WO2024251004A1 - 模型训练方法、装置、电子设备及存储介质 - Google Patents

模型训练方法、装置、电子设备及存储介质 Download PDF

Info

Publication number
WO2024251004A1
WO2024251004A1 PCT/CN2024/096051 CN2024096051W WO2024251004A1 WO 2024251004 A1 WO2024251004 A1 WO 2024251004A1 CN 2024096051 W CN2024096051 W CN 2024096051W WO 2024251004 A1 WO2024251004 A1 WO 2024251004A1
Authority
WO
WIPO (PCT)
Prior art keywords
information
wifi
interest
point
occurrence information
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2024/096051
Other languages
English (en)
French (fr)
Inventor
冯朝阳
尹卜一
焦恒建
王畔
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Douyin Vision Co Ltd
Original Assignee
Douyin Vision Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Douyin Vision Co Ltd filed Critical Douyin Vision Co Ltd
Publication of WO2024251004A1 publication Critical patent/WO2024251004A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
    • H04W4/02—Services making use of location information
    • H04W4/021—Services related to particular areas, e.g. point of interest [POI] services, venue services or geofences
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00—Reducing energy consumption in communication networks
    • Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks

Definitions

  • the present invention relates to a model training method, a model training device, an electronic device and a computer storage medium.
  • WiFi fingerprint positioning technology is one of the important technologies for realizing point of interest (POI) positioning. It is widely used in the industry because it does not require manual deployment of equipment and has good spatial distribution characteristics. Among the common fingerprint positioning technologies, the user's current point of interest is usually determined based on the cross-information between the WiFi information scanned by the user's electronic device and the historical WiFi fingerprint information collected.
  • the embodiments of the present disclosure at least provide a model training method, device, electronic device and storage medium, which can improve the credibility of positioning results.
  • the present disclosure provides a model training method, including:
  • the historical WiFi scanning information includes multiple groups of WiFi scanning information, each group of WiFi scanning information includes at least one first WiFi information, and the historical WiFi distribution information includes second WiFi information associated with each point of interest;
  • each point of interest based on the second WiFi information associated with the point of interest and at least one first WiFi information in each group of WiFi scanning information, generate co-occurrence information corresponding to the point of interest; the co-occurrence information is used to characterize the scanning state of each first WiFi information in each group relative to the second WiFi information associated with the point of interest;
  • each co-occurrence information For each co-occurrence information, generating training sample data based on the co-occurrence information and a label corresponding to the co-occurrence information; the label corresponding to the co-occurrence information is determined based on preset offline behavior information associated with the point of interest corresponding to the co-occurrence information;
  • the model to be trained is trained to obtain a trained model; the trained model is used to determine the corresponding point of interest positioning information based on the current WiFi scanning information.
  • each point of interest has corresponding point of interest identification information; and generating co-occurrence information corresponding to the point of interest based on the second WiFi information associated with the point of interest and at least one first WiFi information in each set of WiFi scan information includes:
  • For each point of interest based on the second WiFi information associated with the point of interest and the first WiFi information in each group, generate scanning distribution identification information corresponding to the point of interest;
  • the co-occurrence information corresponding to the point of interest is generated.
  • co-occurrence information corresponding to the point of interest is generated based on the unique point of interest identification information of the point of interest and the scanning distribution identification information corresponding to the point of interest.
  • the co-occurrence information has the characteristics of the point of interest and the corresponding associated WiFi characteristics.
  • the model can learn the association between the point of interest and the corresponding WiFi, which is beneficial to improving the accuracy of the model.
  • the first WiFi information includes a first WiFi and a signal strength of the first WiFi
  • the second WiFi information includes a second WiFi and a signal strength of the second WiFi
  • the step of generating, for each point of interest, scanning distribution identification information corresponding to the point of interest based on the second WiFi information associated with the point of interest and the first WiFi information in each group, includes:
  • the sub-scanning distribution identification information is spliced according to the distribution order between the second WiFis associated with the points of interest to generate the scanning distribution identification information.
  • each point of interest is associated with at least one second WiFi
  • corresponding sub-scanning distribution identification information can be generated for each second WiFi, and then each sub-scanning The distribution identification information is spliced to obtain the scan distribution identification information, so that the accuracy of the scan distribution identification information can be improved, thereby improving the accuracy of the subsequent generation of training sample data, and further improving the accuracy of model training.
  • the sub-scan distribution identification information corresponds to a scan flag bit and multiple signal strength flag bits, and different signal strength flag bits are used to represent different signal strengths; the generating of the sub-scan distribution identification information for each second WiFi associated with each point of interest based on the second WiFi and each first WiFi in each group includes:
  • the sub-scan distribution identification information is generated based on the value of the scan flag bit and the value of the signal strength flag bit.
  • the sub-scanning distribution identification information is determined by determining the value of the scanning flag bit and the value of the signal strength flag bit, which is helpful to improve the accuracy of the sub-scanning distribution identification information.
  • the format of the co-occurrence information is a character string format; and for each piece of co-occurrence information, generating training sample data based on the co-occurrence information and a label corresponding to the co-occurrence information includes:
  • feature processing is performed on the co-occurrence information in the string format to generate a feature vector corresponding to the co-occurrence information
  • the training sample data is generated based on the feature vector and the true value corresponding to the label.
  • the true value of the feature vector is determined based on the preset offline behavior information, and the true value is calibrated for the feature vector, which can improve the accuracy of the training sample data.
  • determining the true value corresponding to the tag based on the preset offline behavior information includes:
  • the true value corresponding to the label is determined to be 1; or,
  • the true value corresponding to the label is determined to be 0.
