Method and apparatus for determining wi-fi and merchant poi correspondence

Abstract

The application provides a method and device for determining the correspondence between Wi-Fi and a merchant POI, wherein the method comprises: establishing a multilingual word list of merchant POIs, wherein the word list contains the multilingual translation mapping relationship of each merchant POI to be mined; based on each mapping relationship, checking whether a target merchant POI corresponding to a target Wi-Fi exists in the word list; if not, selecting N candidate merchant POIs from the merchant POIs to be mined according to a first preset rule; scoring each of the N candidate merchant POIs based on a pre-trained deep neural network model; and selecting a target merchant POI corresponding to the target Wi-Fi from the N candidate merchant POIs according to each score. The merchant Wi-Fi mining method provided by the application can accurately mine the Wi-Fi corresponding to the merchant POI.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to a method and device for determining the correspondence between Wi-Fi and merchant POI, an electronic device and a storage medium. BACKGROUND

[0002] At present, the network positioning method based on Wi-Fi is widely used. Wi-Fi network positioning learns the position of Wi-Fi by means of Wi-Fi information and positioning results in satellite positioning, so as to perform positioning by means of the position of Wi-Fi in the next positioning demand. This positioning method can also be called general positioning, and the final output of the positioning result is latitude and longitude information. At present, the life service platform has a natural demand for positioning capability, and has risen from the demand for general latitude and longitude to the demand for scene, such as positioning to the building where the user is located, the mall where the user is located, the shop where the user is located, etc., so as to provide more accurate services for the user. Obviously, the existing positioning method cannot meet the positioning demand of the life service platform.

[0003] Therefore, the technical personnel in the field have developed a positioning method based on merchant POI (Point of Interest) and Wi-Fi. POI is any point on the map that has no geographical meaning and has meaning, such as a store, a gas station, etc. When positioning is performed by using this method, the corresponding Wi-Fi information of the user can be determined first, and then the POI where the user is really located can be calculated by means of the pre-established Wi-Fi and POI mapping table, so as to realize accurate positioning of the user. The implementation of this positioning method depends on the Wi-Fi and POI mapping table, and therefore, mining the correspondence between POI and Wi-Fi is an urgent technical problem to be solved by the technical personnel in the field. SUMMARY

[0004] In view of the above problems, the present application is proposed to provide a method and device for determining the correspondence between Wi-Fi and merchant POI, an electronic device and a storage medium, which can overcome the above problems or at least partially solve the above problems.

[0005] In a first aspect, the embodiments of the present application disclose a method for determining a correspondence between a Wi-Fi and a merchant POI, comprising: establishing a multilingual word list of merchant POIs, wherein the word list contains a multilingual translation mapping relationship of each merchant POI to be mined; based on each mapping relationship, searching whether a target merchant POI corresponding to a target Wi-Fi exists in the word list; if not, selecting N candidate merchant POIs from the merchant POIs to be mined according to a first preset rule; based on a pre-trained deep neural network model, scoring each of the N candidate merchant POIs; and according to each scoring result, selecting a target merchant POI corresponding to the target Wi-Fi from the N candidate merchant POIs.

[0006] Optionally, the step of selecting N candidate merchant POIs from the merchant POIs to be mined according to a first preset rule comprises:

[0007] In the case that the target merchant POI corresponding to the target Wi-Fi does not exist in the word list, performing semantic extraction on a SSID (Service Set Identifier) of the target Wi-Fi to obtain a processed SSID;

[0008] Calculating an edit distance score of the processed SSID and each candidate merchant POI, respectively;

[0009] Selecting a candidate merchant POI whose edit distance score is ranked in the top N.

[0010] Optionally, the step of scoring each of the N candidate merchant POIs based on a pre-trained deep neural network model comprises:

[0011] For each of the N candidate merchant POIs, inputting the candidate merchant POI and the processed SSID into the pre-trained deep neural network model;

[0012] Determining an output of the deep neural network model as a scoring result of the candidate merchant POI.

[0013] Optionally, the step of selecting a target merchant POI corresponding to the target Wi-Fi from the N candidate merchant POIs according to each scoring result comprises:

[0014] According to each scoring result, selecting a first candidate merchant POI with the highest score;

[0015] In the case that the first candidate merchant POI is single, determining the first candidate merchant POI as the target merchant POI corresponding to the target Wi-Fi.

[0016] Optionally, after the step of screening out the first candidate merchant POI with the highest score according to each score, the method further comprises:

[0017] In a case where the number of the first candidate merchant POIs is greater than 1, screening out a target merchant POI from each of the first candidate merchant POIs according to preset parameter information, wherein the preset parameter information comprises at least one of the following: a distance between the first candidate merchant corresponding to the first candidate merchant POI and the target Wi-Fi, and a Wi-Fi feature contained in business data of the first candidate merchant.

[0018] Optionally, before the step of establishing the multi-lingual word list of merchant POI, the method further comprises:

[0019] Binding relationships between a plurality of Wi-Fi and merchant POIs analyzed from the business data are determined as positive samples;

[0020] Each positive sample is processed according to a second preset rule to obtain a processed positive sample;

[0021] A negative sample is generated according to a third preset rule;

[0022] The deep neural network model is trained according to each of the processed positive sample and the negative sample.

