An administrative region prediction method and device based on multiple address information

Abstract

The application aims to provide an administrative region prediction method and device based on multiple address information. Compared with the prior art, the application obtains multiple address information, performs similarity screening on the multiple address information according to a preset similarity threshold, and determines several address information in the preset similarity threshold range as a prediction address set, splits each address information in the prediction address set into multiple address words according to a preset address keyword library, and determines a corresponding administrative region based on the multiple address words, a preset level keyword library and a preset administrative region library. Through conditional fusion of multiple address information, the application realizes low-information-conflict administrative region prediction under the premise of fusing as many address information as possible, and greatly improves the accuracy of the prediction result.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, in particular to an administrative region prediction technology based on multiple address information. BACKGROUND

[0002] On October 13, 2021, the People's Bank of China website released the "Notice of the People's Bank of China on Strengthening the Management of Payment Receiving Terminals and Related Businesses (Yinfa

[2021] No. 259)", which strengthens the management of payment receiving terminals and related businesses. Among them, the document requires that the receiving terminal and the clearing institution cooperate to verify the consistency of the actual location of the entity special merchant receiving terminal and the registered geographic location of the terminal.

[0003] Here, the actual location of the receiving terminal can be obtained through the GPS information collected when the transaction occurs, while the registered location of the receiving terminal needs to be inferred and predicted according to the address information in the terminal and merchant management platform. The address information mainly includes terminal installation address, merchant operating address, and merchant operating name, etc. Therefore, based on the above needs, a method is needed to predict the administrative region of the entity special merchant based on multiple address information.

[0004] The prior application with application number 202210602770.8 discloses a method and device for determining an administrative region based on address information, which can realize administrative region prediction based on multiple address information of entity special merchants. However, this application extracts the sub-addresses in each address information and directly uses several sub-addresses as the total address to determine the administrative region based on the total address, that is, it uses a direct address fusion method.

[0005] The direct address fusion method assumes that all address information is consistent, that is, it is assumed that these address information describes the same address from different perspectives. However, in actual scenarios, address information may be similar or completely different. For example, the registered address and the operating address of the same merchant are both address information of the merchant, but the registered address and the operating address may be completely different. In this case, if the registered address and the operating address are directly fused, it will cause information conflict and reduce the accuracy of the determined administrative region. SUMMARY

[0006] The purpose of the present application is to provide an administrative region prediction method and device based on multiple address information to solve the adverse effects of information conflict in multiple address information on the prediction result, and at the same time, improve the accuracy of the prediction result on the premise of fusing as much address information as possible.

[0007] According to one aspect of the present application, an administrative region prediction method based on multiple address information is provided, wherein the method comprises:

[0008] obtain a plurality of address information;

[0009] perform similarity filtering on the plurality of address information according to a preset similarity threshold, and determine a plurality of address information within the preset similarity threshold range as a predicted address set;

[0010] split each address information in the predicted address set into a plurality of address words according to a preset address keyword library;

[0011] determine a corresponding administrative region based on the plurality of address words, a preset level keyword library, and a preset administrative region library.

[0012] Further, wherein the determining a plurality of address information within the preset similarity threshold range as a predicted address set further comprises:

[0013] clean up redundant information in the predicted address set;

[0014] The splitting each address information in the predicted address set into a plurality of address words according to a preset address keyword library comprises:

[0015] splitting each address information in the cleaned predicted address set into a plurality of address words according to a preset address keyword library;

[0016] Further, wherein the obtaining a plurality of address information comprises:

[0017] obtaining a plurality of address information containing level marks, wherein the level marks are used to represent the reliability of the address information;

[0018] The performing similarity filtering on the plurality of address information according to a preset similarity threshold, and determining a plurality of address information within the preset similarity threshold range as a predicted address set comprises:

[0019] sorting each address information in order from high to low according to the reliability represented by the level marks, traversing the sorting from the second address information, and performing similarity comparison between the traversed address information and the address information of a higher level of reliability;

[0020] If the similarity comparison result is within the preset similarity threshold range, the traversed address information is added to the predicted address set, and the traversal is continued;

[0021] If the similarity comparison result is completely consistent, the traversed address information is deleted and the traversal is continued;

[0022] If it is not within the preset similarity threshold range and not completely consistent, the traversed address information is deleted and the traversal is ended.

