store. If the newly acquired store is the same store as the existing store, the newly acquired store will be considered a duplicate store and will be removed. It's time to go through it again The process does not require manual participation, and the location of the store can be used to narrow the range of stock stores for comparison, improving the efficiency of store duplication processing.
  • Figure 1 is a flow chart of a store deduplication processing method provided by an embodiment of the present application
  • Figure 2 is a schematic diagram of an example of a grid area in an embodiment of the present application.
  • Figure 3 is a flow chart of a store deduplication processing method provided by another embodiment of the present application.
  • Figure 4 is a schematic diagram of an example of a coding table in an embodiment of the present application.
  • Figure 5 is a flow chart of a store deduplication processing method provided by yet another embodiment of the present application.
  • Figure 6 is a schematic structural diagram of a store deduplication processing device provided by an embodiment of the present application.
  • Figure 7 is a schematic structural diagram of a store deduplication processing device provided by an embodiment of the present application.
  • This application provides a store deduplication processing method, device, equipment and storage medium, which can determine the grid area where the store is located based on the newly acquired store location information.
  • Grid areas are areas divided into areas on the map.
  • the similarity between the newly acquired store and the existing store is obtained. Based on the similarity, it is judged whether the newly acquired store is the same store as the existing store. If the newly acquired store is the same store as the existing store, the newly acquired store is considered to be a duplicate store and will be removed.
  • This deduplication process does not require manual participation, and the location of the store is used to narrow the range of existing stores for comparison, which improves the efficiency of store deduplication processing.
  • FIG. 1 is a flow chart of a store deduplication processing method provided by an embodiment of the present application. As shown in FIG. 1 , the store deduplication processing method may include steps S101 to S105.
  • step S101 the first store name and first store location information of the target store are obtained.
  • the target store is a store to be determined whether it is a duplicate store. It can be a store corresponding to the newly obtained store information, such as a new store to be added to the existing store database.
  • the first store name can be the store name of the target store.
  • the first store location information may be the store location information of the target store. Store location information is used to represent the location of the store, which may include store address, store longitude and latitude, etc., but is not limited here.
  • step S102 the target grid area where the target store is located is determined based on the first store location information.
  • the map can be pre-divided into grid areas.
  • the sizes of different grid areas may be the same or different, and are not limited here.
  • the shape of the grid area can be rectangular or regular The shape may also be an irregular shape, which is not limited here.
  • a grid area may be a rectangular area 150 meters long and 150 meters wide.
  • the target grid area is the grid area where the target store is located.
  • the first store location information can represent the location of the target store. According to the first store location information, the grid area where the target store is located can be determined, which is the target grid area.
  • step S103 obtain the second store name and second store location information of the existing stores located in the target grid area and the neighbor grid area from the pre-stored stock store database.
  • the existing store database includes relevant data of existing stores.
  • Existing stores are stores that have been identified as non-duplicate stores.
  • the relevant data of the existing stores may include but is not limited to the store name, store location information, grid area, etc. of the existing stores.
  • a geographical area where there may be stock stores that are the same store as the target store can be defined based on the location.
  • This geographical area is the surrounding area of the target store.
  • the target grid area and the neighbor grid area may be determined as surrounding areas of the location of the target store.
  • the neighbor grid area is adjacent to the target grid area, that is, the neighbor grid area is a grid area adjacent to the target grid area.
  • FIG. 2 is a schematic diagram of an example of a grid area in an embodiment of the present application.
  • nine grid areas are shown in dotted squares, namely grid areas A1 to A9.
  • Figure 2 also shows multiple stock stores 21.
  • grid area A5 is the target grid area, correspondingly, grid area A1, grid area A2, grid area A3, grid area A4, grid area A6, grid area A7, grid area A8 and grid area
  • the grid areas A9 are all neighbor grid areas of the target grid area. Taking the deduplication process of the target store located in the grid area A5 as an example, the store name and store location information of each stock store 21 in the grid area A1, and the store name and store location information of each stock store 21 in the grid area A2 can be obtained.
  • Store location information store name and store location information of each stock store 21 in grid area A3, store name and store location information of each stock store 21 in grid area A4, store name of each stock store 21 in grid area A6 and store location information, the store name and store location information of each stock store 21 in grid area A7, the store name and store location information of each stock store 21 in grid area A8, and the store name and store location information of each stock store 21 in grid area A9 Name and store location information.
  • the order of magnitude of the stock stores in the stock store database is very large. If the target store and the stock store are Comparing all existing stores in the database one by one will make the store deduplication process take a long time. Since the target grid area and the neighbor grid area are the surrounding areas of the target store, the stock stores located in the surrounding area of the target store and the target store are more likely to be the same store. You can first add the stock store database located in the target grid The relevant data of the existing stores in the region and neighboring grid areas are filtered out, and the relevant data of the existing stores located in the target grid area and the neighboring grid areas and the relevant data of the target stores are used to compare the existing stores with the target stores. Yes, to shorten the time required for store deduplication processing and improve the efficiency of store deduplication processing.
  • the stock stores located in the target grid area and the neighbor grid area include the stock stores located in the target grid area and the stock stores located in the neighbor grid area.
  • the second store name includes the store name of the stock store located in the target grid area and the store name of the stock store located in the neighboring grid area.
  • the second store location information includes store location information of the stock store located in the target grid area and store location information of the stock store located in the neighbor grid area.
  • step S104 based on the first store name, first store location information, second store name, and second store location information, the target similarity between the target store and the existing stores located in the target grid area and the neighbor grid area is obtained.
  • the similarity in store names between the target store and the existing stores can be obtained.
  • the geographical similarity between the target store and the existing stores can be obtained.
  • the target similarity can be obtained based on the similarity in store names and geographical locations between the target store and the existing stores.
  • the target similarity is the similarity between the target store and the existing stores.
  • the similarity between the target store and each stock store located in the target grid area and neighbor grid area can be calculated. Based on the target similarity, it is determined whether the target store is the same as the stock store located in the target grid area and neighbor grid area. of duplicate stores.
  • step S105 if the target similarity is greater than or equal to the preset deduplication similarity threshold, the target store is removed as a duplicate store.
  • the similarity threshold for deduplication is the threshold for confirming that the target store and the existing store are the same store. It can be set according to scenarios, needs, experience, etc., and is not limited here.
  • the similarity threshold for deduplication can be 0.6. If the target similarity is greater than or equal to the deduplication similarity threshold, it means that the target store and the existing store are the same store, that is, the target store is a duplicate store, and the target store can be removed. target Store removal can refer to discarding data related to the target store. If the target similarity is less than the deduplication similarity threshold, it means that the target store and the stock store are different stores, that is, the target store is not a duplicate store.
  • the relevant data of the target store can be stored in the stock store database, that is to say, the target store can be regarded as It is a newly added stock store in the stock store database.
  • the grid area where the target store is located can be determined based on the store location information of the target store.
  • Grid areas are areas divided into areas on the map. Based on the stock stores located in the target grid area where the target store is located in the database, the stock stores in the grid area surrounding the target grid area, and the store name and store location information of the target store, the similarity between the target store and the stock store is obtained. This similarity determines whether the newly acquired store is the same store as the existing store. If the newly acquired store is the same store as the existing store, the newly acquired store will be considered a duplicate store and will be removed.
  • This deduplication process does not require manual participation, and the location of the store can be used to narrow the range of existing stores for comparison, improving the efficiency of store deduplication processing.
  • the grid area has a grid code
  • neighbor grid areas of the target grid area can be determined based on the grid code of the target grid area and a grid coding algorithm.
  • Figure 3 is a flow chart of a store deduplication processing method provided by another embodiment of the present application. The difference between Figure 3 and Figure 1 is that the store deduplication processing method shown in Figure 3 can also include steps S106 to S108, and the store deduplication processing method shown in Figure 3 can also include steps S109 to S112, or Step S113 to step S115.
  • step S106 the map is divided into multiple grid areas, and a grid coding algorithm is used to assign a grid code to each grid area.
  • Geographic maps can be obtained and divided into grid areas. Each grid area is assigned a grid code, and the grid code can characterize the grid area, that is, different grid areas have different grid codes.
  • the trellis coding can be obtained according to the trellis coding algorithm, and the type of trellis coding algorithm is not limited here.
  • the grid codes calculated based on the position information of different locations in the same grid area are the same.
  • the grid code can be an m-digit string.
  • the first m1 characters in the grid code can represent provinces, cities, districts, etc.
  • the first m1 characters of multiple adjacent grid areas are consistent. Characters in bits m-m1 are different.
  • the last m-m1 characters of the grid code in different grid areas can be selected according to the preset coding table.
  • the coding table includes multiple coded characters arranged in a certain order. The order of the coded characters can be matched with the grid. Correspondence between regions, select the corresponding encoding characters as the last m-m1 characters of the trellis encoding.
  • Each bit in the last m-m1 bits of trellis coding can correspond to a coding table, and the coding tables corresponding to different bits can be the same or different.
  • the grid codes of the multiple grid areas it can be determined whether the multiple grid areas are adjacent. Furthermore, based on the grid codes of the multiple grid areas, the orientation relationship between the grid areas can be determined.
  • FIG. 4 is a schematic diagram of an example of a coding table in an embodiment of the present application.
  • the grid area is shown in Figure 2.
  • the grid code is a 7-digit string. If the first 6 characters in the grid code of adjacent grid areas are consistent, they are all wk2vu1. The last character is as shown in Figure 4.
  • the coding table is used for coding.
  • the grid code of grid area A1 is wk2vu1E
  • the grid code of grid area A2 is wk2vu1R
  • the grid code of grid area A3 is wk2vu1T
  • the grid code of grid area A4 is wk2vu1D.
  • the grid code of grid area A5 is wk2vu1F
  • the grid code of grid area A6 is wk2vu1G
  • the grid code of grid area A7 is wk2vu1C
  • the grid code of grid area A8 is wk2vu1V
  • the grid code of grid area A9 is The grid encoding is wk2vu1B.
  • step S107 the store location information of the existing stores is obtained, and the grid area where the existing stores are located is determined based on the store location information of the existing stores.
  • step S108 a first correspondence relationship between the grid code of the stock store and the grid area where the stock store is located is established, and the first correspondence relationship is stored in the stock store database.
  • the first correspondence includes a correspondence between the stock stores and the grid codes of the grid areas where the stock stores are located.
  • the data of the existing stores can be processed in advance, and a corresponding relationship between the obtained grid code of the grid area where the existing stores are located and the existing stores is established, and the corresponding relationship is stored in Stock store database, so that during the store deduplication process, the stock store corresponding to the grid code of the target grid area and the stock store corresponding to the grid code of the neighbor grid area can be directly found in the stock store database.
  • the target network The stock stores corresponding to the grid code of the grid area are the stock stores located in the target grid area, and the stock stores corresponding to the grid code of the neighbor grid area are the stock stores located in the neighbor grid area.
  • step S109 the grid code of the target grid area is obtained.
  • the grid code of the target grid area can be obtained.
  • step S110 the position of the vertex of the target grid area is obtained according to the grid coding of the target grid area and the grid coding inverse algorithm.
  • the trellis coding inverse algorithm is the inverse algorithm of the trellis coding algorithm. According to the position information of one or more positions in the grid area, using the grid coding algorithm, the grid code of the grid area can be obtained. According to the grid coding of the grid area, the position information of the vertices of the grid area can be obtained by using the grid coding inverse algorithm.
  • step S111 based on the position information of the vertices of the target mesh area, the position information of the auxiliary point located in the neighbor mesh area is determined.
  • the neighbor grid area shares some vertices with the target grid area, and obtaining the position information of the vertices of the target grid area is equivalent to obtaining the position information of some vertices of the neighbor grid area.
  • the position information of the auxiliary points in the neighbor grid area can be obtained.
  • the auxiliary point can be any point or multiple points in the neighboring grid area except the vertices shared with the target grid area, and is not limited here.
  • An auxiliary point can be determined in each neighbor grid area, so that the location information of the auxiliary point can be subsequently used to determine the neighbor grid area.
  • step S112 based on the position information of the auxiliary points in each neighbor grid area and the grid coding algorithm, the grid code of each neighbor grid area is calculated to determine the neighbor grid area.