  • a preset offline behavior occurs, it means that the preset offline behavior has been performed within the point of interest, and the true value corresponding to the label is determined to be 1. If the preset offline behavior does not occur, it means that the preset offline behavior has not been performed within the point of interest, and the true value corresponding to the label is determined to be 0. The accuracy of the true value determination can be improved, and the accuracy of the training sample data can be improved.
  • performing feature processing on the co-occurrence information in a string format to generate a feature vector corresponding to the co-occurrence information includes:
  • the feature dictionary is used to convert the co-occurrence information in the string format into a feature vector
  • one-hot encoding is performed on the co-occurrence information in the string format based on the feature dictionary to generate a feature vector corresponding to the co-occurrence information.
  • the co-occurrence information in string format is converted into features by constructing a feature dictionary, which is conducive to improving the efficiency of subsequent model training.
  • the one-hot encoding can represent the corresponding feature vector through a state code, the logic of feature conversion can be simplified by the one-hot encoding method.
  • performing one-hot encoding processing on the co-occurrence information in the string format based on the feature dictionary to generate a feature vector corresponding to the co-occurrence information includes:
  • the value of the target index bit in the zero vector that is the same as the number is modified to 1, and a feature vector corresponding to the co-occurrence information is generated.
  • the target index bit in the zero vector is modified based on the serial number of the co-occurrence information in the feature dictionary, so that the accuracy of the feature vector can be improved.
  • the step of training the model to be trained based on each training sample data to obtain a trained model includes:
  • the model parameters of the model to be trained are adjusted;
  • the model is supervised trained based on the prediction result corresponding to each training sample data and the label of each training sample data. In this way, the performance of the model training can be improved, thereby improving the prediction accuracy of the model.
  • the present disclosure provides a method for locating a point of interest, including:
  • the location information of the points of interest of the current device is determined based on the trained model, so that the accuracy of the location of the points of interest can be improved.
  • the present disclosure provides a model training device, including:
  • An information acquisition module used to acquire historical wireless fidelity WiFi scanning information and historical WiFi distribution information;
  • the historical WiFi scanning information includes multiple groups of WiFi scanning information, each group of WiFi scanning information includes at least one first WiFi information, and the historical WiFi distribution information includes second WiFi information associated with each point of interest;
  • An information generation module configured to generate, for each point of interest, co-occurrence information corresponding to the point of interest based on the second WiFi information associated with the point of interest and at least one first WiFi information in each group of WiFi scanning information; the co-occurrence information is used to characterize a scanning state of each first WiFi information in each group relative to the second WiFi information associated with the point of interest;
  • a sample generation module for generating training sample data for each co-occurrence information based on the co-occurrence information and a label corresponding to the co-occurrence information; the label corresponding to the co-occurrence information is determined based on preset offline behavior information associated with the point of interest corresponding to the co-occurrence information;
  • the model training module is used to train the model to be trained based on various training sample data to obtain a trained model; the trained model is used to determine the corresponding point of interest positioning information based on the current WiFi scanning information.
  • each point of interest has corresponding point of interest identification information; the information generation module is specifically used to:
  • For each point of interest based on the second WiFi information associated with the point of interest and the first WiFi information in each group, generate scanning distribution identification information corresponding to the point of interest;
  • the co-occurrence information corresponding to the point of interest is generated.
  • the first WiFi information includes a first WiFi and a signal strength of the first WiFi
  • the second WiFi information includes a second WiFi and a signal strength of the second WiFi
  • the information generation module is specifically used to:
  • the sub-scanning distribution identification information is spliced according to the distribution order between the second WiFis associated with the points of interest to generate the scanning distribution identification information.
  • the sub-scan distribution identification information corresponds to a scan flag bit and a plurality of signal strength flag bits, and different signal strength flag bits are used to represent different signal strengths;
  • the information generation module is specifically used to:
  • the sub-scan distribution identification information is generated based on the value of the scan flag bit and the value of the signal strength flag bit.
  • the format of the co-occurrence information is a character string format; and the sample generation module is specifically used to:
  • feature processing is performed on the co-occurrence information in the string format to generate a feature vector corresponding to the co-occurrence information
  • the training sample data is generated based on the feature vector and the true value corresponding to the label.
  • the sample generation module is specifically used to:
  • the true value corresponding to the label is determined to be 1; or,
  • the true value corresponding to the label is determined to be 0.
  • the sample generation module is specifically used to:
  • the feature dictionary is used to convert the co-occurrence information in the string format into a feature vector
  • one-hot encoding is performed on the co-occurrence information in the string format based on the feature dictionary to generate a feature vector corresponding to the co-occurrence information.
  • the sample generation module is specifically used to:
  • the value of the target index bit in the zero vector that is the same as the number is modified to 1, and a feature vector corresponding to the co-occurrence information is generated.
  • model training module is specifically used to:
  • the present disclosure provides a device for locating a point of interest, including:
  • An acquisition module used to acquire current WiFi scanning information of a current device and associated WiFi information corresponding to each point of interest corresponding to an area where the current device is located;
  • a generating module configured to generate current co-occurrence information for each point of interest based on the current WiFi scanning information and the associated WiFi information corresponding to the point of interest;
  • a positioning module is used to input each current co-occurrence information into a trained model to obtain the point of interest positioning information of the current device; wherein the trained model is obtained by the model training method described in any one of claims 1-9.
  • An embodiment of the present disclosure also provides an electronic device, including: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate through the bus, and when the machine-readable instructions are executed by the processor, the steps of the model training method described in any possible implementation manner described above or the steps of the point of interest positioning method described above are performed.