[0023] Further, each level mark has a corresponding similarity threshold, and the higher the reliability represented by the level mark, the lower the corresponding similarity threshold.

[0024] Further, the cleaning of the redundant information in the predicted address set further comprises:

[0025] A plurality of cleaning rules are set by regular expressions;

[0026] The cleaning of the redundant information in the predicted address set comprises:

[0027] The predicted address set is traversed by each cleaning rule;

[0028] When the address information traversed meets the cleaning rule, the address information is deleted of redundant information according to the regular expression content of the cleaning rule.

[0029] Further, the redundant information includes at least any one of a zip code, a positioning point information, a non-Chinese character, and a purpose information, wherein the cleaning rule corresponds to different types of redundant information one by one;

[0030] When the redundant information includes at least a zip code and a non-Chinese character, the predicted address set is traversed by each cleaning rule, comprising:

[0031] The order of traversing the predicted address set by the cleaning rule corresponding to the zip code is prior to the order of traversing the predicted address set by the cleaning rule corresponding to the non-Chinese character.

[0032] Further, before the address information in the predicted address set is split into a plurality of address words according to a preset address keyword library, the method further comprises:

[0033] All keywords with a character length of 1 in the address keyword library are determined;

[0034] All keywords with a character length of 1 are combined two by two to obtain combined words;

[0035] The combined words are added to the address keyword library.

[0036] Further, a preset screening threshold is provided, wherein before the combined words are added to the address keyword library, the method further comprises:

[0037] The number of occurrences of each combined word in the administrative region word library is counted;

[0038] The combined words are added to the address keyword library, comprising:

[0039] When the occurrence number of the combination word exceeds the screening threshold, the combination word is added to the address keyword library.

[0040] Further, the splitting further comprises:

[0041] Manually checking the splitting result, and if the splitting result is wrong, adding the address information corresponding to the splitting result to a special place library.

[0042] Further, a province mapping library is established according to the administrative region library, wherein the province mapping library comprises mapping relationships between each prefectural administrative region word and the provincial administrative region word to which the prefectural administrative region word belongs, and the determining of the corresponding administrative region based on the plurality of address words, the preset level word library and the preset administrative region library comprises:

[0043] determining a word group set containing administrative region words of all levels based on the preset level word library and the plurality of address words;

[0044] matching the word group set in the preset administrative region library to determine a matching set corresponding to the word group set;

[0045] determining the administrative region corresponding to the matching set as the administrative region corresponding to the address information;

[0046] Further, the determining of the word group set containing administrative region words of all levels based on the preset level word library and the plurality of address words further comprises:

[0047] acquiring the prefectural administrative region word and the provincial administrative region word in each address word in the word group set;

[0048] judging whether there is a mapping relationship between the prefectural administrative region word and the provincial administrative region word according to the province mapping library;

[0049] if there is no mapping relationship, replacing the prefectural administrative region word and the provincial administrative region word in the word group set with original combination fields corresponding to the prefectural administrative region word and the provincial administrative region word.

[0050] According to another aspect of the present application, a computer readable medium having computer readable instructions stored thereon is also provided, the computer readable instructions being executable by a processor to implement the operations of the foregoing method.

[0051] According to still another aspect of the present application, an administrative region prediction device based on a plurality of address information is also provided, wherein the device comprises:

[0052] one or more processors; and

[0053] A memory storing computer readable instructions that, when executed, cause the processor to perform the operations of the above method.