[0023] Optionally, the step of processing each of the positive samples according to the second preset rule to obtain a processed positive sample comprises:

[0024] For each of the binding relationships between the Wi-Fi and the merchant POI, a preset processing is performed on the SSID of the Wi-Fi in the binding relationship in sequence to obtain a processed SSID, wherein the preset processing comprises: SSID semantic extraction, Chinese to Pinyin conversion, and letter string segmentation;

[0025] The merchant POI in the binding relationship is subjected to word segmentation processing to obtain a processed merchant POI;

[0026] The SSID of the Wi-Fi, the processed SSID, the merchant POI, and the processed merchant POI are determined as the processed positive sample.

[0027] Optionally, the step of generating a negative sample according to a third preset rule comprises:

[0028] A first preset number of positive samples are selected;

[0029] For each of the positive samples, at least one letter of the SSID of the Wi-Fi of the positive sample is modified to generate a modified SSID;

[0030] The modified SSID and the binding relationship of the merchant POI in the positive sample are determined as a negative sample.

[0031] Optionally, the step of generating the negative sample according to the third preset rule comprises:

[0032] selecting a second preset number of positive samples;

[0033] For each of the positive samples, a third preset number of second merchant POIs satisfying a preset matching condition are found according to Wi-Fi in the positive sample;

[0034] A binding relationship between each of the second merchant POIs and the Wi-Fi is established respectively to generate the third preset number of negative samples.

[0035] In a second aspect, an embodiment of the present application discloses a device for determining a Wi-Fi and merchant POI corresponding relationship, comprising:

[0036] A building module is configured to build a multilingual word list of merchant POIs, wherein the word list contains a multilingual translation mapping relationship of each merchant POI to be mined;

[0037] A finding module is configured to find whether a target merchant POI corresponding to a target Wi-Fi exists in the word list based on each of the mapping relationships;

[0038] A first screening module is configured to screen N candidate merchant POIs from the merchant POIs to be mined according to a first preset rule if not;

[0039] A scoring module is configured to score each of the N candidate merchant POIs based on a pre-trained deep neural network model;

[0040] A second screening module is configured to screen a target merchant POI corresponding to the target Wi-Fi from the N candidate merchant POIs according to each of the scoring results.

[0041] Optionally, the first screening module comprises:

[0042] A first sub-module is configured to perform semantic extraction on an SSID of the target Wi-Fi to obtain a processed SSID in a case where the target merchant POI corresponding to the target Wi-Fi does not exist in the word list;

[0043] A second sub-module is configured to calculate an edit distance score of the processed SSID and each of the candidate merchant POIs respectively;

[0044] A third sub-module is configured to screen a candidate merchant POI whose edit distance score is ranked in TopN.

[0045] Optionally, the scoring module comprises:

[0046] a fourth submodule configured to input each candidate merchant POI in the N candidate merchant POIs and the processed SSID into a pre-trained deep neural network model;

[0047] a fifth submodule configured to determine an output of the deep neural network model as a scoring result of the candidate merchant POI.

[0048] Optionally, the second screening module comprises:

[0049] a sixth submodule configured to screen a first candidate merchant POI with the highest score according to the scoring results;

[0050] a seventh submodule configured to determine the first candidate merchant POI as a target merchant POI corresponding to the target Wi-Fi in a case where the first candidate merchant POI is single.

[0051] Optionally, the apparatus further comprises:

[0052] an eighth submodule configured to, after the sixth submodule screens the first candidate merchant POI with the highest score according to the scoring results, screen a target merchant POI from the first candidate merchant POIs according to preset parameter information in a case where the number of the first candidate merchant POIs is greater than 1, wherein the preset parameter information comprises at least one of the following: a distance between a first candidate merchant corresponding to the first candidate merchant POI and the target Wi-Fi, and a Wi-Fi feature contained in business data of the first candidate merchant.

[0053] Optionally, the apparatus further comprises:

[0054] a positive sample determination module configured to determine a plurality of binding relationships between Wi-Fi and merchant POIs analyzed from the business data as positive samples before the establishment module establishes the multilingual word list of merchant POIs.

[0055] a processing module configured to process the positive samples according to a second preset rule to obtain processed positive samples;

[0056] a generation module configured to generate negative samples according to a third preset rule;

[0057] a training module configured to train a deep neural network model according to the processed positive samples and the negative samples.

[0058] Optionally, the processing module comprises:

[0059] a ninth sub-module, configured to perform preset processing on the SSID of the Wi-Fi in the binding relationship to obtain a processed SSID, for each of the binding relationships between the Wi-Fi and the merchant POI, wherein the preset processing includes: SSID semantic extraction, Chinese-to-pinyin conversion, and letter string segmentation;

[0060] a tenth sub-module, configured to perform word segmentation processing on the merchant POI in the binding relationship to obtain a processed merchant POI;

[0061] an eleventh sub-module, configured to determine the SSID of the Wi-Fi, the processed SSID, the merchant POI, and the processed merchant POI as a processed positive sample.