  • the embodiments of the present disclosure also provide a computer-readable storage medium having a computer program stored thereon.
  • the computer program is executed by a processor, the steps of the model training method described in any of the above possible implementation modes or the steps of the above point of interest positioning method are executed.
  • FIG1 shows a flow chart of a model training method provided by an embodiment of the present disclosure
  • FIG2 shows a flow chart of a method for generating co-occurrence information provided by an embodiment of the present disclosure
  • FIG3 is a schematic diagram showing a process of generating sub-scan distribution identification information provided by an embodiment of the present disclosure
  • FIG4 shows a flow chart of a method for generating training data samples provided by an embodiment of the present disclosure
  • FIG5 shows a flow chart of a method for processing features of co-occurrence information provided by an embodiment of the present disclosure
  • FIG6 shows a flow chart of a method for locating a point of interest provided by an embodiment of the present disclosure
  • FIG7 shows a schematic diagram of a model training device provided by an embodiment of the present disclosure
  • FIG8 shows a schematic diagram of a device for locating a point of interest provided by an embodiment of the present disclosure.
  • FIG. 9 shows a schematic diagram of an electronic device provided by an embodiment of the present disclosure.
  • a and/or B may represent the following three situations: A exists alone, A and B exist at the same time, and B exists alone.
  • at least one herein represents any combination of at least two of any one or more of a plurality of.
  • including at least one of A, B, and C may represent including any one or more elements selected from the set consisting of A, B, and C.
  • the click-through rate (CTR) prediction model is one of the important technologies used to learn and predict user feedback information, where the user feedback information can be information about the user's clicks, favorites, or purchases.
  • the CTR model achieves the above-mentioned information recommendation and other functions through its memory ability, where the memory ability refers to the model's ability to directly learn and use the "co-occurrence frequency" of requests and historical fingerprints in historical data.
  • models such as collaborative filtering models and logistic regression models have strong "memory". Because of the simple structure of this type of model, historical data can often directly affect the recommendation results. That is, the model can learn the distribution characteristics of historical data and use its own memory to predict the results.
  • the present disclosure provides a model training method, device, electronic device and storage medium, firstly obtaining historical wireless fidelity WiFi scanning information and historical WiFi distribution information;
  • the historical WiFi scanning information includes multiple groups of WiFi scanning information, each group of WiFi scanning information includes at least one first WiFi information, and the historical WiFi distribution information includes the second WiFi information associated with each point of interest; secondly, for each point of interest, based on the second WiFi information associated with the point of interest and at least one first WiFi information in each group of WiFi scanning information, co-occurrence information corresponding to the point of interest is generated; the co-occurrence information is used to characterize the scanning state of each first WiFi information in each group relative to the second WiFi information associated with the point of interest; then, for each co-occurrence information, based on the co-occurrence information and the label corresponding to the co-occurrence information, training sample data is generated; the label corresponding to the co-occurrence information is determined based on the preset offline behavior information associated with the point of interest corresponding to the co-occurrence information; finally, based on each training sample data, the model to
  • corresponding co-occurrence information is constructed based on historical WiFi distribution information and historical WiFi scanning information. Since the co-occurrence information represents the scanning status of the WiFi information scanned by each first user through the electronic device relative to at least one WiFi information covering the point of interest, therefore, training the model by generating training sample data based on the co-occurrence information can enhance the model's memory of the association between the WiFi scanning information and the point of interest, and obtain a trained model. Furthermore, the point of interest can be predicted based on the trained model, thereby improving the credibility of the positioning result.
  • the executor of the model training method provided in the embodiment of the present disclosure is an electronic device.
  • the electronic device is a server, which can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud storage, big data and artificial intelligence platforms.
  • the electronic device can also be a terminal device.
  • the terminal device can be a mobile device, a user terminal, a handheld device, a computing device, a wearable device, etc.
  • the model training method can also be called by the processor in the memory This is accomplished by storing computer-readable instructions.
  • FIG. 1 is a flow chart of the model training method provided by the embodiment of the present disclosure, the method includes steps S101 to S104, wherein:
  • the historical WiFi scanning information includes multiple groups of WiFi scanning information, each group of WiFi scanning information includes at least one first WiFi information, and the historical WiFi distribution information includes second WiFi information associated with each point of interest.
  • the historical WiFi scanning information may be WiFi information scanned by an electronic device, that is, may refer to multiple groups of WiFi scanning information scanned by different electronic devices, and since WiFi signals have corresponding coverage ranges, each group of WiFi scanning information includes at least one first WiFi information.
  • the first WiFi information includes a first WiFi and a signal strength of the first WiFi, and each first WiFi corresponds to a point of interest.
  • the electronic device may be a terminal device, such as a smart phone, a tablet computer, or a smart watch, etc., which is not limited here.
  • the historical WiFi information is the WiFi information collected in the historical time (such as the previous month).
  • the historical WiFi distribution information includes second WiFi information associated with each point of interest, the second WiFi information includes the second WiFi and the signal strength of the second WiFi, and there is a distribution order between the second WiFi associated with each point of interest, and the distribution order is determined by the signal strength of each second WiFi.
  • the historical WiFi fingerprint distribution information may be WiFi distribution information in a point of interest (POI) dimension constructed based on historical WiFi scanning information, that is, the historical WiFi fingerprint distribution information may be a WiFi distribution constructed based on multiple sets of WiFi scanning information scanned for each point of interest within a historical time period, and since the signal strengths of the second WiFis are different, for each point of interest, there is a corresponding distribution order between the second WiFis associated with the point of interest.
  • POI point of interest
  • its corresponding second WiFi may include W1, W2, and W3, wherein W1, W2, and W3 have different distances from the point of interest A, resulting in different signal strengths of the second WiFis associated with the point of interest A.