[0054] Compared with the prior art, the application obtains a plurality of address information; performs similarity screening on the plurality of address information according to a preset similarity threshold, and determines a plurality of address information in the preset similarity threshold range as a predicted address set; splits each address information in the predicted address set into a plurality of address words according to a preset address keyword library; and determines a corresponding administrative region based on the plurality of address words, a preset level keyword library and a preset administrative region library. By conditionally fusing a plurality of address information, the application realizes low-information-conflict administrative region prediction under the premise of fusing as many address information as possible, and greatly improves the accuracy of the prediction result. BRIEF DESCRIPTION OF DRAWINGS

[0055] Other features, objects, and advantages of the application will become more apparent from the following detailed description of non-limiting embodiments, when read in conjunction with the accompanying drawings:

[0056] Figure 1 A flow chart of an administrative region prediction method based on a plurality of address information according to one aspect of the application is shown;

[0057] Figure 2 A flow chart of an administrative region prediction method based on a plurality of address information according to one preferred embodiment of the application is shown.

[0058] The same or similar reference signs in the drawings represent the same or similar components. DETAILED DESCRIPTION

[0059] The application is described in further detail below in conjunction with the accompanying drawings.

[0060] In one typical configuration of the application, the terminal, the device of the service network and the trusted party each include one or more processors (CPU), input / output interfaces, network interfaces and memories.

[0061] The memory can include a non-permanent memory in a computer readable medium, random access memory (RAM) and / or non-volatile memory such as read-only memory (ROM) or flash memory (flash RAM). The memory is an example of the computer readable medium.

[0062] Computer-readable media includes permanent and non-permanent, removable and non-removable media implemented in any method or technology for information storage. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible to a computing device. According to the definition herein, computer-readable media does not include non-transitory computer-readable media, such as modulated data signals and carriers.

[0063] In order to further illustrate the technical means adopted by the present application and the effects achieved, the technical solutions of the present application will be described clearly and completely in conjunction with the preferred embodiments and the accompanying drawings.

[0064] Figure 1 An administrative region prediction method based on multiple address information is shown, wherein the method comprises:

[0065] S11 acquiring multiple address information;

[0066] S12 performing similarity screening on the multiple address information according to a preset similarity threshold, and determining several address information within the preset similarity threshold range as a prediction address set;

[0067] S13 splitting each address information in the prediction address set into multiple address words according to a preset address keyword library;

[0068] S14 determining a corresponding administrative region based on the multiple address words, a preset level keyword library and a preset administrative region library.

[0069] In this embodiment, in the step S11, multiple address information is acquired.

[0070] In this embodiment, the plurality of address information can be obtained from the merchant management platform and / or the merchant terminal. The obtained address information includes, but is not limited to, the terminal installation address, the merchant operating address, and the merchant operating name. It should be noted that the method of the present application can be applied to the scenario of illegal machine migration prevention, and can also be applied to the scenario of merchant or customer distribution statistics. Similarly, other scenarios that require prediction of administrative regions based on multiple address information are also within the scope of application of the present application. It should be noted that the predicted administrative region can be a specific address or a region according to the requirements of different application scenarios. For example, when the method of the present application is applied to the scenarios of illegal machine migration prevention and merchant or customer distribution statistics, since such scenarios do not require detailed addresses, for example, illegal machine migration prevention needs to determine whether the terminal has conducted cross-city or cross-province operation, and merchant or customer distribution statistics needs to determine the distribution of merchants or customers in various provincial, municipal, or district regions, therefore, the method of the present application is required to predict a region range based on multiple address information; and in other cases where a specific address needs to be predicted, the method of the present application can also be used to predict a specific address based on multiple address information.

[0071] In this embodiment, in the step S12, the plurality of address information is similarity filtered according to a preset similarity threshold, and several address information within the preset similarity threshold range is determined as a predicted address set.

[0072] In this embodiment, the plurality of address information includes a reference address with high credibility, which can be manually selected or specified by an information provider, and other address information is a candidate address that supplements the reference address. When determining the similarity, the candidate address is compared with the reference address one by one to determine the similarity. The method of determining the text similarity is not limited, and any method of calculating the text similarity can be applied, including Jaro-Walker similarity, Jacard similarity, simhash algorithm, LSI model, and matching matrix model. According to the preset similarity threshold, several candidate addresses within the threshold range are determined as the predicted address set together with the reference address. In this embodiment, when the candidate address is completely the same as the reference address, fusion will increase the workload of subsequent address splitting, cleaning, and matching. When the candidate address is significantly different from the reference address, fusion will have the risk of information conflict. Therefore, the similarity threshold should be determined according to the actual requirements, so that the selected candidate address and the reference address have overlapping fields but are not completely the same, so as to achieve the optimal fusion effect.