[0062] Optionally, the generating module includes:

[0063] a twelfth sub-module, configured to select a first preset number of positive samples;

[0064] a thirteenth sub-module, configured to modify at least one letter of the SSID of the Wi-Fi in each of the positive samples to generate a modified SSID;

[0065] a fourteenth sub-module, configured to determine the binding relationship between the modified SSID and the merchant POI in the positive sample as a negative sample.

[0066] Optionally, the generating module includes:

[0067] a fifteenth sub-module, configured to select a second preset number of positive samples;

[0068] a sixteenth sub-module, configured to find, according to the Wi-Fi in each of the positive samples, a third preset number of second merchant POIs that satisfy a preset matching condition;

[0069] a seventeenth sub-module, configured to establish a binding relationship between each of the second merchant POIs and the Wi-Fi to generate a third preset number of negative samples.

[0070] In a third aspect, an embodiment of the present application discloses an electronic device, including: one or more processors; and one or more machine-readable media having instructions stored thereon; when the instructions are executed by the one or more processors, the processor executes the method for determining the corresponding relationship between the Wi-Fi and the merchant POI according to any one of the above.

[0071] In a fourth aspect, an embodiment of the present application discloses a computer-readable storage medium, having a computer program stored thereon, which is executed by a processor to implement the method for determining the corresponding relationship between the Wi-Fi and the merchant POI according to any one of the above.

[0072] In the embodiment of the present application, based on the establishment of the merchant POI multilingual vocabulary; based on each mapping relationship, it is found whether there is a target merchant POI corresponding to the target Wi-Fi in the vocabulary; if not, N candidate merchant POIs are selected from the merchant POIs to be mined according to the first preset rule; based on the pre-trained deep neural network model, the N candidate merchant POIs are scored respectively; according to each scoring result, the target merchant POI corresponding to the target Wi-Fi is selected from the N candidate merchant POIs. The scheme for determining the correspondence between Wi-Fi and merchant POI provided by the embodiment of the present application combines the multilingual vocabulary, the first preset rule and the deep neural network model to calculate the semantic correlation between Wi-Fi and merchant POI, which can accurately determine the merchant POI corresponding to the Wi-Fi and indirectly accurately mine the Wi-Fi corresponding to the merchant POI. BRIEF DESCRIPTION OF DRAWINGS

[0073] Figure 1 is a step flow chart of a method for determining the correspondence between Wi-Fi and merchant POI according to an embodiment of the present application;

[0074] Figure 2 is a step flow chart of a deep neural network model training method according to an embodiment of the present application;

[0075] Figure 3 is a structural block diagram of a device for determining the correspondence between Wi-Fi and merchant POI according to an embodiment of the present application. DETAILED DESCRIPTION

[0076] Exemplary embodiments of the present disclosure will be described in greater detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be accurately conveyed to those skilled in the art.

[0077] Referring to Figure 1 , a step flow chart of a method for determining the correspondence between Wi-Fi and merchant POI according to an embodiment of the present application is shown.

[0078] The method for determining the correspondence between Wi-Fi and merchant POI according to an embodiment of the present application can include the following steps:

[0079] Step 101: Establish a merchant POI multilingual vocabulary.

[0080] The merchant POI multilingual glossary contains a multilingual translation mapping relationship of each merchant POI to be mined.

[0081] POI is a point of interest, and a company, a store or a shopping mall can be regarded as a POI in the implementation process. The merchant POI can uniquely mark a merchant. The merchant POI in the glossary can be continuously expanded and modified during system use.

[0082] Step 102: Based on the mapping relationship, it is determined whether the target merchant POI corresponding to the target Wi-Fi exists in the glossary.

[0083] In the embodiment of the application, a Wi-Fi corresponding to a merchant POI is taken as an example for mining. In the embodiment of the application, the target merchant POI is mined based on the target Wi-Fi, so as to achieve the purpose of mining the target Wi-Fi for the target merchant POI.

[0084] When the target merchant POI corresponding to the Wi-Fi is searched, the SSID of the Wi-Fi can be used for searching. When the target merchant POI corresponding to the target Wi-Fi is searched from the glossary, the SSID of the target Wi-Fi can be matched with the name of each merchant POI in the glossary. If the matching is successful, it is determined that the target merchant POI corresponding to the target Wi-Fi exists in the glossary. The target Wi-Fi is directly bound to the target merchant POI found. If the matching fails, the subsequent process for searching the target merchant POI in the embodiment of the application needs to be performed.

[0085] Step 103: If not, N candidate merchant POIs are selected from the merchant POIs to be mined according to a first preset rule.

[0086] The first preset rule can be set to select the candidate merchant POI according to the edit distance. In the actual selection process, the edit distance between the SSID of the target Wi-Fi and each merchant POI in the glossary can be calculated, and the N merchant POIs corresponding to the edit distances in the order from small to large are determined as the candidate merchant POIs. The merchant POIs corresponding to the edit distances less than the preset distance can also be determined as the candidate merchant POIs.