  • the grid code has a corresponding relationship with the grid area.
  • the calculated grid code is the grid code of the neighbor grid area. Using the correspondence between grid codes and grid areas, neighbor grid areas can be determined.
  • step S113 the grid code of the target grid area is obtained.
  • step S114 the grid code of the candidate grid area is obtained according to the grid code of the target grid area.
  • the characters of a part of the grid codes of adjacent grid areas are the same.
  • This feature can be used to filter out grid areas adjacent to the target grid area, that is, candidate grid areas, from a large number of grid areas.
  • the candidate grid area includes a grid area in which characters of a part of the digits in the grid code are the same as characters of a part of the digits in the grid code of the target grid area. For example, neighbor The first m1-digit characters of the grid code in the nearest grid area are the same.
  • the first m1-digit characters of the grid code can be the same grid as the first m1-digit characters of the grid code in the target grid area.
  • the area is determined as a candidate grid area.
  • step S115 according to the corresponding relationship between the grid area arrangement and the characters of the coded digits in the grid coding algorithm, the grid coding of the neighbor grid area is determined in the grid coding of the candidate grid area to determine the neighbor network Grid area.
  • the grid coding algorithm may include the corresponding relationship between the grid area arrangement and the characters of the coded digits.
  • the grid area is arranged as shown in Figure 2.
  • the grid code is a 7-bit string.
  • the first 6 characters of the grid code of the candidate grid area are the same as the first 6 characters of the grid code of the target grid area.
  • the characters in the digits are the same.
  • the target grid area is grid area A5, and its grid code is wk2vu1D.
  • the corresponding relationship between the grid area arrangement and the last character of the grid code in the grid coding algorithm is specifically implemented as shown in the figure.
  • the target grid area has 8 neighbor grid areas, and the 8 neighbor grid areas are located at the upper left, upper, upper right, left, right, lower left, lower, and lower right of the target grid area.
  • the characters located at the upper left, upper, upper right, left, right, lower left, lower, and lower right of character D are W, E, R, S, F, X, C, and V respectively.
  • the eight neighbor grid areas located at the upper left, upper, upper right, left, right, lower left, lower, and lower right of the target grid area namely grid area A1, grid area A2, grid area A3, and grid area
  • the grid codes of grid area A4, grid area A6, grid area A7, grid area A8, and grid area A9 are wk2vu1W, wk2vu1E, wk2vu1R, wk2vu1S, wk2vu1F, wk2vu1X, wk2vu1C, and wk2vu1V respectively.
  • the grid code represents the grid area, and by determining the grid code of the neighbor grid area, the neighbor grid area can be determined.
  • the target similarity may be comprehensively obtained based on the similarity related to the store name and the similarity related to the store location information.
  • Figure 5 is a flow chart of a store deduplication processing method provided by yet another embodiment of the present application. The difference between Figure 5 and Figure 1 is that step S104 in Figure 1 can be specifically detailed into steps S1041 to step S1043 in Figure 5 .
  • step S1041 the target store is obtained based on the first store name and the second store name. Similarities related to N names of existing stores located in the target grid area and neighboring grid areas.
  • N is an integer greater than or equal to 1.
  • the name-related similarity is the similarity related to the store name, which can be obtained based on the first store name and the second store name.
  • Name-related similarity may include, but is not limited to, any one or more of character similarity, semantic similarity, and store type similarity.
  • Character similarity is the similarity of the characters that make up the store name.
  • Semantic similarity is the semantic similarity of store names.
  • the store type similarity is the similarity of the store type based on the store name.
  • name-related similarity includes character similarity.
  • the first store name and the second store name can be segmented separately to obtain the vocabulary corresponding to the first store name and the vocabulary corresponding to the second store name; calculate the word frequency of the vocabulary corresponding to the first store name and the vocabulary corresponding to the second store name (Term Frequency, TF) and inverse document frequency index (Inverse Document Frequency, IDF); select words whose word frequency is lower than or equal to the redundant word frequency threshold and whose inverse text frequency index is greater than the redundant frequency index threshold; based on the selected first store name correspondence
  • the vocabulary and the vocabulary corresponding to the selected second store name are used to obtain the character similarity between the target store and the existing stores located in the target grid area and neighbor grid area.
  • Word frequency represents the frequency of word occurrence.
  • the inverse text frequency index is used to characterize the discriminative ability of words.
  • the redundant word frequency threshold is a word frequency threshold used to distinguish whether a word is a redundant word.
  • the redundant frequency index threshold is the threshold of the inverse text frequency index used to distinguish whether a word is a redundant word.
  • the word frequency of a word is greater than the redundant word frequency threshold, it means that the word is a redundant word; if the inverse text frequency index of a word is less than or equal to the redundant frequency index threshold, it means that the word is a redundant word. Redundant words do not help in the calculation of character similarity, and may even have adverse effects, and do not need to participate in the calculation of character similarity. Words whose word frequency is lower than or equal to the redundant word frequency threshold and whose inverse text frequency index is greater than the redundant frequency index threshold are valid words that participate in the character similarity calculation.
  • the character similarity calculation can refer to the Bilingual Evaluation Understudy (BLEU) algorithm used in machine translation, and is evaluated by the N-gram overlap between the vocabulary corresponding to the selected first store name and the vocabulary corresponding to the second store name. Character similarity between the first store name and the second store name.
  • BLEU Bilingual Evaluation Understudy
  • name-related similarity includes semantic similarity. Convert the first store name and the second store name into the first name numeric sequence and the second name numeric sequence respectively; input the first name numeric sequence and the second name numeric sequence into the first model to obtain the target store output by the first model Semantic similarity with existing stores located in the target grid area and neighbor grid areas.
  • the first model is used to output the semantic similarity of the two store names based on the numerical sequences converted from the two input store names.
  • a certain number of labeled store names can be obtained in advance as positive samples of the training set, and a similar number of store names can be randomly selected as negative samples of the training set.
  • the positive samples and negative samples of the training set can be converted into digital sequences respectively, and the digital sequence can be used for training.
  • the first model may include a classification model, and may be a deep learning classification model or other types of classification models, which is not limited here.
  • the BERT (BidirectionalEncoder Representations from Transformer) model can be used to train the first model by taking "[CLS] + the number sequence corresponding to a certain store name + [SEP] + the number sequence corresponding to another store name" as input.
  • the first model is enabled to fit the semantic similarity between one store name and another store name, that is, the first model is enabled to output the semantic similarity between one store name and another store name based on the input.
  • the first name numeric sequence is the numeric sequence converted into the first store name.
  • the second name numeric sequence is the numeric sequence converted into the second store name.
  • the store name can be divided into characters, the divided characters can be converted into numbers, and the numbers corresponding to each character can be combined to obtain a number sequence.
  • the first model can output the store name of the target store and the store name of this stock store. Semantic similarity of names.
  • name-related similarity includes store type similarity.
  • store type similarity can be introduced to improve the store quality. Accuracy of deduplication.
  • the first store name information can be obtained based on the first store name; input the first store name information into the second model to obtain the store type probability vector of the target store output by the second model; search for the corresponding second store name in the inventory database store type probability vector; calculate the similarity between the store type probability vector of the target store and the store type probability vector corresponding to the second store name, and determine the similarity between the target store and the stock stores located in the target grid area and neighbor grid area store type similarity.
  • the second model is used to output a store type probability vector based on the input store name information.
  • the store type probability vector is used to represent the probability that the store indicated by the store name belongs to each store type.
  • Each element in the store type probability vector can represent the probability that the store belongs to a store type, and the store type corresponding to the element with the highest probability represented in the store type probability vector can be determined as the store type of the store.
  • the store type probability vector may be a normalized vector of length M, but is not limited to this.
  • a certain number of labeled store names and store types can be obtained in advance as a training set, such as ⁇ XXXX1 (store in B1 region), supermarket>, ⁇ YYYY2 (store in B2 region), coffee shop>, among which, XXXX1 (store in B1 region) and YYYY2 (B2 area store) are store names, and supermarket and cafe are store types.
  • a training set such as ⁇ XXXX1 (store in B1 region), supermarket>, ⁇ YYYY2 (store in B2 region), coffee shop>, among which, XXXX1 (store in B1 region) and YYYY2 (B2 area store) are store names, and supermarket and cafe are store types.
  • the second model may include a classification model, and may be a deep learning classification model or other types of classification models, which is not limited here.
  • the BERT model can be used to take "[CLS] + the number sequence corresponding to a certain store name" as input to train the second model, so that the second model can fit the correspondence between the store name and the store type, that is, , so that the second model can output the store type probability vector of the store name based on the input.
  • the first store name information is obtained based on the first store name. It can be the first store name, or it can be the processed information of the first store name, such as a digital sequence. The method of converting the store name into a digital sequence can be found in the above embodiment. The relevant instructions will not be repeated here.
  • the store type probability vector corresponding to the second store name includes the store type probability vectors corresponding to the existing stores located in the target grid area and the neighbor grid area.
  • the similarity between the store type probability vector of the target store and the store type probability vector corresponding to the second store name may be the cosine similarity of the two store type probability vectors.
  • the store type probability vector of the existing stores can be obtained in advance based on the store name of each existing store, so that when it is necessary to calculate the store type similarity, it can be obtained directly from the existing store database.
  • the store name of the stock store can be obtained, and store name information is obtained based on the store name; the store name information of the stock store is input into the second model, and the store type probability vector of the stock store output by the second model is obtained; the stock store and The second correspondence relationship of the store type probability vector of the stock store, and the second correspondence relationship is stored in the stock store database.
  • the store type probability vector corresponding to the second store name can be found in the existing store database according to the second correspondence relationship.
  • step S1042 based on the first store location information and the second store location information, the location similarity between the target store and the existing stores located in the target network area and the neighbor grid area is obtained.
  • the location similarity is the similarity related to the store location information, which can be obtained based on the first store location information and the second store location information.
  • Location similarity may be determined based on the distance between two store locations indicated by the two store location information and the amount of deviation that may result from the location information.
  • the geographical distance between the target store and the existing stores can be obtained based on the location information of the first store and the location information of the second store; based on the ratio of the geographical distance and the location deviation threshold, the distance between the target store and the target network area and the neighbor grid area can be obtained
  • the location similarity of existing stores may be positioning coordinate information, such as Global Positioning System (GPS) coordinate information.
  • GPS Global Positioning System
  • the address information can be converted into coordinate information, such as latitude and longitude information, and then the geographical distance between the target store and the existing stores is determined based on the coordinate information.
  • the position deviation threshold may be the maximum amount of deviation that the position information may cause.
  • the ratio of geographical distance and location deviation threshold can be used for normalization to obtain location similarity. For example, the location similarity can be obtained according to the following formula (1):
  • step S1043 the target similarity is calculated based on the N name-related similarities, position similarities and corresponding weight coefficients.
  • the weight coefficient can be used as an index or a product coefficient to participate in the calculation of target similarity, and is not limited here. In some examples, the weight coefficient can be used as an index to participate in the calculation of target similarity.
  • name-related similarity includes character similarity, semantic similarity and store type similarity.
  • sim target store, stock store
  • sim (character) is the character similarity
  • sim (semantic) is the semantic similarity
  • sim (type) is the store type similarity
  • sim (location) is the location similarity degree
  • ⁇ is the weight coefficient of character similarity
  • ⁇ is the weight coefficient of semantic similarity
  • ⁇ is the weight coefficient of store type similarity
  • ⁇ is the weight coefficient of location similarity.
  • name-related similarity includes character similarity, semantic similarity, and store type similarity.
  • the store name and store address of the target store and convert the store address into longitude and latitude coordinates.
  • the converted longitude and latitude coordinates are ⁇ 30.193, 120.173 ⁇ .
  • the grid code of the grid area where the target store is located is calculated as wtm7y8e.
  • the first 6 characters of the grid code of the neighbor grid area are the same as the first 6 characters of the grid code of the target grid area.
  • the grid codes of the 8 neighbor grid areas can be obtained by using the encoding table as shown in Figure 4. .
  • the grid codes of the eight neighbor grid areas are wtm7y82, wtm7y83, wtm7y84, wtm7y8W, wtm7y8R, wtm7y8S, wtm7y8D and wtm7y8F.
  • the following takes the calculation of the target similarity between the target store and one of the existing stores as an example.