  • the signal strengths of W1, W2, and W3 are W1>W2>W3
  • the distribution order of the second WiFis may be W1, W2, and W3. W3.
  • the point of interest may refer to a store in a target location, which may be a shopping mall, an office building, etc.
  • the point of interest may be a store in a shopping mall.
  • Each point of interest has corresponding unique point of interest identification information, which may be identification information composed of numbers, letters, etc.
  • S102 For each point of interest, based on the second WiFi information associated with the point of interest and at least one first WiFi information in each group of WiFi scanning information, generate co-occurrence information corresponding to the point of interest; the co-occurrence information is used to characterize a scanning state of each first WiFi information in each group relative to the second WiFi information associated with the point of interest.
  • the scanning status may refer to a scanned state and a not scanned state.
  • the second WiFi information associated with the point of interest can be compared with at least one first WiFi information in each group of WiFi scanning information. In this way, it can be determined whether there is first WiFi information in the at least one first WiFi information that is the same as the second WiFi information associated with the point of interest, and further the scanning status of each first WiFi information relative to the second WiFi information associated with the point of interest can be determined.
  • step S102 when generating co-occurrence information corresponding to each point of interest based on the second WiFi information associated with the point of interest and at least one first WiFi information in each set of WiFi scan information, referring to FIG. 2, the following steps S1021 to S1022 may be included:
  • S1021 For each point of interest, based on the second WiFi information associated with the point of interest and the first WiFi information in each group, generate scanning distribution identification information corresponding to the point of interest.
  • the second WiFi information associated with the point of interest and the first WiFi information in each group of WiFi scanning information may be compared, thereby generating scanning distribution identification information corresponding to the point of interest.
  • sub-scan distribution identification information is generated based on the second WiFi and each first WiFi in each group.
  • the sub-scan distribution identification information corresponds to a scan flag bit and a plurality of signal strength flag bits, and the scan flag bit is used to indicate whether the corresponding second WiFi is scanned.
  • the signal strength flag is used to characterize the signal strength, wherein different signal strength flags correspond to different signal strengths.
  • the signal strength flag includes 8 bits, for example, -70dBm to -60dBm corresponds to signal strength flag 0, and -60dBm to -50dBm corresponds to signal strength flag 1.
  • a scan flag bit can be determined based on whether the first WiFi in each group of first WiFi information is the same as the second WiFi, and a signal strength flag bit can be determined based on the signal strength of each first WiFi in the first WiFi information, and sub-scan distribution identification information can be generated based on the scan flag bit and the signal strength flag bit.
  • steps (a) to (c) can be included:
  • ID n is the sub-scan distribution identification information
  • the scan flag Tag scan is used to indicate that there is a target first WiFi in the first WiFi information that has the same media access control address MAC address as the second WiFi. If so, the value of the scan flag is determined to be 1; if not, the value of the scan flag is determined to be 0; I represents the index value of the signal strength of the target first WiFi falling into the signal bucket, and the signal bucket refers to the signal range.
  • FIG3, is a schematic diagram of a generation process of sub-scan distribution identification information provided by an embodiment of the present disclosure.
  • there are 5 second WiFis (W1, W2, W3, W4, and W5) distributed at the point of interest POI-1.
  • W1 is compared with each first WiFi in each group of first WiFi information. If there is a target first WiFi with the same MAC address as W1, the value of the scan flag is determined to be 1. Then, according to the signal strength of the target first WiFi, the target signal strength flag corresponding to its signal strength is determined.
  • the corresponding sub-scan distribution identification information bit can be determined to be 000
  • the corresponding sub-scan distribution identification information for W3 is 272
  • the corresponding sub-scan distribution identification information for W4 is 258, and the corresponding sub-scan distribution identification information for W5 is 288.
  • Each sub-scanning distribution identification information is spliced according to the distribution order between the second WiFi associated with the point of interest to generate the scanning distribution identification information.
  • each sub-scan distribution identification information After obtaining each sub-scan distribution identification information, it can be spliced according to the distribution order of each second WiFi. As shown in Figure 3, the obtained sub-scan distribution identification information is 272, 000, 272, 258 and 288, which are spliced to obtain the scan distribution identification information 272000272258288.
  • S1022 Generate the co-occurrence information corresponding to the point of interest based on the point of interest identification information of the point of interest and the scan distribution identification information corresponding to the point of interest.
  • the co-occurrence information corresponding to the point of interest can be generated based on the point of interest identification information and the scanning distribution identification information corresponding to the point of interest.
  • the point of interest identification information and the scanning distribution identification information corresponding to the point of interest can be spliced to generate co-occurrence information.
  • the format of the co-occurrence information is a string format.
  • the interest point identification information is 22535659086281011, and the scan distribution identification information corresponding to the interest point is 272000272258288, then the co-occurrence information is: 272000272258288_22535659086281011.
  • the preset offline behavior information may refer to whether a preset offline behavior occurs.
  • the preset offline behavior may refer to offline consumption behavior, such as coupon redemption and the like.
  • the first WiFi in the first WiFi scanning information includes W11, W12, ..., W19, which correspond to points of interest POI1, POI2, ..., POI9, respectively. No offline behavior occurs at points of interest POI0, POI2, ..., POI9, and the corresponding co-occurrence information will not be used to generate training sample data.
  • step S103 when generating training sample data for each co-occurrence information based on the co-occurrence information and the label corresponding to the co-occurrence information, referring to FIG. 4 , the following S1031 to S1033 may be included:
  • the co-occurrence information is in a string format, it needs to be converted into a corresponding feature vector in order to facilitate model training in subsequent steps.