[0073] Preferably, a plurality of address information with a level mark is acquired, wherein the level mark is used to represent the reliability of the address information; each address information is compared with the address information of a higher level according to the level mark in a preset similarity threshold; and several address information in the preset similarity threshold range is determined as a prediction address set.

[0074] Here, a plurality of address information with a level mark is acquired from the information provider, each address information is sorted from high to low according to the reliability represented by the level mark, and the sorting is traversed from the second address information (because the first address information has no higher level comparison object) and compared with the address information of a higher level, if in the preset similarity threshold range, it is added to the prediction address set, and the traversal is continued; if the similarity comparison result is completely consistent, the address information is discarded and the traversal is continued; if it is not in the preset similarity threshold range and not completely consistent, the address information is discarded and the traversal is ended. Here, in order to reduce unnecessary fusion, the address information with higher similarity and not completely consistent is screened out through the similarity threshold, and the number of similarity comparisons is minimized through sequential single comparison. When the similarity of a certain address information is not in the similarity threshold range and not completely consistent, it means that the address information has the risk of information conflict and is unreliable information, so the reliability of the level mark represented by the other address information of the address information is also unreliable, and therefore the traversal comparison process can be terminated. In this way, the reliable information is screened out while the number of similarity comparisons is greatly reduced, the resource overhead is saved, and the time cost is reduced.

[0075] Further, each level mark has a corresponding similarity threshold, and the higher the reliability represented by the level mark, the lower the corresponding similarity threshold. Accordingly, the more unreliable the address information, the higher the similarity threshold, so that unreliable information can only be added to the address set in the case of high similarity, as the basis for prediction. The overall consistency of the address information is improved, and the possibility of information conflict is further reduced.

[0076] In a preferred embodiment, referring to Figure 2 , wherein, Figure 2 , steps S21 and S22 are the same as or substantially the same as steps S11 and S12 in the Figure 1 embodiment, and therefore will not be described here again, but are included herein by reference. The step S23 includes: cleaning redundant information in the prediction address set; and splitting each address information in the cleaned prediction address set into a plurality of address words according to a preset address keyword library.

[0077] The determined prediction address set is the basis for predicting the administrative region, and the prediction address set is cleaned of redundant information to improve the accuracy of the prediction result.

[0078] In the step, a plurality of cleaning rules are set by means of regular expressions, the prediction address set is traversed by the cleaning rules, and when the address information traversed meets the cleaning rule, the address information is deleted of redundant information according to the regular expression content of the cleaning rule.

[0079] In this way, each cleaning rule can clean a type of redundant information uniformly, and compared with the scheme of establishing a redundant feature library by enumeration, the scheme summarizes the features of various types of redundant information, reduces the workload of the preliminary preparation of the cleaning link, and only matches the prediction address set with a plurality of cleaning rules, which greatly reduces the workload in the cleaning link compared with matching with the enumerated redundant feature library.

[0080] Further, the redundant information includes at least any one of a postal code, location point information, non-Chinese characters, and use information, wherein the cleaning rules correspond to different types of redundant information one by one, and when the redundant information includes at least a postal code and non-Chinese characters, the traversing of the prediction address set by the cleaning rules includes that the order of traversing the prediction address set by the cleaning rule corresponding to the postal code is earlier than the order of traversing the prediction address set by the cleaning rule corresponding to the non-Chinese characters.

[0081] Here, the use information refers to unnecessary orientation or place description, such as “commercial and residential building”, “office building”, “store”, “west side”, etc., and the location point information refers to detailed information of a specific address, such as the Xth courtyard, building, floor, and room, etc. As described above, when the method is applied to the scene of illegal machine migration prevention, merchant or customer distribution statistics, the location point information is redundant information, and in the scene of predicting a specific address, the location point information is not considered as redundant information.