[0087] N is an integer greater than or equal to 1, and the specific value of N can be set by those skilled in the art according to actual needs, and the embodiment of the application does not make specific limitations.

[0088] Step 104: scoring the N candidate merchant POIs respectively based on the pre-trained deep neural network model.

[0089] After inputting the target Wi-Fi and the candidate merchant POI as input information into the deep neural network model, the deep neural network model can output the scoring result of the candidate merchant POI.

[0090] Step 105: selecting the target merchant POI corresponding to the target Wi-Fi from the N candidate merchant POIs according to the respective scoring results.

[0091] In actual implementation, the candidate merchant POI with the highest score in the scoring result can be selected as the target merchant POI. When there are two or more candidate merchant POIs with the highest score, one target merchant POI can be selected according to a preset parameter, or one target merchant POI can be selected randomly. The specific selection rule for selecting the target merchant POI corresponding to the target Wi-Fi from the N candidate merchant POIs is not limited in the embodiment of the present application.

[0092] The method for determining the correspondence between the Wi-Fi and the merchant POI provided in the embodiment of the present application is based on the establishment of the multi-lingual word table of the merchant POI, based on the respective mapping relationship, and whether the target merchant POI corresponding to the target Wi-Fi exists in the word table is searched. If not, the N candidate merchant POIs are selected from the merchant POIs to be mined according to the first preset rule. The N candidate merchant POIs are scored respectively based on the pre-trained deep neural network model. The target merchant POI corresponding to the target Wi-Fi is selected from the N candidate merchant POIs according to the respective scoring results. The method for determining the correspondence between the Wi-Fi and the merchant POI provided in the embodiment of the present application combines the multi-lingual word table, the first preset rule and the deep neural network model, calculates the correlation degree between the Wi-Fi and the merchant POI in terms of semantics, and can accurately determine the merchant POI corresponding to the Wi-Fi, which is equivalent to accurately mining the Wi-Fi corresponding to the merchant POI.

[0093] In an alternative embodiment, when the N candidate merchant POIs are selected from the merchant POIs to be mined according to the first preset rule, the following method can be used:

[0094] First, in the case where the target merchant POI corresponding to the target Wi-Fi does not exist in the multi-lingual word table of the merchant POI, the SSID of the target Wi-Fi is subjected to semantic extraction to obtain the processed SSID.

[0095] Wi-Fi can be uniquely determined by a MAC address and an SSID, which can be understood as the name of the Wi-Fi. In the embodiment of the present application, the similarity between the SSID of the Wi-Fi and the semantics of the merchant POI is to be measured. However, not all SSIDs in the real world have semantic information. First, the SSID may contain some specific characters, such as "5G", "free", "first floor", etc. It may also contain information on the router brand, such as "TP_Link", "huawei", or even consist of a random string of characters. Such characters are meaningless for our target, so they need to be deleted and the semantic part needs to be extracted. In addition, there are some Wi-Fi that are mobile terminal hotspots. Such Wi-Fi should not be bound to the merchant POI and should also be filtered out.

[0096] Secondly, the edit distance score of the processed SSID and each candidate merchant POI is calculated respectively;

[0097] When calculating the edit distance score, the existing related algorithm can be used for implementation, and the present embodiment does not make specific restrictions on this.

[0098] Finally, the candidate merchant POI with the edit distance score ranked in the TopN is selected.

[0099] This optional way of selecting the candidate merchant POI makes the selected merchant POI more reliable.

[0100] In an optional embodiment, the scoring method based on the pre-trained deep neural network model for the N candidate merchant POIs can be as follows:

[0101] For each candidate merchant POI in the N candidate merchant POIs, the candidate merchant POI and the processed SSID are input into the pre-trained deep neural network model. The output of the deep neural network model is determined as the scoring result of the candidate merchant POI.

[0102] This optional scoring method based on the deep neural network model for the candidate merchant POI is accurate.

[0103] In an optional embodiment, the way of selecting the target merchant POI corresponding to the target Wi-Fi from the N candidate merchant POIs according to the scoring results can be as follows:

[0104] According to the scoring results, the first candidate merchant POI with the highest score is selected.

[0105] In the case where the first candidate merchant POI is single, the first candidate merchant POI is determined as the target merchant POI corresponding to the target Wi-Fi.

[0106] If the number of first-candidate merchant POIs is greater than 1, a target merchant POI is filtered out from each first-candidate merchant POI based on preset parameter information.

[0107] The preset parameter information includes at least one of the following: the distance between the first candidate merchant (POI) and the target Wi-Fi network, and the Wi-Fi characteristics contained in the first candidate merchant's business data. The first candidate merchant's business data may include, but is not limited to, the Wi-Fi information connected by the user when making online payments.

[0108] The smaller the distance between the first candidate merchant and the target Wi-Fi, the greater the likelihood that the target Wi-Fi belongs to the first candidate merchant; the higher the number of times the Wi-Fi connected to by users when making online payments in the first candidate merchant's business data is the target Wi-Fi, the greater the likelihood that the target Wi-Fi belongs to the first candidate merchant.

[0109] This method of optionally determining the target merchant POI from N candidate merchant POIs results in a more reliable target merchant POI.