  • feature processing is performed on the co-occurrence information in the string format to generate a feature vector corresponding to the co-occurrence information, referring to FIG. 5 , including the following S10311 to S10312:
  • each piece of co-occurrence information may be uniformly numbered so that each piece of co-occurrence information uniquely corresponds to an integer. For example, see Table 1, which shows the correspondence between the co-occurrence information and the number.
  • S10312 For each piece of co-occurrence information in the string format, perform one-hot encoding processing on the co-occurrence information in the string format based on the feature dictionary to generate a feature vector corresponding to the co-occurrence information.
  • the feature vector is a high-dimensional sparse feature vector.
  • the co-occurrence information in the string format can be encoded.
  • a one-hot encoding method is adopted, wherein the one-hot encoding is also called One-hot encoding, and the method is to use an N-bit state register to encode N states, each state has an independent register bit, and only one register bit is valid at any time.
  • the number of numbers contained in the feature dictionary can be first determined, and a zero vector with a vector length of the number can be created. Then, for each co-occurrence information, according to the number corresponding to the co-occurrence information in the feature dictionary, the value of the target index bit in the zero vector that is the same as the number is modified to 1 to generate a feature vector corresponding to the co-occurrence information.
  • Table 2 shows the conversion process of the feature vector obtained by performing feature conversion on the co-occurrence information based on the feature dictionary.
  • a zero vector with a vector length of 9 is created.
  • the number corresponding to the co-occurrence information is determined to be 2 in the feature dictionary, and the value of the register bit with index bit 2 in the zero vector is modified to 1, generating a feature vector 001000000 corresponding to the co-occurrence information.
  • S1032 Obtain preset offline behavior information associated with the point of interest corresponding to the co-occurrence information, and determine a true value corresponding to the tag based on the preset offline behavior information.
  • the true value corresponding to the label is determined to be 1.
  • the true value corresponding to the label is determined to be 0.
  • S1033 Generate the training sample data based on the feature vector and the true value corresponding to the label.
  • the true value of each feature vector determined in the above embodiment can be calibrated to obtain training sample data.
  • the model to be trained may refer to a logistic regression model.
  • the model to be trained may also be other models, which is not limited here.
  • the model to be trained can be trained based on the training sample data (or the training sample data can be fitted based on the logistic regression model to be trained) to obtain a trained model.
  • the model to be trained can be supervised.
  • corresponding co-occurrence information is constructed based on historical WiFi distribution information and historical WiFi scanning information. Since the co-occurrence information represents the scanning status of each group of scanned first WiFi scanning information relative to at least one second WiFi information associated with the point of interest, therefore, training the model by generating training sample data based on the co-occurrence information can enhance the model's memory of the association relationship between the WiFi scanning information and the point of interest, and obtain a trained model. Furthermore, the point of interest can be predicted based on the trained model, thereby improving the credibility of the positioning result.
  • each training sample data can be input into the model to be trained to obtain a prediction result corresponding to each training sample data, and then, based on the prediction result corresponding to each training sample data and the label of each training sample data, the model parameters of the model are adjusted.
  • the loss function can be set in advance, and the loss value between the prediction result corresponding to each training sample data and the label of each training sample data is calculated, and the model parameters of the model are adjusted based on the loss value. In this way, the above process is repeated until the training result meets the preset requirements to obtain the trained model.
  • FIG6 is a method for locating a point of interest provided by an embodiment of the present disclosure.
  • S601 obtaining current WiFi scanning information of a current device and associated WiFi information corresponding to each point of interest corresponding to an area where the current device is located.
  • current WiFi scanning information may be obtained through the current electronic device, wherein the current WiFi scanning information includes at least one WiFi currently scanned and the signal strength of each WiFi.
  • the area where the current device is located can be determined according to the location information of the current device, and the associated WiFi information corresponding to each point of interest can be pre-constructed.
  • S602 For each point of interest, generate current co-occurrence information based on the current WiFi scan information and associated WiFi information corresponding to the point of interest.
  • the current co-occurrence information is obtained by combining the current WiFi scan information with the POI identification information of each POI respectively, and performing one-hot encoding (One-hot) on the combined information, wherein the current co-occurrence information is in vector form.
  • S603 Input each current co-occurrence information into the trained model to obtain the interest point positioning information of the current device.
  • the trained model is obtained by any of the model training methods described above.
  • the probabilities corresponding to each point of interest corresponding to the area where the current device is located can be output. Therefore, in the process of POI positioning, it is only necessary to input the current co-occurrence information in the form of a vector obtained above into the trained model, and the model can output the probabilities corresponding to each point of interest. That is, according to the probabilities corresponding to each point of interest, the point of interest positioning information of the current device can be determined. Specifically, the point of interest corresponding to the maximum probability can be determined as the point of interest where the current device is located.
  • information recommendations can be made to users who want to use the current device based on the point of interest where the device is located.
  • the embodiment of the present disclosure also provides a model training device corresponding to the model training method. Since the principle of solving the problem by the device in the embodiment of the present disclosure is similar to the above-mentioned model training method in the embodiment of the present disclosure, the implementation of the device can refer to the implementation of the method, and the repetition is not repeated. Elaborate.