[0082] For example, in python, the cleaning rules of the above redundant information set by means of regular expressions can be:

[0083] Non-Chinese character cleaning rule: [^\u4e00-\u9fff]+;

[0084] Postal code cleaning rule: (? : postal code | post code | code) [: 0-9] +;

[0085] Use information cleaning rule: \s*(?: door face room | door face east | south square | commercial building | clinic | stall number | south block | commercial and residential building | office building | commercial building | commercial and residential area | underground of department store | underground garage | underground mall) \s* $;

[0086] Positioning point cleaning rule: the first [-0-9a-zA-Z / \\\& ,、 # negative one two three four five six seven eight nine ten and to] number (?: store number | store | street building | counter | door face | door face room | door city | courtyard | road | room | shop | card shop | shop | counter | store | shop | shop | store | shop | number | building | room | door city | shop | unit | store | area | section | axis | period).

[0087] In this step, in actual application scenarios, address information may contain postal code information, for example, the address information contains the "Shanghai Huangpu District Wuliqiao Street Postcode: 200023" field, and here "Postcode: 200023" is redundant information. If the redundant information is cleaned by the non-Chinese character cleaning rule first, the "Postcode" field cannot be cleaned. Therefore, the order of traversing the predicted address set by the postal code cleaning rule is earlier than the order of traversing the predicted address set by the non-Chinese character cleaning rule, so as to avoid the situation that a specific field cannot be cleaned.

[0088] Further, according to the preset address keyword library, each address information in the cleaned predicted address set is split into multiple address words. Here, the prior patent with the patent name of "A method and device for determining an administrative region based on address information" with the application number of 202210602770.8 is continued for address word splitting. Specifically, "all sub-addresses are split into multiple address words according to the preset address keyword library. Preferably, the keywords in the address keyword library are arranged in descending order of string length, wherein the operation of splitting all sub-addresses into multiple address words according to the preset address keyword library comprises: sequentially traversing the address keyword library; sequentially matching the traversed keywords with the sub-addresses; if a match is found, splitting the field before the keyword in the sub-address into an address word; determining the field after the keyword in the sub-address as the sub-address, and repeating the above operation".

[0089] Based on the above, the present application accordingly splits all sub-addresses into a plurality of address words according to a preset address keyword library in the present application, the keywords in the address keyword library are arranged in descending order of string length, sequentially traversing the address keyword library, sequentially matching the keywords traversed with each address information in the predicted address set; if matched, the field before the keyword in the address information is split into an address word; the field after the keyword is determined as the address information, and the above operation is repeated.

[0090] Preferably, all keywords with a character length of 1 in the address keyword library are determined; the keywords with a character length of 1 are combined two by two to obtain combined words; the combined words are added to the address keyword library. Thus, the above address word splitting process is performed according to the address keyword library after the keywords are expanded.

[0091] Herein, in actual application scenarios, there are some special administrative regions, such as "Shashi District", "Ningxiang City" and "Qingzhen City", etc., which have address keywords in the address keyword library in their names. If the original address keyword library is used for splitting, mis-splitting will occur. Therefore, for this situation, the keywords in the address keyword library are combined to expand the original address keyword library, thereby reducing the mis-splitting of such special administrative regions. At the same time, since such special administrative regions mostly contain only single-character address keywords, such as "town", "village", "city", "district", etc., only the keywords with a character length of 1 are combined two by two, thereby covering most of the special administrative regions while reducing the additional workload.

[0092] Preferably, the administrative region library in the prior patent is inherited, and the number of occurrences of the combined words in the administrative region word library is counted; when the number of occurrences of the combined word exceeds the screening threshold, the combined word is added to the address keyword library.