  • the store name of the target store is "X1X2 (Hangzhou Binjiang Baolong City Plaza Store)", and the existing store name is "Hangzhou Binjiang District X3X4 Convenience Store”.
  • X1, X2, X3 and Chinese characters are examples of the target store and one of the existing stores.
  • the vocabulary corresponding to the target store includes ⁇ X1X2 ⁇ , ⁇ ( ⁇ , ⁇ Hangzhou City ⁇ , ⁇ Binjiang ⁇ , ⁇ Baolong ⁇ , ⁇ city ⁇ , ⁇ square ⁇ , ⁇ store ⁇ and ⁇ ) ⁇ .
  • the vocabulary corresponding to the existing stores includes ⁇ Hangzhou City ⁇ , ⁇ Binjiang District ⁇ , ⁇ X3X4 ⁇ and ⁇ convenience store ⁇ . Calculate the word frequency and inverse text frequency index of each vocabulary.
  • the word frequency and inverse text frequency index of ⁇ ( ⁇ , ⁇ Hangzhou City ⁇ and ⁇ ) ⁇ do not meet the requirement that the word frequency is lower than or equal to the redundant word frequency threshold and the inverse text frequency index is greater than the redundant word frequency index.
  • residual frequency The condition of the rate index threshold, so the words ⁇ ( ⁇ , ⁇ Hangzhou City ⁇ and ⁇ ) ⁇ are discarded.
  • the selected vocabulary combination corresponding to the target store is "X1X2 Binjiang Baolong City Plaza Store”
  • the selected vocabulary combination corresponding to the existing store is "Binjiang District X3X4 convenience store”.
  • X1X2 Binjiang Baolong City Plaza Store contains 11 1-grams
  • "Binjiang District X3X4 Convenience Store” contains 8 1-grams. Calculate the co-occurrence of the two 1-grams respectively. times, it can be seen that the three 1-grams of ⁇ bin ⁇ , ⁇ jiang ⁇ and ⁇ dian ⁇ appear together once each. Therefore, the character similarity between "X1X2 Binjiang Powerlong City Plaza Store” and "Binjiang District X3X4 Convenience Store” is (3 /11+3/8)/2 ⁇ 0.32.
  • X1X2 (Binjiang Baolong City Plaza Store, Hangzhou)
  • X3X4 Convenience Store (Binjiang Baolong City Plaza Store, Hangzhou)
  • the same Chinese characters correspond to the same numbers.
  • the above two digital sequences are spliced with [CLS] and [SEP], combined into a single vector, and input into the first model to obtain the semantic similarity between the two output by the first model.
  • the two digital sequences converted from "X1X2 (Hangzhou Binjiang Baolong City Plaza Store)" and "Hangzhou Binjiang District X3X4 Convenience Store” can be input into the second model respectively to obtain the store type probability vector of the target store and the store type probability vector of the existing stores.
  • Store type probability vector The values of elements in the three store type dimensions of "shopping", "supermarket” and “convenience store” are relatively high between the target store and the existing store.
  • the store type probability vector obtained based on the store type probability vector of the target store and the store type probability vector of the existing store The store types represented by type similarity are relatively close.
  • the geographical distance between the two is determined to be 285 meters. Based on the geographical distance and the location deviation threshold, the location similarity can be calculated to be 0.8585.
  • the deduplication similarity threshold is 0.6.
  • the target similarity calculated using the above formula (2) is less than 0.6, it can be determined that the target store and the stock store are not the same store.
  • FIG. 6 is a schematic structural diagram of a store deduplication processing device provided by an embodiment of the present application. As shown in Figure 6, the store’s deduplication process
  • the setup 300 may include a first acquisition module 301, a grid area determination module 302, a second acquisition module 303, a calculation module 304 and a deduplication module 305.
  • the first acquisition module 301 may be used to acquire the first store name and first store location information of the target store.
  • the grid area determination module 302 may be used to determine the target grid area where the target store is located based on the first store location information.
  • the second acquisition module 303 may be used to obtain the second store name and second store location information of the existing stores located in the target grid area and the neighbor grid area from the pre-stored inventory store database.
  • the neighbor grid area is adjacent to the target grid area.
  • the calculation module 304 may be used to obtain the target similarity between the target store and the stock stores located in the target grid area and the neighbor grid area based on the first store name, the first store location information, the second store name, and the second store location information.
  • the deduplication module 305 may be used to remove the target store as a duplicate store when the target similarity is greater than or equal to the preset deduplication similarity threshold.
  • the grid area where the target store is located can be determined based on the store location information of the target store.
  • Grid areas are areas divided into areas on the map. Based on the stock stores located in the target grid area where the target store is located in the database, the stock stores in the grid area surrounding the target grid area, and the store name and store location information of the target store, the similarity between the target store and the stock store is obtained. This similarity determines whether the newly acquired store is the same store as the existing store. If the newly acquired store is the same store as the existing store, the newly acquired store will be considered a duplicate store and will be removed.
  • This deduplication process does not require manual participation, and the location of the store can be used to narrow the range of existing stores for comparison, improving the efficiency of store deduplication processing.
  • the target store in addition to comparing the target store with the existing stores in the target grid area, it also compares the target store with the existing stores in the neighbor grid area to avoid missing the store that is located near the boundary of the target grid area and is the same store as the target store. existing stores to further improve the comprehensiveness and accuracy of store duplication processing.
  • the grid areas have grid coding.
  • the store deduplication processing device 200 may also include a neighbor grid area determination module.
  • the neighbor grid area determination module can be used to: obtain the grid code of the target grid area; obtain the position information of the vertices of the target grid area according to the grid code of the target grid area and the grid coding inverse algorithm ; Based on the position information of the vertices of the target grid area, determine the location information of the auxiliary points located in the neighbor grid area; Based on the location information of the auxiliary points in each neighbor grid area and the grid coding algorithm, calculate each neighbor network Grid coding of grid areas to determine neighbor grid areas.
  • adjacent grid regions have the same value for a portion of the digits in the grid code.
  • the neighbor grid area determination module can be used to: obtain the grid code of the target grid area; obtain the grid code of the candidate grid area based on the grid code of the target grid area, and the candidate grid area includes a part of the digits in the grid code.
  • the characters in the grid area are the same as the characters of a part of the digits in the grid coding of the target grid area; according to the corresponding relationship between the grid area arrangement and the characters of the coded digits in the grid coding algorithm, the network in the candidate grid area Determine the grid code of the neighbor grid area in the grid code to determine the neighbor grid area.
  • the store deduplication device 200 may also include a first preprocessing module.
  • the first preprocessing module can be used to: divide the map into multiple grid areas, and use the grid coding algorithm to assign grid codes to each grid area; obtain the store location information of the stock stores, and based on the store locations of the stock stores Information, determine the grid area where the stock store is located; establish a first correspondence between the stock store and the grid code of the grid area where the stock store is located, and store the first correspondence in the stock store database.
  • the calculation module 304 can be used to: based on the first store name and the second store name, obtain the name-related similarities between the target store and the N names of the stock stores located in the target grid area and the neighbor grid area, N is An integer greater than or equal to 1; based on the location information of the first store and the location information of the second store, the location similarity between the target store and the existing stores located in the target network area and neighbor grid area is obtained; based on N name-related similarity, location similarity degree and the corresponding weight coefficient to calculate the target similarity.
  • name-related similarity includes character similarity.
  • the calculation module 304 can be used to: segment the first store name and the second store name respectively to obtain the vocabulary corresponding to the first store name and the vocabulary corresponding to the second store name; calculate the vocabulary corresponding to the first store name and the second store name.
  • the word frequency and inverse text frequency index of the corresponding vocabulary select word frequency less than or equal to redundancy
  • the word frequency threshold and the inverse text frequency index are greater than the redundant frequency index threshold; based on the vocabulary corresponding to the selected first store name and the vocabulary corresponding to the selected second store name, the target store is located in the target grid area and the neighbor grid. Character similarity of existing stores in the region.
  • name-related similarity includes semantic similarity.
  • the calculation module 304 can be used to: convert the first store name and the second store name into a first name number sequence and a second name number sequence respectively; input the first name number sequence and the second name number sequence into the first model to obtain the first name number sequence and the second name number sequence.
  • the first model outputs the semantic similarity between the target store and the existing stores located in the target grid area and neighbor grid area.
  • the first model is used to output the semantic similarity of the two store names based on the numerical sequence converted from the two input store names. Spend.
  • name-related similarity includes store type similarity.
  • the calculation module 304 can be used to: obtain the first store name information according to the first store name; input the first store name information into the second model to obtain the store type probability vector of the target store output by the second model, and the second model is used to obtain the store type probability vector of the target store according to the first store name.
  • the input store name information outputs a store type probability vector.
  • the store type probability vector is used to represent the probability that the store indicated by the store name belongs to each store type; search for the store type probability vector corresponding to the second store name in the inventory database; calculate the target store
  • the similarity between the store type probability vector and the store type probability vector corresponding to the second store name is determined as the store type similarity between the target store and the stock stores located in the target grid area and neighbor grid area.
  • the calculation module 304 can be used to: obtain the geographical distance between the target store and the existing stores based on the first store location information and the second store location information; obtain the target store and the target store based on the ratio of the geographical distance and the location deviation threshold. The location similarity of the existing stores in the network area and neighboring grid areas.
  • the store deduplication processing device may also include a second preprocessing module.
  • the second preprocessing module can be used to: obtain the store name of the existing store, and obtain store name information based on the store name; input the store name information of the existing store into the second model, and obtain the store type probability vector of the existing store output by the second model; A second correspondence relationship between the stock stores and the store type probability vectors of the stock stores is established, and the second correspondence relationship is stored in the stock store database.
  • FIG. 7 is a schematic structural diagram of a store deduplication processing device provided by an embodiment of the present application.
  • the store deduplication processing equipment 400 includes a memory 401, a processor 402, and a computer program stored on the memory 401 and executable on the processor 402.
  • the above-mentioned processor 402 may include a central processing unit (CPU), or an Application Specific Integrated Circuit (ASIC), or one or more integrated circuits that may be configured to implement embodiments of the present application.
  • CPU central processing unit
  • ASIC Application Specific Integrated Circuit
  • Memory 401 may include read-only memory (ROM), random access memory (Random Access Memory, RAM), magnetic disk storage media devices, optical storage media devices, flash memory devices, electrical, optical or other physical/tangible devices Memory storage device.
  • ROM read-only memory
  • RAM random access memory
  • magnetic disk storage media devices e.g., magnetic disks
  • optical storage media devices e.g., magnetic disks
  • flash memory devices e.g., electrical, optical or other physical/tangible devices Memory storage device.
  • memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or multiple processors), it is operable to perform the operations described with reference to the store deduplication processing method according to the embodiment of the present application.
  • the processor 402 reads the executable program code stored in the memory 401 to run the computer program corresponding to the executable program code, so as to implement the store deduplication processing method in the above embodiment.
  • the store deduplication processing device 400 may also include a communication interface 403 and a bus 404. Among them, as shown in Figure 7, the memory 401, the processor 402, and the communication interface 403 are connected through the bus 404 and complete communication with each other.
  • the communication interface 403 is mainly used to implement communication between modules, devices, units and/or equipment in the embodiments of this application. Input devices and/or output devices can also be accessed through the communication interface 403.
  • Bus 404 includes hardware, software, or both, coupling the components of store deduplication processing device 400 to one another.
  • the bus 404 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), Hyper Transport (HT) interconnect, Industry Standard Architecture (ISA) bus, infinite bandwidth interconnect, low pin count (LPC) bus, memory bus, Micro Channel architecture Architecture, MCA) bus, Peripheral Component Interconnect (PCI) bus, PCI-Express (PCI-E) bus, Serial Advanced Technology Attachment (Serial Advanced Technology Attachment, SATA) bus, Video Electronics Standards Association Local Bus (VLB) bus or other suitable bus or a combination of two or more of these.
  • bus 404 may include one or more buses.
  • the fourth aspect of the present application provides a computer-readable storage medium.
  • Computer program instructions are stored on the computer-readable storage medium.
  • the store deduplication processing method in the above embodiment can be implemented, and can achieve the same technical effect, so to avoid repetition, we will not repeat them here.
  • the above-mentioned computer-readable storage media may include non-transitory computer-readable storage media, such as read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), magnetic disks or optical disks etc. are not limited here.
  • An embodiment of the present application provides a computer program product.
  • the electronic device can execute the store deduplication processing method in the above embodiment and achieve the same technical effect. , to avoid repetition, will not be repeated here.
  • Such a processor may be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It will also be understood that each block in the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations, can also be implemented by special purpose hardware that performs the specified functions or actions, or can be implemented by special purpose hardware and A combination of computer instructions.