  • the device includes:
  • the information acquisition module 701 is used to acquire historical wireless fidelity WiFi scanning information and historical WiFi distribution information;
  • the historical WiFi scanning information includes multiple groups of WiFi scanning information, each group of WiFi scanning information includes at least one first WiFi information, and the historical WiFi distribution information includes second WiFi information associated with each point of interest;
  • the information generating module 702 is used to generate, for each point of interest, co-occurrence information corresponding to the point of interest based on the second WiFi information associated with the point of interest and at least one first WiFi information in each group of WiFi scanning information; the co-occurrence information is used to represent the scanning state of each first WiFi information in each group relative to the second WiFi information associated with the point of interest;
  • the sample generation module 703 is used to generate training sample data for each co-occurrence information based on the co-occurrence information and a label corresponding to the co-occurrence information; the label corresponding to the co-occurrence information is determined based on the preset offline behavior information associated with the point of interest corresponding to the co-occurrence information;
  • the model training module 704 is used to train the model to be trained based on various training sample data to obtain a trained model; the trained model is used to determine the corresponding point of interest positioning information based on the current WiFi scanning information.
  • each point of interest has corresponding point of interest identification information; the information generation module 702 is specifically used to:
  • For each point of interest based on the second WiFi information associated with the point of interest and the first WiFi information in each group, generate scanning distribution identification information corresponding to the point of interest;
  • the co-occurrence information corresponding to the point of interest is generated.
  • the first WiFi information includes a first WiFi and a signal strength of the first WiFi
  • the second WiFi information includes a second WiFi and a signal strength of the second WiFi
  • the information generation module 702 is specifically used to:
  • the sub-scanning distribution identification information is divided according to the second WiFi associated with the point of interest.
  • the scan distribution identification information is generated by splicing the scan distribution sequence.
  • the sub-scan distribution identification information corresponds to a scan flag bit and a plurality of signal strength flag bits, and different signal strength flag bits are used to represent different signal strengths;
  • the information generation module 702 is specifically used to:
  • the sub-scan distribution identification information is generated based on the value of the scan flag bit and the value of the signal strength flag bit.
  • the format of the co-occurrence information is a string format; the sample generation module 703 is specifically used to:
  • feature processing is performed on the co-occurrence information in the string format to generate a feature vector corresponding to the co-occurrence information
  • the training sample data is generated based on the feature vector and the true value corresponding to the label.
  • sample generation module 703 is specifically configured to:
  • the true value corresponding to the label is determined to be 1; or,
  • the true value corresponding to the label is determined to be 0.
  • sample generation module 703 is specifically configured to:
  • the feature dictionary is used to convert the co-occurrence information in the string format into a feature vector
  • one-hot encoding is performed on the co-occurrence information in the string format based on the feature dictionary to generate a feature vector corresponding to the co-occurrence information.
  • sample generation module 703 is specifically configured to:
  • the value of the target index bit in the zero vector that is the same as the number is modified to 1, and a feature vector corresponding to the co-occurrence information is generated.
  • model training module 704 is specifically used to:
  • FIG8 is a device for locating points of interest provided in an embodiment of the present disclosure, and the device for locating points of interest includes:
  • the acquisition module 801 is used to acquire the current WiFi scanning information of the current device and the associated WiFi information corresponding to each point of interest corresponding to the area where the current device is located;
  • a generating module 802 configured to generate current co-occurrence information for each point of interest based on the current WiFi scanning information and the associated WiFi information corresponding to the point of interest;
  • the positioning module 803 is used to input each current co-occurrence information into a trained model to obtain the interest point positioning information of the current device; wherein the trained model is obtained by the model training method described in any one of claims 1-9.
  • an embodiment of the present disclosure also provides an electronic device.
  • a schematic diagram of the structure of an electronic device 900 provided in an embodiment of the present disclosure includes a processor 901, a memory 902, and a bus 903.
  • the memory 902 is used to store execution instructions, including a memory 9021 and an external memory 9022; the memory 9021 here is also called an internal memory, which is used to temporarily store the operation data in the processor 901, and the data exchanged with the external memory 9022 such as a hard disk.
  • the processor 901 exchanges data with the external memory 9022 through the memory 9021.
  • the memory 902 is specifically used to store the application code for executing the solution of the present application, and the execution is controlled by the processor 901. That is, when the electronic device 900 is running, the processor 901 communicates with the memory 902 through the bus 903, so that the processor 901 executes the application code stored in the memory 902, and then executes the method described in any of the above embodiments.
  • the memory 902 can be, but is not limited to, random access memory (Random Access Memory, RAM), read-only memory (Read Only Memory, ROM), programmable read-only memory (Programmable Read-Only Memory, PROM), erasable read-only memory (Erasable Programmable Read-Only Memory, EPROM), electrically erasable read-only memory (Electric Erasable Programmable Read-Only Memory, EEPROM), etc.
  • RAM Random Access Memory
  • ROM read-only memory
  • PROM programmable read-only memory
  • PROM Programmable Read-Only Memory
  • EPROM Erasable Programmable Read-Only Memory
  • EEPROM Electrically erasable read-only memory
  • Processor 901 may be an integrated circuit chip with signal processing capabilities.
  • the above-mentioned processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components.
  • DSP digital signal processor
  • ASIC application-specific integrated circuit
  • FPGA field programmable gate array
  • the various methods, steps and logic block diagrams disclosed in the embodiments of the present invention may be implemented or executed.
  • the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.
  • the structure illustrated in the embodiment of the present application does not constitute a specific limitation on the electronic device 900.
  • the electronic device 900 may include more or fewer components than shown in the figure, or combine some components, or split some components, or arrange the components differently.
  • the components shown in the figure may be implemented in hardware, software, or a combination of software and hardware.
  • the present disclosure also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the steps of model training in the above method embodiment are executed.
  • the storage medium may be a volatile or non-volatile computer-readable storage medium.
  • the embodiments of the present disclosure also provide a computer program product, which carries a program code.
  • the instructions included in the program code can be used to execute the model training steps in the above method embodiment. For details, please refer to the above method embodiment, which will not be repeated here.