[0093] Herein, if all single-character address keywords are combined two by two, it may increase the address keywords of multiple administrative regions without matching, thereby increasing unnecessary matching traversal links in the address word splitting process. Therefore, in order to further reduce the traversal and matching of unnecessary address keywords, a screening threshold is set, and statistics and screening are performed in the administrative region library to avoid expanding unnecessary combined words into the address keyword library. The screening threshold is set according to the requirements, for example, if the highest splitting accuracy is required and the computing resources are sufficient, the screening threshold can be set to the minimum value, so that as long as there is a matching administrative region for the combined word, the combined word is added to the key address keyword library, thereby ensuring the accuracy of the splitting; correspondingly, when the computing resources and time cost need to be saved, the screening threshold can be appropriately increased.

[0094] Further, the splitting of each address information in the predicted address set into a plurality of address words according to the preset address keyword library further comprises: manually checking the splitting result, and if the splitting result is wrong, adding the address information corresponding to the splitting result to the special place library.

[0095] Here, the special place library inherited from the previous patent has some relatively complex administrative region names, such as "Jishishan Bao'an Dongxiang Sala Autonomous County" and "Buyluq Tajik Ethnic Township Kuiyibag District Office", etc., which are difficult to split. Therefore, during the business process, the address splitting abnormal result can be captured, and the corresponding original address information can be added to the special place library, thereby improving the accuracy of address splitting. It should be clear that the possibility of such complex administrative region names appearing in the business process is relatively low, so the operation of manually checking the splitting result is only performed in special situations where such administrative region names need to be processed.

[0096] The process of determining the corresponding administrative region based on the plurality of address words, the preset level keyword library and the preset administrative region library follows the method of the previous patent, that is: determining a word group set containing administrative region words of all levels based on the preset level keyword library and the plurality of address words; the word group set is matched in the preset administrative region library to determine a matching set corresponding to the word group set; the administrative region corresponding to the matching set is determined as the administrative region corresponding to the address information. Specifically, the determination of the word group set containing administrative region words of all levels based on the preset level keyword library and the plurality of address words comprises: sequentially traversing the address words according to the level keyword library; if there is a full name field or a short name field of a national-level administrative region in the traversed address words, the full name of the national-level administrative region in the level keyword library is determined as the national-level administrative region word; only the field after the full name field or the short name field of the national-level administrative region in the address words is reserved, and the reserved field is updated as the traversed address words, and the determination of the provincial-level and prefectural-level administrative region words is sequentially performed according to the above method; the reserved field after the determination of the national-level, provincial-level and prefectural-level administrative region words is determined as the tail word; the national-level, provincial-level, prefectural-level administrative region words and the tail word are determined as the word group set.

[0097] Further, the calculation content of the above-mentioned prior patent is used to establish a province-city mapping library based on the administrative region library. After the word set containing administrative region words of each level is determined based on the preset level word library and the plurality of address words, the method further includes: obtaining the determined city-level administrative region word and the province-level administrative region word in each address word in the word set; judging whether a mapping relationship exists between the city-level administrative region word and the province-level administrative region word according to the province-city mapping library; and if no mapping relationship exists, replacing the city-level administrative region word and the province-level administrative region word in the word set with original combination fields corresponding to the city-level administrative region word and the province-level administrative region word.