Landscapes

  • Business, Economics & Management (AREA)
  • Strategic Management (AREA)
  • Engineering & Computer Science (AREA)
  • Accounting & Taxation (AREA)
  • Development Economics (AREA)
  • Finance (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Game Theory and Decision Science (AREA)
  • Data Mining & Analysis (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

本申请公开了一种门店去重处理方法、装置、设备及存储介质，属于数据处理领域。该方法包括：获取目标门店的第一门店名称和第一门店位置信息；根据第一门店位置信息，确定目标门店所在的目标网格区域；在预存的存量门店数据库中，获取位于目标网格区域和邻居网格区域的存量门店的第二门店名称和第二门店位置信息；基于第一门店名称、第一门店位置信息、第二门店名称和第二门店位置信息，得到目标门店与位于目标网格区域和邻居网格区域的存量门店的目标相似度；在目标相似度大于等于预设的去重相似度阈值的情况下，将目标门店作为重复门店去除。

Description

门店去重处理方法、装置、设备及存储介质
相关申请的交叉引用
本申请要求享有于2022年08月10日提交的名称为“门店去重处理方法、装置、设备及存储介质”的中国专利申请202210957641.0的优先权，该申请的全部内容通过引用并入本文中。
技术领域
本申请属于数据处理领域，尤其涉及一种门店去重处理方法、装置、设备及存储介质。
背景技术
随着电子支付技术的推广，用户在商户线下的门店中可利用电子支付技术进行支付。为了便于处理商户线下的门店中的电子支付，需要对商户线下的门店进行信息管理。但在门店数据由不同来源上送的情况下，不同来源可能会上送同一门店的门店数据，且不同来源上送的同一门店的门店数据可能会有所不同，导致根据门店数据将同一门店误判为两个不同的门店，即同一门店被反复统计。
为了避免同一门店被反复统计，需要派遣人员前往门店现场进行巡检，人工判断同一门店是否被反复统计。但人工巡检花费的时间、人力非常大，门店去重处理的效率很低。
发明内容
本申请实施例提供一种门店去重处理方法、装置、设备及存储介质，能够提高门店去重处理的效率。
第一方面，本申请实施例提供一种门店去重处理方法，包括：获取目标门店的第一门店名称和第一门店位置信息；根据第一门店位置信息，确 定目标门店所在的目标网格区域；在预存的存量门店数据库中，获取位于目标网格区域和邻居网格区域的存量门店的第二门店名称和第二门店位置信息，邻居网格区域与目标网格区域相邻；基于第一门店名称、第一门店位置信息、第二门店名称和第二门店位置信息，得到目标门店与位于目标网格区域和邻居网格区域的存量门店的目标相似度；在目标相似度大于等于预设的去重相似度阈值的情况下，将目标门店作为重复门店去除。
第二方面，本申请实施例提供一种门店去重处理装置，包括：第一获取模块，用于获取目标门店的第一门店名称和第一门店位置信息；网格区域确定模块，用于根据第一门店位置信息，确定目标门店所在的目标网格区域；第二获取模块，用于在预存的存量门店数据库中，获取位于目标网格区域和邻居网格区域的存量门店的第二门店名称和第二门店位置信息，邻居网格区域与目标网格区域相邻；计算模块，用于基于第一门店名称、第一门店位置信息、第二门店名称和第二门店位置信息，得到目标门店与位于目标网格区域和邻居网格区域的存量门店的目标相似度；去重模块，用于在目标相似度大于等于预设的去重相似度阈值的情况下，将目标门店作为重复门店去除。
第三方面，本申请实施例提供一种门店去重处理设备，设备包括：处理器以及存储有计算机程序指令的存储器；处理器执行计算机程序指令时实现第一方面的门店去重处理方法。
第四方面，本申请实施例提供一种计算机可读存储介质，计算机可读存储介质上存储有计算机程序指令，计算机程序指令被处理器执行时实现第一方面的门店去重处理方法。
本申请实施例提供一种门店去重处理方法、装置、设备及存储介质，可根据目标门店的门店位置信息，确定目标门店所在的网格区域。网格区域为地图中划分的区域。基于数据库中位于目标门店所在的目标网格区域的存量门店、目标网格区域周边的网格区域的存量门店以及目标门店的门店名称、门店位置信息，得到目标门店与存量门店的相似度，根据该相似度判断新获取的门店是否与存量门店为同一门店，若新获取的门店与存量门店为同一门店，则认为新获取的门店为重复门店，予以去除。该去重过 程不需人工参与，且利用门店的位置可缩小用于比对的存量门店的范围，提高了门店去重处理的效率。
附图说明
为了更清楚地说明本申请实施例的技术方案，下面将对本申请实施例中所需要使用的附图作简单的介绍，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1为本申请一实施例提供的门店去重处理方法的流程图；
图2为本申请实施例中网格区域的一示例的示意图；
图3为本申请另一实施例提供的门店去重处理方法的流程图；
图4为本申请实施例中编码表的一示例的示意图；
图5为本申请又一实施例提供的门店去重处理方法的流程图；
图6为本申请一实施例提供的门店去重处理装置的结构示意图；
图7为本申请一实施例提供的门店去重处理设备的结构示意图。
具体实施方式
下面将详细描述本申请的各个方面的特征和示例性实施例，为了使本申请的目的、技术方案及优点更加清楚明白，以下结合附图及具体实施例，对本申请进行进一步详细描述。应理解，此处所描述的具体实施例仅意在解释本申请，而不是限定本申请。对于本领域技术人员来说，本申请可以在不需要这些具体细节中的一些细节的情况下实施。下面对实施例的描述仅仅是为了通过示出本申请的示例来提供对本申请更好的理解。
随着电子支付技术的推广，用户在商户线下的门店中可利用电子支付技术进行支付。为了便于处理商户线下的门店中的电子支付，需要对商户线下的门店进行信息管理。但在门店数据由不同来源上送的情况下，不同来源可能会上送同一门店的门店数据，且不同来源上送的同一门店的门店数据可能会有所不同，导致根据门店数据将同一门店误判为两个不同的门店，即同一门店被反复统计。在一些情况下，同一门店被反复统计的情况也可能会被利用，从而造成存储门店信息的数据库中的漏洞。
为了避免同一门店被反复统计，需要派遣人员前往门店现场进行巡检，人工判断同一门店是否被反复统计。但人工巡检花费的时间、人力非常大，门店去重处理的效率很低。
本申请提供一种门店去重处理方法、装置、设备及存储介质，可根据新获取的门店的门店位置信息，确定该门店所在的网格区域。网格区域为地图中划分的区域。利用数据库中位于新获取的门店所在的目标网格区域和目标网格区域周边的网格区域的存量门店的数据，以及新获取的门店的数据，得到新获取的门店与存量门店的相似度，根据该相似度判断新获取的门店是否与存量门店为同一门店，若新获取的门店与存量门店为同一门店，则认为新获取的门店为重复门店，予以去除。该去重过程不需人工参与，且利用门店的位置缩小用于比对的存量门店的范围，提高了门店去重处理的效率。
下面对本申请提供的门店去重处理方法、装置、设备及存储介质分别进行说明。
本申请第一方面提供一种门店去重处理方法，可应用于根据不同来源收集来的门店信息进行门店去重的场景，可由门店去重装置、设备等执行，在此并不限定。图1为本申请一实施例提供的门店去重处理方法的流程图，如图1所示，门店去重处理方法可包括步骤S101至步骤S105。
在步骤S101中，获取目标门店的第一门店名称和第一门店位置信息。
目标门店为待判断是否为重复门店的门店，可以为新获取到的门店信息对应的门店，如新的欲加入存量门店数据库中的门店。第一门店名称可为目标门店的门店名称。第一门店位置信息可为目标门店的门店位置信息。门店位置信息用于表征门店的位置，可包括门店地址、门店经纬度等，在此并不限定。
在步骤S102中，根据第一门店位置信息，确定目标门店所在的目标网格区域。
为了便于处理，可预先将地图划分为多个网格区域。不同网格区域的大小可以相同，也可不同，在此并不限定。网格区域的形状可为矩形等规 则形状，也可为不规则形状，在此并不限定。例如，网格区域可为长为150米，宽为150米的矩形区域。
目标网格区域为目标门店所在的网格区域。第一门店位置信息可表征目标门店的位置，根据第一门店位置信息可确定目标门店所在的网格区域即目标网格区域。
在步骤S103中，在预存的存量门店数据库中，获取位于目标网格区域和邻居网格区域的存量门店的第二门店名称和第二门店位置信息。
存量门店数据库包括存量门店的相关数据。存量门店为已确定为非重复门店的门店。存量门店的相关数据可包括但不限于存量门店的门店名称、门店位置信息、所在网格区域等。
为了缩小与目标门店比对的存量门店的范围，可根据位置先行划定一个可能存在与目标门店为同一门店的存量门店的地理区域，该地理区域为目标门店的位置的周边区域。可将目标网格区域和邻居网格区域确定为目标门店的位置的周边区域。邻居网格区域与目标网格区域相邻，即，邻居网格区域为与目标网格区域相邻的网格区域。
例如，图2为本申请实施例中网格区域的一示例的示意图。图2中以虚线方格示出了9个网格区域，分别为网格区域A1至A9。图2还示出了多个存量门店21。若网格区域A5为目标网格区域，对应地，网格区域A1、网格区域A2、网格区域A3、网格区域A4、网格区域A6、网格区域A7、网格区域A8和网格区域A9均为目标网格区域的邻居网格区域。以对位于网格区域A5中的目标门店进行去重处理为例，可获取网格区域A1中各存量门店21的门店名称和门店位置信息、网格区域A2中各存量门店21的门店名称和门店位置信息、网格区域A3中各存量门店21的门店名称和门店位置信息、网格区域A4中各存量门店21的门店名称和门店位置信息、网格区域A6中各存量门店21的门店名称和门店位置信息、网格区域A7中各存量门店21的门店名称和门店位置信息、网格区域A8中各存量门店21的门店名称和门店位置信息以及网格区域A9中各存量门店21的门店名称和门店位置信息。
存量门店数据库中存量门店的数量级很大，若将目标门店与存量门店 数据库中所有存量门店一一比对，会使得门店去重处理所需时间较长。由于目标网格区域和邻居网格区域为目标门店的周边区域，位于目标门店的周边区域中的存量门店和目标门店为同一门店的可能性较大，可先将存量门店数据库中位于目标网格区域的和邻居网格区域的存量门店的相关数据筛选出来，利用位于目标网格区域的和邻居网格区域的存量门店的相关数据和目标门店的相关数据，来进行存量门店与目标门店的比对，以缩短门店去重处理所需时间，提高门店去重处理的效率。
位于目标网格区域和邻居网格区域的存量门店包括位于目标网格区域的存量门店和位于邻居网格区域的存量门店。第二门店名称包括位于目标网格区域的存量门店的门店名称和位于邻居网格区域的存量门店的门店名称。第二门店位置信息包括位于目标网格区域的存量门店的门店位置信息和位于邻居网格区域的存量门店的门店位置信息。
在步骤S104中，基于第一门店名称、第一门店位置信息、第二门店名称和第二门店位置信息，得到目标门店与位于目标网格区域和邻居网格区域的存量门店的目标相似度。
基于第一门店名称和第二门店名称，可得到目标门店与存量门店在门店名称方面的相似度。基于第一门店位置信息和第二门店位置信息，可得到目标门店与存量门店在地理位置方面的相似度。根据目标门店与存量门店在门店名称方面的相似度和在地理位置方面的相似度，可得到目标相似度。目标相似度为目标门店与存量门店的相似度。可计算得到目标门店与位于目标网格区域和邻居网格区域的每个存量门店的相似度，根据目标相似度，确定目标门店是否为与位于目标网格区域和邻居网格区域的存量门店相同的重复门店。
在步骤S105中，在目标相似度大于等于预设的去重相似度阈值的情况下，将目标门店作为重复门店去除。
去重相似度阈值为确认目标门店与存量门店为同一门店的相似度的阈值，可根据场景、需求、经验等设定，在此并不限定，例如，去重相似度阈值可为0.6。目标相似度大于等于去重相似度阈值，表示目标门店与存量门店为同一门店，即目标门店为重复门店，可将目标门店去除。将目标 门店去除可指舍弃目标门店的相关数据。目标相似度小于去重相似度阈值，表示目标门店与存量门店为不同的门店，即目标门店不是重复门店，可将目标门店的相关数据存储入存量门店数据库，也就是说，可将目标门店视为新加入存量门店数据库中的存量门店。
在本申请实施例中，可根据目标门店的门店位置信息，确定目标门店所在的网格区域。网格区域为地图中划分的区域。基于数据库中位于目标门店所在的目标网格区域的存量门店、目标网格区域周边的网格区域的存量门店以及目标门店的门店名称、门店位置信息，得到目标门店与存量门店的相似度，根据该相似度判断新获取的门店是否与存量门店为同一门店，若新获取的门店与存量门店为同一门店，则认为新获取的门店为重复门店，予以去除。该去重过程不需人工参与，且利用门店的位置可缩小用于比对的存量门店的范围，提高了门店去重处理的效率。
而且，除了比对目标门店与目标网格区域中的存量门店以外，还可比对目标门店与邻居网格区域中的存量门店，避免漏查位于目标网格区域的边界附近与目标门店为同一门店的存量门店，进一步提高门店去重处理的全面性和准确性。
在一些实施例中，网格区域具有网格编码，可基于目标网格区域的网格编码和网格编码算法，确定目标网格区域的邻居网格区域。图3为本申请另一实施例提供的门店去重处理方法的流程图。图3与图1的不同之处在于，图3所示的门店去重处理方法还可包括步骤S106至步骤S108，图3所示的门店去重处理方法还可包括步骤S109至步骤S112，或步骤S113至步骤S115。
在步骤S106中，将地图划分为多个网格区域，并利用网格编码算法，为每个网格区域分配网格编码。
可获取地理地图，将地理地图划分为多个网格区域。为每个网格区域分配一个网格编码，网格编码可表征网格区域，即，不同的网格区域的网格编码不同。网格编码可根据网格编码算法得到，在此并不限定网格编码算法的类型。根据同一网格区域中不同位置的位置信息计算得到网格编码相同。
在一些示例中，网格编码可为m位字符串，网格编码中的前m1位的字符可表征省、市、区等，邻近的多个网格区域的前m1位的字符一致，后m-m1位的字符不同。不同网格区域的网格编码的后m-m1位的字符可按照预设的编码表选取，编码表包括多个按一定顺序排布的编码字符，可按照编码字符的排布顺序与网格区域的对应关系，选择对应的编码字符作为网格编码的后m-m1位的字符。网格编码的后m-m1位中每一位可对应一张编码表，不同位对应的编码表可以相同，也可以不同。根据多个网格区域的网格编码，可确定多个网格区域是否邻近，进一步地，还可根据多个网格区域的网格编码，确定网格区域之间的方位关系。