  • the computer program product may be implemented in hardware, software or a combination thereof.
  • the computer program product is embodied as a computer storage medium.
  • the computer program product is embodied as a software product, such as a software development kit (SDK).
  • SDK software development kit
  • the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
  • each functional unit in each embodiment of the present disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.
  • the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor.
  • the technical solution of the present disclosure or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling an electronic device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present disclosure.
  • the aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory, a random access memory, a magnetic disk or an optical disk.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • Signal Processing (AREA)
  • Evolutionary Biology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Medical Informatics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

本公开提供了一种模型训练方法、装置、设备及介质，该方法包括：获取历史无线保真WiFi扫描信息以及历史WiFi分布信息；历史WiFi扫描信息包括多组WiFi扫描信息，每组WiFi扫描信息包括至少一个第一WiFi信息，历史WiFi分布信息包括每个兴趣点关联的第二WiFi信息；针对每个兴趣点，基于兴趣点关联的第二WiFi信息以及每组WiFi扫描信息中的至少一个第一WiFi信息，生成与兴趣点对应的共现信息；针对每个共现信息，基于共现信息以及与共现信息对应的标签，生成训练样本数据；基于各个训练样本数据，对待训练的模型进行训练，得到训练好的模型；训练好的模型用于基于当前WiFi扫描信息确定对应的兴趣点定位信息。

Description

模型训练方法、装置、电子设备及存储介质
本申请要求于2023年6月8日递交的中国专利申请第202310678757.5号的优先权，在此全文引用上述中国专利申请公开的内容以作为本申请的一部分。
技术领域
本公开涉及一种模型训练方法、模型训练装置、电子设备以及计算机存储介质。
背景技术
WiFi指纹定位技术是实现兴趣点(Point of Interest，POI)定位的重要技术之一，由于其无需人工部署设备以及具备良好的空间分布特征，在业内被广泛使用。常见的指纹定位技术中，通常是基于用户通过电子设备扫描到的WiFi信息与采集到的历史WiFi指纹信息之间的交叉信息，来确定用户当前所在的兴趣点。
发明内容
本公开实施例至少提供一种模型训练方法、装置、电子设备以及存储介质，可以提升定位结果的可信度。
本公开实施例提供了一种模型训练方法，包括：
获取历史无线保真WiFi扫描信息以及历史WiFi分布信息；所述历史WiFi扫描信息包括多组WiFi扫描信息，每组WiFi扫描信息包括至少一个第一WiFi信息，所述历史WiFi分布信息包括每个兴趣点关联的第二WiFi信息；
针对每个兴趣点，基于所述兴趣点关联的第二WiFi信息以及每组WiFi扫描信息中的至少一个第一WiFi信息，生成与所述兴趣点对应的共现信息；所述共现信息用于表征每组中的每个第一WiFi信息相对于所述兴趣点关联的第二WiFi信息的扫描状态；
针对每个共现信息，基于所述共现信息以及与所述共现信息对应的标签，生成训练样本数据；所述共现信息对应的标签基于与所述共现信息对应的兴趣点所关联的预设线下行为信息确定；
基于各个训练样本数据，对待训练的模型进行训练，得到训练好的模型；所述训练好的模型用于基于当前WiFi扫描信息确定对应的兴趣点定位信息。
在一种可能的实施方式中，每个兴趣点具有相应的兴趣点标识信息；所述基于所述兴趣点关联的第二WiFi信息以及每组WiFi扫描信息中的至少一个第一WiFi信息，生成与所述兴趣点对应的共现信息，包括：
针对每个兴趣点，基于所述兴趣点关联的第二WiFi信息以及所述每组中的第一WiFi信息，生成与所述兴趣点对应的扫描分布标识信息；
基于所述兴趣点的兴趣点标识信息以及与所述兴趣点对应的扫描分布标识信息，生成所述与所述兴趣点对应的共现信息。
本公开实施例中，基于兴趣点的唯一的兴趣点标识信息，以及兴趣点对应的扫描分布标识信息，生成与兴趣点对应的共现信息，如此，使得共现信息具有兴趣点的特征以及对应的关联WiFi特征，这样，在后续步骤基于共现信息进行模型训练时，即可使得模型能够学习到兴趣点与对应WiFi之间的关联关系，有利于提升模型的精度。