[0110] Reference Figure 2 The diagram illustrates a flowchart of the steps involved in training a deep neural network model according to an embodiment of the present invention. The training of the deep neural network model is performed during... Figure 1 The merchant Wi-Fi mining method shown was completed previously. The deep neural network model can be a Transformer model. Transformers are suitable for translation between different languages ​​and consist of an encoder and a decoder. In this embodiment, only the encoder is used, with the string as input data. The neural network is then directly connected after the encoder for classification learning. The deep neural network model training method includes the following steps:

[0111] Step 201: Analyze the business data to obtain the binding relationships between multiple Wi-Fi networks and merchant POIs, and identify them as positive samples.

[0112] The specific number of positive samples collected can be set by those skilled in the art according to actual needs, and no specific limitation is made in this embodiment of the invention.

[0113] Step 202: Process each positive sample according to the second preset rule to obtain the processed positive sample.

[0114] Each positive sample includes: a Wi-Fi network and a merchant POI.

[0115] The manner of processing each positive sample according to a second preset rule to obtain a processed positive sample can be as follows:

[0116] First, for each binding relationship between Wi-Fi and a merchant POI, a preset processing is performed on the SSID of the Wi-Fi in the binding relationship in sequence to obtain a processed SSID.

[0117] The preset processing includes SSID semantic extraction, conversion of Chinese into pinyin, and letter string segmentation.

[0118] SSID semantic extraction:

[0119] The SSID can contain some specific characters, such as "5G", "free", "first floor", and can also contain information on the router brand, such as "TP_Link", "huawei", or even completely composed of random strings. Such characters are meaningless for the technical problem to be solved by the present application, so the part containing semantics needs to be deleted and extracted. In addition, some Wi-Fi are mobile terminal hotspots, which should not be bound with the merchant POI, and should also be filtered out, so the SSID semantic extraction needs to be performed on the Wi-Fi in the positive sample to remove the SSID without semantics.

[0120] For the SSID without semantics, the Chinese needs to be converted into pinyin in sequence, and then letter string segmentation is performed. When converting Chinese into pinyin, the Chinese, letters, and other characters of the SSID can be segmented, and the Chinese part is converted into corresponding pinyin. When performing letter string segmentation on the SSID, the first letter, pinyin, and English word recognition and segmentation can be performed on the letter string: by collecting corpus and counting the frequency of occurrence of English words therein, and adding each pinyin, an English word library is constructed. When segmenting the letter string, a dynamic programming algorithm can be used.

[0121] Secondly, the merchant POI in the binding relationship is subjected to word segmentation processing to obtain a processed merchant POI.

[0122] In the actual implementation process, any appropriate Chinese morphological analysis toolkit can be used to segment the name of the merchant POI, such as "Nantong Suobo Education Consulting Co., Ltd." being segmented into "Nantong", "Suobo", "Education", "Consulting", and "Co., Ltd.".

[0123] Finally, the SSID of the Wi-Fi, the processed SSID, the merchant POI, and the processed merchant POI are determined as the processed positive sample.

[0124] That is, the positive sample data of the final input model can be divided into four parts: SSID of Wi-Fi, processed SSID, merchant POI, and processed merchant POI. For example, "gmxx#gm;x;x <sep>Gu Miao School, Xiaoniu School <sep>Shenzhen Xinao Pu Hui Investment Management Co., Ltd. of Shenzhen, China.

[0125] Step 203: generating negative samples according to a third preset rule.

[0126] During the training of the deep neural network model, both positive samples and negative samples are needed to effectively train the model, so in this step, negative samples for model training are generated.

[0127] An optional way of generating negative samples according to the third preset rule can include the following sub-steps:

[0128] Sub-step one: selecting a first preset number of positive samples.

[0129] The first preset number can be set by those skilled in the art according to actual needs, and the present embodiment does not make specific limitations on this.

[0130] In step two: for each positive sample, at least one letter of the SSID of the positive sample Wi-Fi is modified to generate a modified SSID.

[0131] In the introduction of the construction of positive samples, the first letter, pinyin, and English word recognition and segmentation of the SSID are introduced. For each substring, it can be modified to other letter strings with a probability of 0.1: when the substring is composed of pinyin, it is modified to other pinyin with a containing relationship but not a front and back nasal sound relationship, such as "ji" is changed to "jin", "hong" is changed to "zhong", etc.; when the substring is of other forms, it is modified to other English or pinyin strings with a length not exceeding the length of the substring plus 2.

[0132] When performing letter replacement, the replacement can be performed based on a pre-created letter role assignment rule. Letters can appear in both SSIDs and merchant POIs, but the roles of the letters are quite different. For example, the same "h" can be the first letter of the pinyin "hong", and can also be an ordinary character inside "zhong". Similarly, it can also be an English word inside "hotel". The different meanings can affect the model's judgment, so in the letter role assignment rule, each letter from a-z is assigned multiple roles, i.e., ordinary letter, pinyin first letter, pinyin internal letter, and English letter.

[0133] Sub-step three: determining the binding relationship between the modified SSID and the merchant POI in the positive sample as a negative sample.