例如，图4为本申请实施例中编码表的一示例的示意图。网格区域如图2所示，网格编码为7位字符串，若邻近的网格区域的网格编码中前6位的字符一致，均为wk2vu1，最后一位的字符按照图4所示的编码表进行编码，网格区域A1的网格编码为wk2vu1E，则网格区域A2的网格编码为wk2vu1R，网格区域A3的网格编码为wk2vu1T，网格区域A4的网格编码为wk2vu1D，网格区域A5的网格编码为wk2vu1F，网格区域A6的网格编码为wk2vu1G，网格区域A7的网格编码为wk2vu1C，网格区域A8的网格编码为wk2vu1V，网格区域A9的网格编码为wk2vu1B。
在步骤S107中，获取存量门店的门店位置信息，根据存量门店的门店位置信息，确定存量门店所在的网格区域。
在步骤S108中，建立存量门店和存量门店所在的网格区域的网格编码的第一对应关系，并将第一对应关系存储于存量门店数据库。
第一对应关系包括存量门店和存量门店所在的网格区域的网格编码的对应关系。为了进一步缩短门店去重处理所需的时间，可预先对存量门店的数据进行处理，将得到的存量门店所在的网格区域的网格编码与存量门店建立对应关系，并将该对应关系存储入存量门店数据库，以便于在门店去重处理过程中可在存量门店数据库中直接查找到目标网格区域的网格编码对应的存量门店以及邻居网格区域的网格编码对应的存量门店，目标网格区域的网格编码对应的存量门店为位于目标网格区域的存量门店，邻居网格区域的网格编码对应的存量门店为位于邻居网格区域的存量门店。
在步骤S109中，获取目标网格区域的网格编码。
确定目标网格区域后，可获取目标网格区域的网格编码。
在步骤S110中，根据目标网格区域的网格编码和网格编码逆算法，获取目标网格区域的顶点的位置。
网格编码逆算法为网格编码算法的逆算法。根据网格区域中一个或多个位置的位置信息，利用网格编码算法，可得到该网格区域的网格编码。根据网格区域的网格编码，利用网格编码逆算法，可得到该网格区域的顶点的位置信息。
在步骤S111中，根据目标网格区域的顶点的位置信息，确定位于邻居网格区域中辅助点的位置信息。
邻居网格区域与目标网格区域共用部分顶点，得到目标网格区域的顶点的位置信息，相当于得到邻居网格区域的部分顶点的位置信息，根据邻居网格区域的部分顶点的位置信息，可得到邻居网格区域中辅助点的位置信息。辅助点可为邻居网格区域中除与目标网格区域共用的顶点外的任意一点或多点，在此并不限定。可在每个邻居网格区域中确定辅助点，以便于后续利用辅助点的位置信息，确定邻居网格区域。
在步骤S112中，基于每个邻居网格区域中辅助点的位置信息和网格编码算法，计算得到每个邻居网格区域的网格编码，以确定邻居网格区域。
网格编码与网格区域具有对应关系，根据邻居网格区域中辅助点的位置信息，利用网格编码算法，计算得到的网格编码为邻居网格区域的网格编码。利用网格编码与网格区域的对应关系，可确定邻居网格区域。
在步骤S113中，获取目标网格区域的网格编码。
在步骤S114中，根据目标网格区域的网格编码，获取候选网格区域的网格编码。
在一些示例中，邻近的网格区域的网格编码的一部分数位的字符是相同的，可利用该特征在大量的网格区域中筛选出目标网格区域邻近的网格区域即候选网格区域。候选网格区域包括网格编码中一部分数位的字符与目标网格区域的网格编码中一部分数位的字符相同的网格区域。例如，邻 近的网格区域的网格编码的前m1个数位的字符相同，可将网格编码的前m1个数位的字符与目标网格区域的网格编码的前m1个数位的字符相同的网格区域确定为候选网格区域。
在步骤S115中，按照网格编码算法中的网格区域排布与编码数位的字符的对应关系，在候选网格区域的网格编码中确定邻居网格区域的网格编码，以确定邻居网格区域。
网格编码算法中可包括网格区域排布与编码数位的字符的对应关系。例如，网格区域的排布如图2所示，网格编码为7位的字符串，候选网格区域的网格编码的前6位的字符与目标网格区域的网格编码的前6位的字符相同，目标网格区域为网格区域A5，其网格编码为wk2vu1D，网格编码算法中网格区域排布与网格编码的最后一位的字符的对应关系具体实现为如图4所示的编码表，则可知目标网格区域具有8个邻居网格区域，8个邻居网格区域分别位于目标网格区域的左上、上、右上、左、右、左下、下、右下，按照图4所示的编码表，位于字符D的左上、上、右上、左、右、左下、下、右下的字符分别为W、E、R、S、F、X、C、V，对应地，位于目标网格区域的左上、上、右上、左、右、左下、下、右下的8个邻居网格区域，即网格区域A1、网格区域A2、网格区域A3、网格区域A4、网格区域A6、网格区域A7、网格区域A8、网格区域A9的网格编码分别为wk2vu1W、wk2vu1E、wk2vu1R、wk2vu1S、wk2vu1F、wk2vu1X、wk2vu1C、wk2vu1V。
网格编码表征网格区域，确定邻居网格区域的网格编码，即可确定邻居网格区域。
利用网格区域排布与编码数位的字符的对应关系来确定邻居网格区域的方式更为简便，耗时更短，效率更高。
在一些实施例中，目标相似度可基于与门店名称相关的相似度、与门店位置信息相关的相似度综合得到。图5为本申请又一实施例提供的门店去重处理方法的流程图。图5与图1的不同之处在于，图1中的步骤S104可具体细化为图5中的步骤S1041至步骤S1043。
在步骤S1041中，基于第一门店名称和第二门店名称，得到目标门店 与位于目标网格区域和邻居网格区域的存量门店的N个名称相关相似度。
N为大于等于1的整数。名称相关相似度为与门店名称相关的相似度，可基于第一门店名称和第二门店名称得到。名称相关相似度可包括但不限于字符相似度、语义相似度、门店类型相似度中的任意一种或两种以上。字符相似度为组成门店名称的字符的相似度。语义相似度为门店名称的语义的相似度。门店类型相似度为基于门店名称得到的门店类型的相似度。
在一些示例中，名称相关相似度包括字符相似度。可对第一门店名称和第二门店名称分别进行分词，得到第一门店名称对应的词汇和第二门店名称对应的词汇；计算第一门店名称对应的词汇和第二门店名称对应的词汇的词频(Term Frequency，TF)和逆文本频率指数(Inverse Document Frequency，IDF)；选取词频低于等于冗余词频阈值且逆文本频率指数大于冗余频率指数阈值的词汇；基于选取的第一门店名称对应的词汇和选取的第二门店名称对应的词汇，得到目标门店与位于目标网格区域和邻居网格区域的存量门店的字符相似度。
可利用分词工具对第一门店名称进行切分，得到第一门店名称对应的词汇；利用分词工具对第二门店名称进行切分，得到第二门店名称对应的词汇。词频表征词汇出现的频率。逆文本频率指数用于表征词汇具有的区分能力。冗余词频阈值为用于区分词汇是否为冗余词汇的词频的阈值。冗余频率指数阈值为用于区分词汇是否为冗余词汇的逆文本频率指数的阈值。若某词汇的词频大于冗余词频阈值，表示该词汇为冗余词汇；若某词汇的逆文本频率指数小于等于冗余频率指数阈值，表示该词汇为冗余词汇。冗余词汇对字符相似度的运算没有帮助，甚至可能会有不良影响，不需参与字符相似度的运算。词频低于等于冗余词频阈值且逆文本频率指数大于冗余频率指数阈值的词汇为参与字符相似度运算的有效词汇。字符相似度运算可参考机器翻译所使用的双语评估研究(Bilingual Evaluation Understudy，BLEU)算法，通过选取的第一门店名称对应的词汇和第二门店名称对应的词汇间的N-gram重合度来评价第一门店名称和第二名称在字符方面的相似性。
在一些示例中，名称相关相似度包括语义相似度。将第一门店名称和第二门店名称分别转化为第一名称数字序列和第二名称数字序列；将第一名称数字序列和第二名称数字序列输入第一模型，得到第一模型输出的目标门店与位于目标网格区域和邻居网格区域的存量门店的语义相似度。
第一模型用于根据输入的两个门店名称转化为的数字序列输出两个门店名称的语义相似度。可预先获取一定数量的具有标注的门店名称作为训练集正样本，随机抽取数量相当的门店名称作为训练集负样本，将训练集正样本和训练集负样本分别转换为数字序列，利用数字序列训练得到第一模型。第一模型可包括分类模型，可为深度学习分类模型或其他类型的分类模型，在此并不限定。例如，可利用BERT(即BidirectionalEncoder Representations from Transformer)模型，将“[CLS]+某一门店名称对应的数字序列+[SEP]+另一门店名称对应的数字序列”作为输入，训练第一模型，使第一模型可拟合一门店名称与另一门店名称的语义相似度，即，使第一模型可根据输入输出一门店名称与另一门店名称的语义相似度。
第一名称数字序列为第一门店名称转化为的数字序列。第二名称数字序列为第二门店名称转化为的数字序列。具体可将门店名称按字分割，将分割得到的字转化为数字，将每个字对应的数字组合，得到数字序列。将第一名称数字序列和位于目标网格区域和邻居网格区域的一个存量门店对应的第二名称数字序列输入第一模型，第一模型可输出目标门店的门店名称与这一个存量门店的门店名称的语义相似度。
在一些示例中，名称相关相似度包括门店类型相似度。在门店去重处理过程中可能会出现门店为连锁店且距离较近、不同门店名称类似所产生的误去重的可能，为了降低甚至避免误去重的可能，可引入门店类型相似度来提高门店去重的准确性。可根据第一门店名称，得到第一门店名称信息；将第一门店名称信息输入第二模型，得到第二模型输出的目标门店的门店类型概率向量；在存量数据库中查找与第二门店名称对应的门店类型概率向量；计算目标门店的门店类型概率向量与第二门店名称对应的门店类型概率向量的相似度，将相似度确定为目标门店与位于目标网格区域和邻居网格区域的存量门店的门店类型相似度。
第二模型用于根据输入的门店名称信息输出门店类型概率向量。门店类型概率向量用于表征门店名称指示的门店属于各门店类型的概率。门店类型概率向量中的每个元素可表征门店属于一门店类型的概率，可将门店类型概率向量中表征的概率最大元素对应的门店类型确定为该门店的门店类型。门店类型概率向量可为长度为M的归一化向量，但并不限于此。可预先获取一定数量的具有标注的门店名称和门店类型作为训练集，如<XXXX1(B1地区店)，超市>、<YYYY2(B2地区店)，咖啡厅>，其中，XXXX1(B1地区店)和YYYY2(B2地区店)为门店名称，超市和咖啡厅为门店类型。利用训练集训练得到第二模型。第二模型可包括分类模型，可为深度学习分类模型或其他类型的分类模型，在此并不限定。例如，可利用BERT模型，将“[CLS]+某一门店名称对应的数字序列”作为输入，训练第二模型，使第二模型可拟合该门店名称与门店类型之间的对应关系，即，使第二模型可根据输入输出该门店名称的门店类型概率向量。
第一门店名称信息基于第一门店名称得到，可为第一门店名称，也可为第一门店名称经处理后的信息，如数字序列，门店名称转化为数字序列的方式可参见上述实施例中的相关说明，在此不再赘述。第二门店名称对应的门店类型概率向量包括位于目标网格区域和邻居网格区域的存量门店对应的门店类型概率向量。在一些示例中，目标门店的门店类型概率向量与第二门店名称对应的门店类型概率向量的相似度可为两门店类型概率向量的余弦相似度。
为了进一步缩短门店去重处理所需的时间，可预先根据各存量门店的门店名称，得到存量门店的门店类型概率向量，以便于需要计算门店类型相似度时，直接从存量门店数据库中获取。具体地，可获取存量门店的门店名称，根据门店名称，得到门店名称信息；将存量门店的门店名称信息输入第二模型，得到第二模型输出的存量门店的门店类型概率向量；建立存量门店和存量门店的门店类型概率向量的第二对应关系，并将第二对应关系存储于存量门店数据库。在计算门店类型相似度时，可根据第二对应关系，在存量门店数据库中查找得到第二门店名称对应的门店类型概率向量。
在步骤S1042中，基于第一门店位置信息和第二门店位置信息，得到目标门店与位于目标网络区域和邻居网格区域的存量门店的位置相似度。
位置相似度为与门店位置信息相关的相似度，可基于第一门店位置信息和第二门店位置信息得到。位置相似度可根据两个门店位置信息指示的两个门店位置之间的距离和位置信息可能导致的偏差量确定。具体地，可根据第一门店位置信息和第二门店位置信息，得到目标门店与存量门店的地理距离；根据地理距离和位置偏差阈值的比值，得到目标门店与位于目标网络区域和邻居网格区域的存量门店的位置相似度。第一门店位置信息和第二门店位置信息可为定位坐标信息，如全球定位系统(Global Positioning System，GPS)坐标信息。若第一门店位置信息和第二门店位置信息为地址信息，则可将地址信息转换为坐标信息，如经纬度信息，再根据坐标信息确定目标门店与存量门店的地理距离。位置偏差阈值可为位置信息可能导致的偏差量的最大值。可利用地理距离和位置偏差阈值的比值进行归一化，从而得到位置相似度。例如，位置相似度可根据下式(1)得到：
在步骤S1043中，根据N个名称相关相似度、位置相似度以及对应的权重系数，计算得到目标相似度。
权重系数可作为指数或乘积系数参与目标相似度的计算，在此并不限定。在一些示例中，权重系数可作为指数参与目标相似度的计算，例如，名称相关相似度包括字符相似度、语义相似度和门店类型相似度，则目标相似度可根据下式(2)得到：
sim(目标门店，存量门店)＝sim(字符)α×sim(语义)β×sim(类型)γ×sim(位置）δ  
(2)
其中，sim(目标门店，存量门店)为目标相似度；sim(字符)为字符相似度；sim(语义)为语义相似度；sim(类型)为门店类型相似度；sim(位置)为位置相似度；α为字符相似度的权重系数；β为语义相似度的权重系数；γ为门店类型相似度的权重系数；δ为位置相似度的权重系数。在一些示例中，为了方便计算，可使α＝β＝γ＝δ＝1。
为了便于理解，下面以一示例对门店去重处理方法进行说明。在该示例中，名称相关相似度包括字符相似度、语义相似度和门店类型相似度。
获取目标门店的门店名称和门店地址，将门店地址转换为经纬度坐标，转换得到的经纬度坐标为{30.193，120.173}。利用网格编码算法，计算得到目标门店所在网格区域即目标网格区域的网格编码为wtm7y8e。邻居网格区域的网格编码的前6位字符与目标网格区域的网格编码的前6位字符相同，可利用如图4所示的编码表得到8个邻居网格区域的网格编码。8个邻居网格区域的网格编码分别为wtm7y82、wtm7y83、wtm7y84、wtm7y8W、wtm7y8R、wtm7y8S、wtm7y8D和wtm7y8F。在存量门店数据库中查询，确定目标网格区域中具有158个存量门店，网格编码为wtm7y82的邻居网格区域中具有0个存量门店，网格编码为wtm7y83的邻居网格区域中具有4个存量门店，网格编码为wtm7y84的邻居网格区域中具有1个存量门店，网格编码为wtm7y8W的邻居网格区域中具有0个存量门店，网格编码为wtm7y8R的邻居网格区域中具有18个存量门店，网格编码为wtm7y8S的邻居网格区域中具有1个存量门店，网格编码为wtm7y8D的邻居网格区域中具有0个存量门店，网格编码为wtm7y8F的邻居网格区域中具有0个存量门店。即，目标网格区域和邻居网格区域中共具有181个存量门店。需计算得到目标门店与目标网格区域和邻居网格区域中每一个存量门店的目标相似度。