[0098] In this case, the process of establishing the province-city mapping library is to establish a mapping relationship between the city-level administrative region and the province-level administrative region to which the city-level administrative region belongs in the administrative region library, and save the mapping relationship in the province-city mapping library. In actual application scenarios, there may be some streets named after administrative regions, such as Zhongshan Road and Nanjing Road. The prior patent adds such special places to the special place library, and before determining the word set based on the level word library, the special place library is traversed based on the address word to determine whether such special places exist in the address word. The establishment of the special place library by enumeration has the following defects. First, this method is difficult to exhaust all special places, and may miss some special places, resulting in incorrect prediction results. Second, if all special places are exhausted in the special place library, a large amount of time and resources will be consumed when the special place library is traversed. Therefore, the present application improves the prior patent by establishing the province-city mapping library, and only saves the mapping relationship in the province-city mapping library, such as the mapping relationship between the codes in the form of codes. For example, in the administrative region library, “Beijing” corresponds to code 01, “Chaoyang District, Beijing” corresponds to code 12, and “Changping District, Beijing” corresponds to code 13. The mapping relationship between 01 and {12, 13} is established in the reverse index library. The establishment form of the mapping relationship includes but is not limited to queue, linked list, tree structure, etc. In another embodiment, the province-city mapping library is established as follows: the code of each lower-level administrative region retains the code of the upper-level administrative region, and an additional bit code is supplemented after the code of the upper-level administrative region, and the codes of all levels are fixed length. Taking Beijing as an example, “Beijing” corresponds to code 01001, “Chaoyang District, Beijing” corresponds to code 0100101, and “Changping District, Beijing” corresponds to code 0100102. When determining whether a mapping relationship exists between the city-level administrative region word and the province-level administrative region word, only the part of the code field of the city-level administrative region word is obtained according to the fixed length of the province-level code, the code field is used as the province-level code, and it is determined whether the province-level administrative region corresponding to the province-level code is the province-level administrative region word. The method of establishing the province-city mapping library of the present application completely avoids the misidentification of special places while greatly reducing the time cost required for identifying special places.

[0099] Compared with the prior art, the application obtains multiple address information, performs similarity screening on the multiple address information according to a preset similarity threshold, and determines several address information in the preset similarity threshold range as a predicted address set; each address information in the predicted address set is split into multiple address words according to a preset address keyword library; and a corresponding administrative region is determined based on the multiple address words, a preset level keyword library and a preset administrative region library. By conditionally fusing multiple address information, the application realizes low-information-conflict administrative region prediction under the premise of fusing as many address information as possible, and greatly improves the accuracy of the prediction result.

[0100] In addition, an embodiment of the application further provides a computer readable medium, which has computer readable instructions stored thereon, and the computer readable instructions can be executed by a processor to implement the foregoing method.

[0101] An embodiment of the application further provides an administrative region prediction device based on multiple address information, and the device comprises:

[0102] one or more processors; and

[0103] a memory in which computer readable instructions are stored, and the computer readable instructions, when executed, cause the processor to perform the operations of the foregoing method.

[0104] For example, the computer readable instructions, when executed, cause the one or more processors to: obtain multiple address information; perform similarity screening on the multiple address information according to a preset similarity threshold, and determine several address information in the preset similarity threshold range as a predicted address set; split each address information in the predicted address set into multiple address words according to a preset address keyword library; and determine a corresponding administrative region based on the multiple address words, a preset level keyword library and a preset administrative region library.

[0105] It will be obvious to a person skilled in the art that the application is not limited to the details of the above-described exemplary embodiments, but that the application can be implemented in other concrete forms without deviating from the spirit or the basic characteristics of the application. The embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the application being defined by the appended claims rather than by the above Description, which is therefore intended merely as a specification. All changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference signs in the claims should not be construed as limiting the claim concerned. Furthermore, it is to be noted that the term "comprising" does not exclude other elements or steps, that the term "a" or "an" does not exclude a plurality, and that a single processor or other unit can fulfil the functions of several units recited in the claims. The terms first, second and the like do not denote any ordering, but rather are used as names for naming different units.

Claims

1. An administrative region prediction method based on a plurality of address information, wherein, The method comprises: obtaining a plurality of address information containing level marks, wherein the level marks are used to represent the reliability of the address information; performing similarity screening on the plurality of address information according to a preset similarity threshold, and determining a plurality of address information within the preset similarity threshold range as a predicted address set; splitting each address information in the predicted address set into a plurality of address words according to a preset address keyword library; determining a corresponding administrative region based on the plurality of address words, a preset level keyword library and a preset administrative region library; wherein the similarity screening on the plurality of address information according to the preset similarity threshold, and the determination of the plurality of address information within the preset similarity threshold range as the predicted address set comprise: sorting each address information in order from high to low according to the reliability represented by the level marks, traversing the sorting from the second address information, and performing similarity comparison between the traversed address information and the address information of a higher level of reliability; if the similarity comparison result is within the preset similarity threshold range, adding the traversed address information to the predicted address set, and continuing the traversal; if the similarity comparison result is completely consistent, deleting the traversed address information and continuing the traversal; if the similarity comparison result is not within the preset similarity threshold range and is not completely consistent, deleting the traversed address information and ending the traversal.