[0134] This way of constructing negative samples by modifying SSIDs is simple and easy to operate.

[0135] Another optional way of generating negative samples according to a third preset rule can include the following sub-steps:

[0136] Sub-step one: select a second preset number of positive samples.

[0137] The second preset number can be set by those skilled in the art according to actual needs, and the present embodiment does not make specific limitations on this.

[0138] Sub-step two: for each positive sample, find a third preset number of second merchant POIs that satisfy a preset matching condition according to the Wi-Fi in the positive sample.

[0139] The third preset number can be set by those skilled in the art according to actual needs, and the present embodiment does not make specific limitations on this. For example, the third preset number can be set to 3, 4, or 5, etc.

[0140] When finding the second merchant POI, another 3 merchant POIs with higher edit distance scores can be found near the Wi-Fi as the second merchant POI. The SSID matching place of the second merchant POI and the Wi-Fi cannot be exactly the same as the positive sample. For example, when the SSID is "Luo Bing Fen", if the positive sample is "Liu Jie Luo Bing Fen", the negative sample cannot be "Zhang Ji Luo Bing Fen" or other merchant POIs.

[0141] Sub-step three: respectively establish the binding relationship between each second merchant POI and the Wi-Fi, and generate a third preset number of negative samples.

[0142] This optional way of finding a merchant POI with similar semantics as the merchant POI corresponding to the Wi-Fi in the negative sample is simple and easy to operate.

[0143] Step 204: train the deep neural network model according to the processed positive samples and negative samples.

[0144] Both the positive samples and the negative samples are training samples of the deep neural network model. When training the deep neural network model according to the training samples, for each training sample, the training sample is input into the deep neural network model, the model parameters are adjusted according to the model output result, and the training of the deep neural network model is completed until the convergence degree of the deep neural network model reaches a preset convergence degree.

[0145] Referring to Figure 3 , a structural block diagram of a device for determining the correspondence between Wi-Fi and merchant POIs according to an embodiment of the present application is shown.

[0146] The device for determining the correspondence between Wi-Fi and merchant POIs according to an embodiment of the present application can include the following modules:

[0147] The establishing module 301 is configured to establish a multilingual POI word table of a merchant, wherein the word table comprises a multilingual translation mapping relationship of each merchant POI to be mined.

[0148] The searching module 302 is configured to search whether a target merchant POI corresponding to a target Wi-Fi exists in the word table based on the mapping relationships.

[0149] The first screening module 303 is configured to screen N candidate merchant POIs from the merchant POIs to be mined according to a first preset rule if the target merchant POI corresponding to the target Wi-Fi does not exist in the word table.

[0150] The scoring module 304 is configured to score the N candidate merchant POIs based on a pre-trained deep neural network model.

[0151] The second screening module 305 is configured to screen the target merchant POI corresponding to the target Wi-Fi from the N candidate merchant POIs according to the scoring results.

[0152] Optionally, the first screening module comprises:

[0153] The first sub-module is configured to perform semantic extraction on an SSID of the target Wi-Fi to obtain a processed SSID in a case where the target merchant POI corresponding to the target Wi-Fi does not exist in the word table.

[0154] The second sub-module is configured to calculate an edit distance score of the processed SSID and each candidate merchant POI.

[0155] The third sub-module is configured to screen a candidate merchant POI whose edit distance score is ranked in TopN.

[0156] Optionally, the scoring module comprises:

[0157] The fourth sub-module is configured to input each candidate merchant POI in the N candidate merchant POIs and the processed SSID into a pre-trained deep neural network model.

[0158] The fifth sub-module is configured to determine an output of the deep neural network model as a scoring result of the candidate merchant POI.

[0159] Optionally, the second screening module comprises:

[0160] The sixth sub-module is configured to screen a first candidate merchant POI with the highest score according to the scoring results.

[0161] The seventh sub-module is configured to determine the first candidate merchant POI as a target merchant POI corresponding to the target Wi-Fi when the first candidate merchant POI is single.

[0162] Optionally, the apparatus further comprises:

[0163] The eighth sub-module is configured to, after the sixth sub-module filters out the first candidate merchant POI with the highest score according to each score result, filter out a target merchant POI from each first candidate merchant POI according to preset parameter information when the number of the first candidate merchant POIs is greater than one, wherein the preset parameter information comprises at least one of the following: a distance between the first candidate merchant corresponding to the first candidate merchant POI and the target Wi-Fi, and a Wi-Fi feature contained in the business data of the first candidate merchant.

[0164] Optionally, the apparatus further comprises:

[0165] The positive sample determination module is configured to determine a plurality of binding relationships between Wi-Fi and merchant POIs obtained by analyzing the business data as positive samples before the establishment module establishes the multilingual word list of merchant POIs.

[0166] The processing module is configured to process each positive sample according to a second preset rule to obtain a processed positive sample.

[0167] The generation module is configured to generate a negative sample according to a third preset rule.

[0168] The training module is configured to train the deep neural network model according to each processed positive sample and the negative sample.