下面以目标门店与其中一个存量门店的目标相似度的计算为例进行说明。目标门店的门店名称为“X1X2(杭州市滨江宝龙城市广场店)”，存量门店名称为“杭州市滨江区X3X4便利店”，其中，X1、X2、X3和X4均为汉字，且是不同的汉字。
可使用分词工具对目标门店和存量门店的门店名称进行切分，得到目标门店对应的词汇和存量门店对应的词汇。目标门店对应的词汇包括`X1X2`、`(`、`杭州市`、`滨江`、`宝龙`、`城市`、`广场`、`店`和`)`。存量门店对应的词汇包括`杭州市`、`滨江区`、`X3X4`和`便利店`。计算各词汇的词频和逆文本频率指数，上述词汇中`(`、`杭州市`和`)`的词频和逆文本频率指数不符合词频低于等于冗余词频阈值且逆文本频率指数大于冗余频 率指数阈值的条件，因此舍弃词汇`(`、`杭州市`和`)`。舍弃词汇`(`、`杭州市`和`)`后，目标门店对应的选取的词汇组合后为“X1X2滨江宝龙城市广场店”，存量门店对应的选取的词汇组合后为“滨江区X3X4便利店”。利用上述BLEU算法计算字符相似度，“X1X2滨江宝龙城市广场店”包含11个1-gram，“滨江区X3X4便利店”包含8个1-gram，分别计算两者的1-gram的共现次数，可知`滨`、`江`和`店`三个1-gram分别共现一次，因此，“X1X2滨江宝龙城市广场店”和“滨江区X3X4便利店”的字符相似度为(3/11+3/8)/2≈0.32。
可将“X1X2(杭州市滨江宝龙城市广场店)”转化为数字序列[0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]，将“杭州市滨江区X3X4便利店”转化为数字序列[3,4,5,6,7,16,17,18,19,20,14]，相同的汉字对应的数字相同。将上述两个数字序列和[CLS]以及[SEP]拼接，组合为单个向量，并输入第一模型，得到第一模型输出的两者的语义相似度。
可将“X1X2(杭州市滨江宝龙城市广场店)”和“杭州市滨江区X3X4便利店”转换得到的两个数字序列分别输入第二模型，得到目标门店的门店类型概率向量和存量门店的门店类型概率向量。目标门店和存量门店在“购物”、“超市”、“便利店”三个门店类型维度上的元素的值比较高，基于目标门店的门店类型概率向量和存量门店的门店类型概率向量得到的门店类型相似度表征的门店类型比较接近。
基于目标门店的门店位置信息和存量门店的门店位置信息，确定两者的地理距离为285米，根据该地理距离和位置偏差阈值可计算得到位置相似度为0.8585。
设去重相似度阈值为0.6，对于目标门店和存量门店，利用上述式(2)计算得到的目标相似度小于0.6，可确定目标门店和存量门店不是同一门店。
需要说明的是，本申请实施例中对信息、数据的获取、存储、使用、处理等均得到用户或相关机构的授权，符合国家法律法规的相关规定。
本申请第二方面提供一种门店去重处理装置。图6为本申请一实施例提供的门店去重处理装置的结构示意图。如图6所示，该门店去重处理装 置300可包括第一获取模块301、网格区域确定模块302、第二获取模块303、计算模块304和去重模块305。
第一获取模块301可用于获取目标门店的第一门店名称和第一门店位置信息。
网格区域确定模块302可用于根据第一门店位置信息，确定目标门店所在的目标网格区域。
第二获取模块303可用于在预存的存量门店数据库中，获取位于目标网格区域和邻居网格区域的存量门店的第二门店名称和第二门店位置信息。
邻居网格区域与目标网格区域相邻。
计算模块304可用于基于第一门店名称、第一门店位置信息、第二门店名称和第二门店位置信息，得到目标门店与位于目标网格区域和邻居网格区域的存量门店的目标相似度。
去重模块305可用于在目标相似度大于等于预设的去重相似度阈值的情况下，将目标门店作为重复门店去除。
在本申请实施例中，可根据目标门店的门店位置信息，确定目标门店所在的网格区域。网格区域为地图中划分的区域。基于数据库中位于目标门店所在的目标网格区域的存量门店、目标网格区域周边的网格区域的存量门店以及目标门店的门店名称、门店位置信息，得到目标门店与存量门店的相似度，根据该相似度判断新获取的门店是否与存量门店为同一门店，若新获取的门店与存量门店为同一门店，则认为新获取的门店为重复门店，予以去除。该去重过程不需人工参与，且利用门店的位置可缩小用于比对的存量门店的范围，提高了门店去重处理的效率。
而且，除了比对目标门店与目标网格区域中的存量门店以外，还比对目标门店与邻居网格区域中的存量门店，避免漏查位于目标网格区域的边界附近与目标门店为同一门店的存量门店，进一步提高门店去重处理的全面性和准确性。
在一些实施例中，网格区域具有网格编码。门店去重处理装置200还可包括邻居网格区域确定模块。
在一些示例中，邻居网格区域确定模块可用于：获取目标网格区域的网格编码；根据目标网格区域的网格编码和网格编码逆算法，获取目标网格区域的顶点的位置信息；根据目标网格区域的顶点的位置信息，确定位于邻居网格区域中辅助点的位置信息；基于每个邻居网格区域中辅助点的位置信息和网格编码算法，计算得到每个邻居网格区域的网格编码，以确定邻居网格区域。
在一些示例中，相邻的网格区域的网格编码中一部分数位的值相同。邻居网格区域确定模块可用于：获取目标网格区域的网格编码；根据目标网格区域的网格编码，获取候选网格区域的网格编码，候选网格区域包括网格编码中一部分数位的字符与目标网格区域的网格编码中一部分数位的字符相同的网格区域；按照网格编码算法中的网格区域排布与编码数位的字符的对应关系，在候选网格区域的网格编码中确定邻居网格区域的网格编码，以确定邻居网格区域。
在一些实施例中，门店去重装置200还可包括第一预处理模块。第一预处理模块可用于：将地图划分为多个网格区域，并利用网格编码算法，为每个网格区域分配网格编码；获取存量门店的门店位置信息，根据存量门店的门店位置信息，确定存量门店所在的网格区域；建立存量门店和存量门店所在的网格区域的网格编码的第一对应关系，并将第一对应关系存储于存量门店数据库。
在一些实施例中，计算模块304可用于：基于第一门店名称和第二门店名称，得到目标门店与位于目标网格区域和邻居网格区域的存量门店的N个名称相关相似度，N为大于等于1的整数；基于第一门店位置信息和第二门店位置信息，得到目标门店与位于目标网络区域和邻居网格区域的存量门店的位置相似度；根据N个名称相关相似度、位置相似度以及对应的权重系数，计算得到目标相似度。
在一些示例中，名称相关相似度包括字符相似度。计算模块304可用于：对第一门店名称和第二门店名称分别进行分词，得到第一门店名称对应的词汇和第二门店名称对应的词汇；计算第一门店名称对应的词汇和第二门店名称对应的词汇的词频和逆文本频率指数；选取词频低于等于冗余 词频阈值且逆文本频率指数大于冗余频率指数阈值的词汇；基于选取的第一门店名称对应的词汇和选取的第二门店名称对应的词汇，得到目标门店与位于目标网格区域和邻居网格区域的存量门店的字符相似度。
在一些示例中，名称相关相似度包括语义相似度。计算模块304可用于：将第一门店名称和第二门店名称分别转化为第一名称数字序列和第二名称数字序列；将第一名称数字序列和第二名称数字序列输入第一模型，得到第一模型输出的目标门店与位于目标网格区域和邻居网格区域的存量门店的语义相似度，第一模型用于根据输入的两个门店名称转化为的数字序列输出两个门店名称的语义相似度。
在一些示例中，名称相关相似度包括门店类型相似度。计算模块304可用于：根据第一门店名称，得到第一门店名称信息；将第一门店名称信息输入第二模型，得到第二模型输出的目标门店的门店类型概率向量，第二模型用于根据输入的门店名称信息输出门店类型概率向量，门店类型概率向量用于表征门店名称指示的门店属于各门店类型的概率；在存量数据库中查找与第二门店名称对应的门店类型概率向量；计算目标门店的门店类型概率向量与第二门店名称对应的门店类型概率向量的相似度，将相似度确定为目标门店与位于目标网格区域和邻居网格区域的存量门店的门店类型相似度。
在一些示例中，计算模块304可用于：根据第一门店位置信息和第二门店位置信息，得到目标门店与存量门店的地理距离；根据地理距离和位置偏差阈值的比值，得到目标门店与位于目标网络区域和邻居网格区域的存量门店的位置相似度。
在一些实施例中，门店去重处理装置还可包括第二预处理模块。第二预处理模块可用于：获取存量门店的门店名称，根据门店名称，得到门店名称信息；将存量门店的门店名称信息输入第二模型，得到第二模型输出的存量门店的门店类型概率向量；建立存量门店和存量门店的门店类型概率向量的第二对应关系，并将第二对应关系存储于存量门店数据库。
本申请第三方面提供一种门店去重处理设备。图7为本申请一实施例提供的门店去重处理设备的结构示意图。如图7所示，门店去重处理设备 400包括存储器401、处理器402及存储在存储器401上并可在处理器402上运行的计算机程序。
在一些示例中，上述处理器402可以包括中央处理器(CPU)，或者特定集成电路(Application Specific Integrated Circuit，ASIC)，或者可以被配置成实施本申请实施例的一个或多个集成电路。
存储器401可包括只读存储器(Read-Only Memory，ROM)，随机存取存储器(Random Access Memory，RAM)，磁盘存储介质设备，光存储介质设备，闪存设备，电气、光学或其他物理/有形的存储器存储设备。因此，通常，存储器包括一个或多个编码有包括计算机可执行指令的软件的有形(非暂态)计算机可读存储介质(例如，存储器设备)，并且当该软件被执行(例如，由一个或多个处理器)时，其可操作来执行参考根据本申请实施例中门店去重处理方法所描述的操作。
处理器402通过读取存储器401中存储的可执行程序代码来运行与可执行程序代码对应的计算机程序，以用于实现上述实施例中的门店去重处理方法。
在一些示例中，门店去重处理设备400还可包括通信接口403和总线404。其中，如图7所示，存储器401、处理器402、通信接口403通过总线404连接并完成相互间的通信。
通信接口403，主要用于实现本申请实施例中各模块、装置、单元和/或设备之间的通信。也可通过通信接口403接入输入设备和/或输出设备。
总线404包括硬件、软件或两者，将门店去重处理设备400的部件彼此耦接在一起。举例来说而非限制，总线404可包括加速图形端口(Accelerated Graphics Port，AGP)或其他图形总线、增强工业标准架构(Enhanced Industry Standard Architecture，EISA)总线、前端总线(Front Side Bus，FSB)、超传输(Hyper Transport，HT)互连、工业标准架构(Industry Standard Architecture，ISA)总线、无限带宽互连、低引脚数(Low pin count，LPC)总线、存储器总线、微信道架构(Micro Channel Architecture，MCA)总线、外围组件互连(Peripheral Component Interconnect，PCI)总线、PCI-Express(PCI-E)总线、串行高级技术附件 (Serial Advanced Technology Attachment，SATA)总线、视频电子标准协会局部(Video Electronics Standards Association Local Bus，VLB)总线或其他合适的总线或者两个或更多个以上这些的组合。在合适的情况下，总线404可包括一个或多个总线。尽管本申请实施例描述和示出了特定的总线，但本申请考虑任何合适的总线或互连。
本申请第四方面提供一种计算机可读存储介质，该计算机可读存储介质上存储有计算机程序指令，该计算机程序指令被处理器执行时可实现上述实施例中的门店去重处理方法，且能达到相同的技术效果，为避免重复，这里不再赘述。其中，上述计算机可读存储介质可包括非暂态计算机可读存储介质，如只读存储器(Read-Only Memory，简称ROM)、随机存取存储器(Random Access Memory，简称RAM)、磁碟或者光盘等，在此并不限定。
本申请实施例提供一种计算机程序产品，该计算机程序产品中的指令由电子设备的处理器执行时，使得电子设备可执行上述实施例中的门店去重处理方法，且能达到相同的技术效果，为避免重复，这里不再赘述。
需要明确的是，本说明书中的各个实施例均采用递进的方式描述，各个实施例之间相同或相似的部分互相参见即可，每个实施例重点说明的都是与其他实施例的不同之处。对于装置实施例、设备实施例、计算机可读存储介质实施例、计算机程序产品实施例而言，相关之处可以参见方法实施例的说明部分。本申请并不局限于上文所描述并在图中示出的特定步骤和结构。本领域的技术人员可以在领会本申请的精神之后，作出各种改变、修改和添加，或者改变步骤之间的顺序。并且，为了简明起见，这里省略对已知方法技术的详细描述。
上面参考根据本申请的实施例的方法、装置(系统)和计算机程序产品的流程图和/或框图描述了本申请的各方面。应当理解，流程图和/或框图中的每个方框以及流程图和/或框图中各方框的组合可以由计算机程序指令实现。这些计算机程序指令可被提供给通用计算机、专用计算机、或其它可编程数据处理装置的处理器，以产生一种机器，使得经由计算机或其它可编程数据处理装置的处理器执行的这些指令使能对流程图和/或框图的 一个或多个方框中指定的功能/动作的实现。这种处理器可以是但不限于是通用处理器、专用处理器、特殊应用处理器或者现场可编程逻辑电路。还可理解，框图和/或流程图中的每个方框以及框图和/或流程图中的方框的组合，也可以由执行指定的功能或动作的专用硬件来实现，或可由专用硬件和计算机指令的组合来实现。
本领域技术人员应能理解，上述实施例均是示例性而非限制性的。在不同实施例中出现的不同技术特征可以进行组合，以取得有益效果。本领域技术人员在研究附图、说明书及权利要求书的基础上，应能理解并实现所揭示的实施例的其他变化的实施例。在权利要求书中，术语“包括”并不排除其他装置或步骤；数量词“一个”不排除多个；术语“第一”、“第二”用于标示名称而非用于表示任何特定的顺序。权利要求中的任何附图标记均不应被理解为对保护范围的限制。权利要求中出现的多个部分的功能可以由一个单独的硬件或软件模块来实现。某些技术特征出现在不同的从属权利要求中并不意味着不能将这些技术特征进行组合以取得有益效果。