2. The method of claim 1, wherein, The determination of the plurality of address information within the preset similarity threshold range as the predicted address set further comprises: cleaning up redundant information in the predicted address set; the splitting of each address information in the predicted address set into a plurality of address words according to the preset address keyword library comprises: splitting each address information in the cleaned predicted address set into a plurality of address words according to the preset address keyword library.

3. The method of claim 1, wherein, Each level mark has a corresponding similarity threshold, and the higher the reliability represented by the level mark, the lower the corresponding similarity threshold.

4. The method of claim 2, wherein, The cleaning up of redundant information in the predicted address set further comprises: setting a plurality of cleaning rules in the form of regular expressions; wherein the cleaning up of redundant information in the predicted address set comprises: traversing the predicted address set through each cleaning rule; when the traversed address information meets the cleaning rule, deleting the redundant information of the address information according to the regular expression content of the cleaning rule.

5. The method of claim 4, wherein the redundancy information comprises at least one of a zip code, a point of location information, non-Chinese characters, and usage information. The cleaning rules correspond to different types of redundant information one by one; wherein when the redundant information at least includes a postal code and a non-Chinese character, the traversing of the predicted address set through the postal code corresponding cleaning rule is prior to the traversing of the predicted address set through the non-Chinese character corresponding cleaning rule. The splitting of each address information in the predicted address set into a plurality of address words according to the preset address keyword library further comprises:

6. The method of claim 1, wherein, determining all keywords with a character length of 1 in the address keyword library; combining all keywords with a character length of 1 to obtain combined words; adding the combined words to the address keyword library. ​ 7. The method of claim 6, presetting a screening threshold, wherein, The adding the combined word into the address keyword library further comprises: counting the number of occurrences of each combined word in the administrative region library; The adding the combined word into the address keyword library comprises: adding the combined word into the address keyword library when the number of occurrences of the combined word exceeds the screening threshold.

8. The method of claim 1, wherein, The splitting each address information in the predicted address set into a plurality of address words according to the preset address keyword library further comprises: manually checking the splitting result, and adding the address information corresponding to the splitting result into a special place library if the splitting result is wrong.

9. The method according to any one of claims 1 to 8, establishing a province mapping repository from the administrative area repository, wherein, The province mapping library comprises mapping relationships between each prefecture-level administrative region word and the province-level administrative region word to which the prefecture-level administrative region word belongs, and the determining the corresponding administrative region based on the plurality of address words, the preset level word library and the preset administrative region library comprises: determining a word group set containing administrative region words of all levels based on the preset level word library and the plurality of address words; matching the word group set in the preset administrative region library to determine a matching set corresponding to the word group set; determining the administrative region corresponding to the matching set as the administrative region corresponding to the address information. The determining a word group set containing administrative region words of all levels based on the preset level word library and the plurality of address words further comprises: acquiring the prefecture-level administrative region word and the province-level administrative region word in each address word in the word group set; determining whether there is a mapping relationship between the prefecture-level administrative region word and the province-level administrative region word according to the province mapping library; if there is no mapping relationship, replacing the prefecture-level administrative region word and the province-level administrative region word in the word group set with original combined fields corresponding to the prefecture-level administrative region word and the province-level administrative region word. 10.A computer readable medium having stored thereon computer readable instructions executable by a processor to implement the method of any one of claims 1 to 9.

11. An administrative region prediction device based on a plurality of address information, wherein, The device comprises: one or more processors; and a memory storing computer readable instructions executable by the processor to implement the method of any one of claims 1 to 9. The device comprises: one or more processors; and a memory storing computer readable instructions executable by the processor to implement the method of any one of claims 1 to 9.

Images