[0169] Optionally, the processing module comprises:

[0170] The ninth sub-module is configured to, for each binding relationship between Wi-Fi and merchant POI, sequentially perform a preset processing on the SSID of the Wi-Fi in the binding relationship to obtain a processed SSID, wherein the preset processing comprises: SSID semantic extraction, Chinese to Pinyin conversion, and letter string segmentation.

[0171] The tenth sub-module is configured to perform word segmentation processing on the merchant POI in the binding relationship to obtain a processed merchant POI.

[0172] The eleventh sub-module is configured to determine the SSID of the Wi-Fi, the processed SSID, the merchant POI, and the processed merchant POI as the processed positive sample.

[0173] Optionally, the generation module comprises:

[0174] The twelfth sub-module is configured to select a first preset number of positive samples;

[0175] The thirteenth sub-module is configured to modify at least one letter of the SSID of the Wi-Fi of each positive sample to generate a modified SSID.

[0176] The fourteenth sub-module is configured to determine the binding relationship between the modified SSID and the merchant POI in the positive sample as a negative sample.

[0177] Optionally, the generating module comprises:

[0178] The fifteenth sub-module is configured to select a second preset number of positive samples.

[0179] The sixteenth sub-module is configured to, for each positive sample, find a third preset number of second merchant POIs satisfying a preset matching condition according to the Wi-Fi in the positive sample.

[0180] The seventeenth sub-module is configured to establish a binding relationship between each second merchant POI and the Wi-Fi respectively to generate the third preset number of negative samples.

[0181] For the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the related parts refer to the part of the method embodiment.

[0182] In the embodiments of the present application, an electronic device is also provided. The electronic device can include one or more processors, and one or more machine-readable media having instructions, such as an application, stored thereon. When the instructions are executed by the one or more processors, the processor performs the method of determining the corresponding relationship between Wi-Fi and merchant POI described above.

[0183] In the embodiments of the present application, a non-transitory computer readable storage medium is also provided, which stores a computer program executable by a processor of an electronic device to complete the method of determining the corresponding relationship between Wi-Fi and merchant POI described above. For example, the non-transitory computer readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, and an optical data storage device, etc.

[0184] Each embodiment in the specification is described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The same and similar parts of each embodiment can be referred to each other.

[0185] Those skilled in the art will appreciate that embodiments of the present application can be readily used as a method, apparatus, or computer program product. Accordingly, embodiments of the present application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, embodiments of the present application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, and the like) embodying computer program instructions.

[0186] Embodiments of the present application are described herein with reference to the drawings, which are as follows: Figure 1 Figure 1

[0187] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the Figure 1 Figure 1

[0188] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the Figure 1 Figure 1

[0189] While preferred embodiments of the present application have been described, additional variations and modifications can be made to these embodiments by those skilled in the art once they have the benefit of the foregoing description. Therefore, the appended claims are intended to encompass within their scope all possible embodiments of the present application.

[0190] ​​​​​​Finally, it is to be understood that the phraseology or terminology such as "first" and "second" etc. used herein is merely intended to differentiate one entity or operation from another entity or operation, without necessarily requiring or implying any actual such relationship or order between such entities or operations. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises... a" does not, without more constraints, exclude the existence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0191] The above describes in detail the method and device for determining the correspondence between Wi-Fi and merchant POI, the electronic device and the storage medium provided by the present application. The principles and implementation manners of the present application are described by using specific examples. The above description of the embodiments is only used to help understand the method of the present application and its core idea. Meanwhile, for those skilled in the art, the specific implementation manners and application ranges can be changed according to the idea of the present application. In summary, the content of the specification should not be understood as a limitation of the present application.< / sep> < / sep>

Claims

1. A method for determining a Wi-Fi and merchant POI correspondence relationship, characterized in that, The method comprises the following steps: establishing a multilingual POI word table of a merchant, wherein the word table contains a multilingual translation mapping relationship of each merchant POI to be mined; based on each mapping relationship, checking whether a target merchant POI corresponding to a target Wi-Fi exists in the word table; if not, selecting N candidate merchant POIs from the merchant POIs to be mined according to a first preset rule; based on a pre-trained deep neural network model, scoring each of the N candidate merchant POIs; according to each scoring result, selecting a target merchant POI corresponding to the target Wi-Fi from the N candidate merchant POIs; before the step of establishing the multilingual POI word table of the merchant, the method further comprises the following steps: determining a plurality of Wi-Fi and merchant POI binding relationships analyzed from business data as positive samples; processing each positive sample according to a second preset rule to obtain a processed positive sample; generating negative samples according to a third preset rule; training a deep neural network model according to the processed positive samples and the negative samples; the step of processing each positive sample according to a second preset rule to obtain a processed positive sample comprises the following steps: for each Wi-Fi and merchant POI binding relationship, sequentially performing a preset processing on the SSID of the Wi-Fi in the binding relationship to obtain a processed SSID, wherein the preset processing includes SSID semantic extraction, Chinese to Pinyin conversion, and letter string segmentation; performing word segmentation processing on the merchant POI in the binding relationship to obtain a processed merchant POI; determining the SSID of the Wi-Fi, the processed SSID, the merchant POI, and the processed merchant POI as the processed positive sample; the step of generating negative samples according to a third preset rule comprises the following steps: selecting a first preset number of positive samples; for each positive sample, modifying at least one letter of the SSID of the Wi-Fi in the positive sample to generate a modified SSID; determining the modified SSID and the binding relationship of the merchant POI in the positive sample as a negative sample; alternatively, the step of generating negative samples according to a third preset rule comprises the following steps: selecting a second preset number of positive samples; for each positive sample, searching for a third preset number of second merchant POIs that satisfy a preset matching condition according to the Wi-Fi in the positive sample; establishing the binding relationship of each second merchant POI and the Wi-Fi to generate the third preset number of negative samples.