Claims (13)

  1. 一种门店去重处理方法，包括：
    获取目标门店的第一门店名称和第一门店位置信息；
    根据所述第一门店位置信息，确定所述目标门店所在的目标网格区域；
    在预存的存量门店数据库中，获取位于所述目标网格区域和邻居网格区域的存量门店的第二门店名称和第二门店位置信息，所述邻居网格区域与所述目标网格区域相邻；
    基于所述第一门店名称、所述第一门店位置信息、所述第二门店名称和所述第二门店位置信息，得到所述目标门店与位于所述目标网格区域和所述邻居网格区域的存量门店的目标相似度；
    在所述目标相似度大于等于预设的去重相似度阈值的情况下，将所述目标门店作为重复门店去除。
  2. 根据权利要求1所述的方法，其中，网格区域具有网格编码，
    在所述在预存的存量门店数据库中，获取位于所述目标网格区域和邻居网格区域的存量门店的第二门店名称和第二门店位置信息之前，还包括：
    获取所述目标网格区域的网格编码；
    根据所述目标网格区域的网格编码和网格编码逆算法，获取所述目标网格区域的顶点的位置信息；
    根据所述目标网格区域的顶点的位置信息，确定位于所述邻居网格区域中辅助点的位置信息；
    基于每个所述邻居网格区域中辅助点的位置信息和网格编码算法，计算得到每个所述邻居网格区域的网格编码，以确定所述邻居网格区域。
  3. 根据权利要求1所述的方法，其中，网格区域具有网格编码，相邻的网格区域的网格编码中一部分数位的值相同，
    在所述在预存的存量门店数据库中，获取位于所述目标网格区域和邻居网格区域的存量门店的第二门店名称和第二门店位置信息之前，还包 括：
    获取所述目标网格区域的网格编码；
    根据所述目标网格区域的网格编码，获取候选网格区域的网格编码，所述候选网格区域包括网格编码中一部分数位的字符与所述目标网格区域的网格编码中一部分数位的字符相同的网格区域；
    按照网格编码算法中的网格区域排布与编码数位的字符的对应关系，在所述候选网格区域的网格编码中确定所述邻居网格区域的网格编码，以确定所述邻居网格区域。
  4. 根据权利要求1所述的方法，还包括：
    将地图划分为多个网格区域，并利用网格编码算法，为每个网格区域分配网格编码；
    获取所述存量门店的门店位置信息，根据所述存量门店的门店位置信息，确定所述存量门店所在的网格区域；
    建立所述存量门店和存量门店所在的网格区域的网格编码的第一对应关系，并将第一对应关系存储于所述存量门店数据库。
  5. 根据权利要求1所述的方法，其中，所述基于所述第一门店名称、所述第一门店位置信息、所述第二门店名称和所述第二门店位置信息，得到所述目标门店与位于所述目标网格区域和所述邻居网格区域的存量门店的目标相似度，包括：
    基于所述第一门店名称和所述第二门店名称，得到所述目标门店与位于所述目标网格区域和所述邻居网格区域的存量门店的N个名称相关相似度，N为大于等于1的整数；
    基于所述第一门店位置信息和所述第二门店位置信息，得到所述目标门店与位于所述目标网络区域和所述邻居网格区域的所述存量门店的位置相似度；
    根据N个所述名称相关相似度、所述位置相似度以及对应的权重系数，计算得到所述目标相似度。
  6. 根据权利要求5所述的方法，其中，所述名称相关相似度包括字符相似度，
    所述基于所述第一门店名称和所述第二门店名称，得到所述目标门店与位于所述目标网格区域和所述邻居网格区域的存量门店的N个名称相关相似度，包括：
    对所述第一门店名称和所述第二门店名称分别进行分词，得到所述第一门店名称对应的词汇和所述第二门店名称对应的词汇；
    计算所述第一门店名称对应的词汇和所述第二门店名称对应的词汇的词频和逆文本频率指数；
    选取词频低于等于冗余词频阈值且逆文本频率指数大于冗余频率指数阈值的词汇；
    基于选取的所述第一门店名称对应的词汇和选取的所述第二门店名称对应的词汇，得到所述目标门店与位于所述目标网格区域和所述邻居网格区域的所述存量门店的所述字符相似度。
  7. 根据权利要求5所述的方法，其中，所述名称相关相似度包括语义相似度，
    所述基于所述第一门店名称和所述第二门店名称，得到所述目标门店与位于所述目标网格区域和所述邻居网格区域的存量门店的N个名称相关相似度，包括：
    将所述第一门店名称和所述第二门店名称分别转化为第一名称数字序列和第二名称数字序列；
    将所述第一名称数字序列和所述第二名称数字序列输入第一模型，得到所述第一模型输出的所述目标门店与位于所述目标网格区域和所述邻居网格区域的所述存量门店的语义相似度，所述第一模型用于根据输入的两个门店名称转化为的数字序列输出两个门店名称的语义相似度。
  8. 根据权利要求5所述的方法，其中，所述名称相关相似度包括门店类型相似度，
    所述基于所述第一门店名称和所述第二门店名称，得到所述目标门店与位于所述目标网格区域和所述邻居网格区域的存量门店的N个名称相关相似度，包括：
    根据第一门店名称，得到第一门店名称信息；
    将所述第一门店名称信息输入第二模型，得到所述第二模型输出的所述目标门店的门店类型概率向量，所述第二模型用于根据输入的门店名称信息输出门店类型概率向量，门店类型概率向量用于表征门店名称指示的门店属于各门店类型的概率；
    在所述存量数据库中查找与所述第二门店名称对应的门店类型概率向量；
    计算所述目标门店的门店类型概率向量与所述第二门店名称对应的门店类型概率向量的相似度，将所述相似度确定为所述目标门店与位于所述目标网格区域和所述邻居网格区域的存量门店的所述门店类型相似度。
  9. 根据权利要求8所述的方法，还包括：
    获取所述存量门店的门店名称，根据门店名称，得到门店名称信息；
    将所述存量门店的门店名称信息输入所述第二模型，得到所述第二模型输出的所述存量门店的门店类型概率向量；
    建立所述存量门店和所述存量门店的门店类型概率向量的第二对应关系，并将第二对应关系存储于所述存量门店数据库。
  10. 根据权利要求5所述的方法，其中，所述基于所述第一门店位置信息和所述第二门店位置信息，得到所述目标门店与位于所述目标网络区域和所述邻居网格区域的所述存量门店的位置相似度，包括：
    根据所述第一门店位置信息和所述第二门店位置信息，得到所述目标门店与所述存量门店的地理距离；
    根据所述地理距离和位置偏差阈值的比值，得到所述目标门店与位于所述目标网络区域和所述邻居网格区域的所述存量门店的所述位置相似度。
  11. 一种门店去重处理装置，包括：
    第一获取模块，用于获取目标门店的第一门店名称和第一门店位置信息；
    网格区域确定模块，用于根据所述第一门店位置信息，确定所述目标门店所在的目标网格区域；
    第二获取模块，用于在预存的存量门店数据库中，获取位于所述目标 网格区域和邻居网格区域的存量门店的第二门店名称和第二门店位置信息，所述邻居网格区域与所述目标网格区域相邻；
    计算模块，用于基于所述第一门店名称、所述第一门店位置信息、所述第二门店名称和所述第二门店位置信息，得到所述目标门店与位于所述目标网格区域和所述邻居网格区域的存量门店的目标相似度；
    去重模块，用于在所述目标相似度大于等于预设的去重相似度阈值的情况下，将所述目标门店作为重复门店去除。
  12. 一种门店去重处理设备，包括：处理器以及存储有计算机程序指令的存储器；
    所述处理器执行所述计算机程序指令时实现如权利要求1至10中任意一项所述的门店去重处理方法。
  13. 一种计算机可读存储介质，所述计算机可读存储介质上存储有计算机程序指令，所述计算机程序指令被处理器执行时实现如权利要求1至10中任意一项所述的门店去重处理方法。
PCT/CN2023/074805 2022-08-10 2023-02-07 门店去重处理方法、装置、设备及存储介质 WO2024031943A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210957641.0A CN115392955B (zh) 2022-08-10 2022-08-10 门店去重处理方法、装置、设备及存储介质
CN202210957641.0 2022-08-10

Publications (1)

Publication Number Publication Date
WO2024031943A1 true WO2024031943A1 (zh) 2024-02-15

Family

ID=84119080

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/074805 WO2024031943A1 (zh) 2022-08-10 2023-02-07 门店去重处理方法、装置、设备及存储介质

Country Status (3)

Country Link
CN (1) CN115392955B (zh)
TW (1) TWI847497B (zh)
WO (1) WO2024031943A1 (zh)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115392955B (zh) * 2022-08-10 2024-03-01 中国银联股份有限公司 门店去重处理方法、装置、设备及存储介质
CN115797109B (zh) * 2023-01-30 2023-05-05 北京四方启点科技有限公司 一种不同供应商的酒店聚合方法、装置及系统

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020167523A1 (en) * 1999-07-16 2002-11-14 Taylor Ralph Clayton Pixel engine
CN105320657A (zh) * 2014-05-30 2016-02-10 中国电信股份有限公司 兴趣点数据融合方法和系统
CN109993184A (zh) * 2017-12-30 2019-07-09 华为技术有限公司 一种数据融合的方法以及数据融合设备
CN113642310A (zh) * 2021-07-14 2021-11-12 广州市玄武无线科技股份有限公司 一种终端数据相似度度量方法
CN113761867A (zh) * 2020-12-29 2021-12-07 京东城市(北京)数字科技有限公司 地址识别方法、装置、计算机设备及存储介质
CN114328656A (zh) * 2021-12-17 2022-04-12 中国银联股份有限公司 真实门店的验证方法、装置、设备及存储介质
CN115392955A (zh) * 2022-08-10 2022-11-25 中国银联股份有限公司 门店去重处理方法、装置、设备及存储介质

Family Cites Families (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TW201501056A (zh) * 2013-06-28 2015-01-01 Oriental Inst Technology 具有開店資訊ｌｂｓ系統
CN103559213B (zh) * 2013-10-10 2016-08-17 河南大学 一种公路网高效空间最近邻查询方法
TW201612836A (en) * 2014-09-26 2016-04-01 Mercuries Data Systems Ltd Method of fast estimation for developing a store in a city in accordance with magnetic distance
WO2017075268A1 (en) * 2015-10-27 2017-05-04 Abbott Laboratories Troponin i and soluble urokinase receptor detection for determining the risk of cardiovascular disease
CN107370702B (zh) * 2016-05-11 2021-08-31 北京三星通信技术研究有限公司 一种通信系统中的信号发射、接收方法和装置
CN109947876B (zh) * 2018-07-19 2021-05-11 北京大学 基于地理网格的倒损房屋统计方法
KR102060991B1 (ko) * 2019-04-19 2019-12-31 한국전력공사 상업시설의 영업정보 제공 시스템 및 방법
CN110399569A (zh) * 2019-07-19 2019-11-01 银联智策顾问（上海）有限公司 一种基于大数据评估土地价值的方法及评估装置
CN111597254B (zh) * 2020-04-14 2023-07-21 口碑(上海)信息技术有限公司 资源数据的共享方法、装置及设备
CN113806456A (zh) * 2020-06-16 2021-12-17 华为技术有限公司 网格编码方法和装置
CN113220395A (zh) * 2021-05-07 2021-08-06 王云鹏 一种基于地图显示的博物馆互动方法及系统
CN113706222B (zh) * 2021-09-18 2023-12-22 口碑(上海)信息技术有限公司 一种门店选址的方法、装置
CN114757712A (zh) * 2022-04-29 2022-07-15 维沃移动通信有限公司 选址的推荐方法、装置、电子设备和可读存储介质

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020167523A1 (en) * 1999-07-16 2002-11-14 Taylor Ralph Clayton Pixel engine
CN105320657A (zh) * 2014-05-30 2016-02-10 中国电信股份有限公司 兴趣点数据融合方法和系统
CN109993184A (zh) * 2017-12-30 2019-07-09 华为技术有限公司 一种数据融合的方法以及数据融合设备
CN113761867A (zh) * 2020-12-29 2021-12-07 京东城市(北京)数字科技有限公司 地址识别方法、装置、计算机设备及存储介质
CN113642310A (zh) * 2021-07-14 2021-11-12 广州市玄武无线科技股份有限公司 一种终端数据相似度度量方法
CN114328656A (zh) * 2021-12-17 2022-04-12 中国银联股份有限公司 真实门店的验证方法、装置、设备及存储介质
CN115392955A (zh) * 2022-08-10 2022-11-25 中国银联股份有限公司 门店去重处理方法、装置、设备及存储介质

Also Published As

Publication number Publication date
TW202407602A (zh) 2024-02-16
TWI847497B (zh) 2024-07-01
CN115392955B (zh) 2024-03-01
CN115392955A (zh) 2022-11-25

Similar Documents

Publication Publication Date Title
WO2024031943A1 (zh) 门店去重处理方法、装置、设备及存储介质
WO2016165538A1 (zh) 一种地址数据的管理方法和装置
CN112016605B (zh) 一种基于边界框角点对齐和边界匹配的目标检测方法
CN106874287B (zh) 一种兴趣点 poi 地址编码的处理方法及装置
CN111522901B (zh) 文本中地址信息的处理方法及装置
TWI848373B (zh) 真實門店的驗證方法、裝置、設備及存儲介質
CN113723330A (zh) 一种图表文档信息理解的方法及系统
CN112395854B (zh) 一种标准要素的一致性检验方法
CN110175219A (zh) 一种k12阶段重复学校识别方法、装置、设备及存储介质
CN113723466B (zh) 文本相似度量化方法、设备及系统
CN115545122B (zh) 对象匹配方法、装置、设备、系统、介质及程序产品
CN113868379B (zh) 面向开放域问答的段落选择方法、装置、设备及存储介质
CN115659984A (zh) 合同标的物识别方法、装置以及计算机可读存储介质
CN115982475A (zh) 确定Wi-Fi与商家POI对应关系的方法及装置
CN111291155A (zh) 基于文本相似度的同名小区辨别方法及系统
CN115953224A (zh) 商户匹配方法、装置、设备及存储介质
CN118171727B (zh) 三元组的生成方法、装置、设备、介质及程序产品
CN116051164B (zh) 建设成本测算方法、装置、设备及可读存储介质
CN111402012B (zh) 一种基于迁移学习的电商缺陷产品识别方法
CN112910674B (zh) 物理站点筛选方法、装置、电子设备及存储介质
CN110083679B (zh) 搜索请求的处理方法、装置、电子设备和存储介质
CN118051646A (zh) 门店名称重复检测方法和门店重复检测方法
CN116363675A (zh) 三维模型的敏感词检替方法、装置、电子设备和存储介质
Ghongade et al. A Novel Method for Validating Addresses Using String Distance Metrics
CN116455866A (zh) 对象间wifi关系的生成方法、装置、设备和存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23851179

Country of ref document: EP

Kind code of ref document: A1