2. The method of claim 1, wherein, the step of selecting N candidate merchant POIs from the merchant POIs to be mined according to a first preset rule comprises the following steps: in the case where the target merchant POI corresponding to the target Wi-Fi does not exist in the word table, performing semantic extraction on the SSID of the target Wi-Fi to obtain a processed SSID; calculating the edit distance score of the processed SSID and each candidate merchant POI; selecting candidate merchant POIs whose edit distance scores are ranked in the top N.

3. The method of claim 2, wherein, The step of scoring each of the N candidate merchant POIs based on the pre-trained deep neural network model includes: Inputting the candidate merchant POI and the processed SSID into the pre-trained deep neural network model; The output of the deep neural network model is determined as the scoring result of the candidate merchant POI.

4. The method of claim 1, wherein, The step of screening the target merchant POI corresponding to the target Wi-Fi from the N candidate merchant POIs according to the scoring results includes: Screening a first candidate merchant POI with the highest score according to the scoring results; In the case where the first candidate merchant POI is a single one, the first candidate merchant POI is determined as the target merchant POI corresponding to the target Wi-Fi.

5. The method of claim 4, wherein, After the step of screening the first candidate merchant POI with the highest score according to the scoring results, the method further includes: In the case where the number of the first candidate merchant POIs is greater than 1, screening a target merchant POI from the first candidate merchant POIs according to preset parameter information, wherein the preset parameter information includes at least one of the following: the distance between the first candidate merchant corresponding to the first candidate merchant POI and the target Wi-Fi, and the Wi-Fi feature contained in the business data of the first candidate merchant.

6. A device for determining a Wi-Fi and merchant POI correspondence relationship, characterized in that, It includes: A building module is configured to build a multi-lingual word table of merchant POIs, wherein the word table contains a multi-lingual translation mapping relationship of each merchant POI to be mined. A searching module is configured to search whether a target merchant POI corresponding to a target Wi-Fi exists in the word table based on each mapping relationship. A first screening module is configured to screen N candidate merchant POIs from the merchant POIs to be mined according to a first preset rule if not. A scoring module is configured to score each of the N candidate merchant POIs based on a pre-trained deep neural network model. A second screening module is configured to screen the target merchant POI corresponding to the target Wi-Fi from the N candidate merchant POIs according to the scoring results. The step of screening the target merchant POI corresponding to the target Wi-Fi from the N candidate merchant POIs according to the scoring results includes: Before the step of building the multi-lingual word table of merchant POIs, the method further includes: Binding relationships between a plurality of Wi-Fi and merchant POIs analyzed from business data are determined as positive samples. Each positive sample is processed according to a second preset rule to obtain a processed positive sample. Negative samples are generated according to a third preset rule. A deep neural network model is trained according to the processed positive samples and the negative samples. The step of processing each positive sample according to a second preset rule to obtain a processed positive sample includes: For each of the Wi-Fi and merchant POI binding relationship, a preset processing is performed on the SSID of the Wi-Fi in the binding relationship in sequence to obtain a processed SSID, wherein the preset processing includes: SSID semantic extraction, Chinese to Pinyin conversion, and letter string segmentation; The merchant POI in the binding relationship is subjected to word segmentation processing to obtain a processed merchant POI; The SSID of the Wi-Fi, the processed SSID, the merchant POI, and the processed merchant POI are determined as a processed positive sample; The step of generating the negative sample according to the third preset rule includes: selecting a first preset number of positive samples; For each of the positive samples, at least one letter of the SSID of the positive sample Wi-Fi is modified to generate a modified SSID; The binding relationship between the modified SSID and the merchant POI in the positive sample is determined as a negative sample; Alternatively, the step of generating the negative sample according to the third preset rule includes: selecting a second preset number of positive samples; For each of the positive samples, a third preset number of second merchant POIs satisfying a preset matching condition are found according to the Wi-Fi in the positive sample; The binding relationship between each of the second merchant POIs and the Wi-Fi is established to generate the third preset number of negative samples.

7. An electronic device, comprising: comprise: one or more processors; and one or more machine-readable media having stored thereon instructions; When the instructions are executed by the one or more processors, the processors perform the method for determining the corresponding relationship between the Wi-Fi and the merchant POI according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, a computer program is stored thereon, which is executed by the processor to implement the method for determining the corresponding relationship between the Wi-Fi and the merchant POI according to any one of claims 1 to 5.

Images