在一种可能的实施方式中，所述第一WiFi信息包括第一WiFi以及所述第一WiFi的信号强度，所述第二WiFi信息包括第二WiFi以及所述第二WiFi的信号强度，每个兴趣点关联的第二WiFi之间存在分布顺序，所述分布顺序由所述各个第二WiFi的信号强度确定；
所述针对每个兴趣点，基于所述兴趣点关联的第二WiFi信息以及所述每组中的第一WiFi信息，生成与所述兴趣点对应的扫描分布标识信息，包括：
针对每个兴趣点关联的每个第二WiFi，基于所述第二WiFi以及所述每组中的每个第一WiFi，生成子扫描分布标识信息；
将各个子扫描分布标识信息按照所述兴趣点关联的第二WiFi之间的分布顺序进行拼接，生成所述扫描分布标识信息。
本公开实施例中，由于每个兴趣点关联有至少一个第二WiFi，因此，可以针对每个第二WiFi，生成对应的子扫描分布标识信息，然后将各个子扫描 分布标识信息进行拼接，得到扫描分布标识信息，如此，可以提升扫描分布标识信息的准确性，从而提升后续生成训练样本数据的准确性，进而提升模型训练的精度。
在一种可能的实施方式中，所述子扫描分布标识信息对应一个扫描标志位以及多个信号强度标志位，不同信号强度标志位用于表征不同信号强度；所述针对每个兴趣点关联的每个第二WiFi，基于所述第二WiFi以及所述每组中的每个第一WiFi，生成子扫描分布标识信息，包括：
针对每个第二WiFi，在存在与所述第二WiFi的媒体访问控制地址MAC地址相同的目标第一WiFi的情况下，确定所述扫描标志位的值为1；
从信号强度标志位中确定所述目标第一WiFi的信号强度对应的目标信号强度标志位，并将所述目标信号强度标志位的值确定为1，将除所述目标信号强度标志位之外的其他信号强度标志位的值确定为0；
基于所述扫描标志位的值以及所述信号强度标志位的值，生成所述子扫描分布标识信息。
本公开实施例中，通过确定扫描标志位的值以及信号强度标志位的值，确定子扫描分布标识信息，有利于提升子扫描分布标识信息的准确性。
在一种可能的实施方式中，所述共现信息的格式为字符串格式；所述针对每个共现信息，基于所述共现信息以及与所述共现信息对应的标签，生成训练样本数据，包括：
针对每个共现信息，对所述字符串格式的共现信息进行特征处理，生成与所述共现信息对应的特征向量；
获取与所述共现信息对应的兴趣点所关联的预设线下行为信息，并基于所述预设线下行为信息确定所述标签对应的真值；
基于所述特征向量以及所述标签对应的真值，生成所述训练样本数据。
本公开实施例中，通过将字符串格式的共现信息转换为特征向量，如此，便于进行后续步骤的模型训练。此外，基于预设线下行为信息确定特征向量的真值，并为特征向量标定真值，可以提升训练样本数据的准确性。
在一种可能的实施方式中，所述基于所述预设线下行为信息确定所述标签对应的真值，包括：
在发生预设线下行为的情况下，将所述标签对应的真值确定为1；或者，
在未发生预设线下行为的情况下，将所述标签对应的真值确定为0。
本公开实施例中，若发生预设线下行为说明在该兴趣点内进行过预设线下行为，该标签对应的真值确定为1，若未发生预设线下行为说明在该兴趣点内未进行过预设线下行为，该标签对应的真值确定为0，可以提升真值确定的准确性，进而可以提升训练样本数据的准确性。
在一种可能的实施方式中，所述针对每个共现信息，对所述字符串格式的共现信息进行特征处理，生成与所述共现信息对应的特征向量，包括：
对字符串格式的各个共现信息按照预设编号方式分别进行编号，并基于所述各个共现信息分别对应的编号，构建特征字典；所述特征字典用于将所述字符串格式的共现信息转换成特征向量；
针对每个字符串格式的共现信息，基于所述特征字典对所述字符串格式的共现信息进行独热编码处理，生成与所述共现信息对应的特征向量。
本公开实施例中，通过构建的特征字典对字符串格式的共现信息进行特征转换，有利于提升后续模型训练的效率。此外，由于独热编码通过一个状态码即可表示对应的特征向量，因此，通过独热编码方式可以简化特征转换的逻辑。
在一种可能的实施方式中，所述基于所述特征字典对所述字符串格式的共现信息进行独热编码处理，生成与所述共现信息对应的特征向量，包括：
确定所述特征字典中包含的编号的数量，并创建向量长度为所述数量的零向量；
针对每个共现信息，根据所述共现信息在所述特征字典中对应的编号，将所述零向量中与所述编号相同的目标索引位的值修改为1，生成与所述共现信息对应的特征向量。
本公开实施例中，针对每个共现信息，基于共现信息在特征字典中的编号，对零向量中的目标索引位进行修改，如此，可以提升特征向量的准确性。
在一种可能的实施方式中，所述基于各个训练样本数据，对待训练的模型进行训练，得到训练好的模型，包括：
将每个训练样本数据输入所述待训练的模型，得到与所述每个训练样本数据对应的预测结果；
基于所述每个训练样本数据对应的预测结果以及所述每个训练样本数 据的标签，对所述待训练的模型的模型参数进行调整；
重复上述过程直到训练结果符合预设要求，得到所述训练好的模型。
本公开实施例中，基于每个训练样本数据对应的预测结果以及每个训练样本数据的标签，对模型进行有监督训练，如此，可以提升模型训练的性能，进而提升模型的预测精度。
本公开实施例提供了一种兴趣点定位方法，包括：
获取当前设备的当前WiFi扫描信息以及所述当前设备所处区域对应的每个兴趣点对应的关联WiFi信息；
针对每个兴趣点，基于所述当前WiFi扫描信息以及所述兴趣点对应的关联WiFi信息，生成当前共现信息；
将各个当前共现信息输入至训练好的模型，得到所述当前设备的兴趣点定位信息；其中，所述训练好的模型通过权利要求1-9中任一项所述的模型训练方法得到。
本公开实施例中，基于训练好的模型来确定当前设备的兴趣点定位信息，可以提升兴趣点定位的精度。
本公开实施例提供了一种模型训练装置，包括：
信息获取模块，用于获取历史无线保真WiFi扫描信息以及历史WiFi分布信息；所述历史WiFi扫描信息包括多组WiFi扫描信息，每组WiFi扫描信息包括至少一个第一WiFi信息，所述历史WiFi分布信息包括每个兴趣点关联的第二WiFi信息；
信息生成模块，用于针对每个兴趣点，基于所述兴趣点关联的第二WiFi信息以及每组WiFi扫描信息中的至少一个第一WiFi信息，生成与所述兴趣点对应的共现信息；所述共现信息用于表征每组中的每个第一WiFi信息相对于所述兴趣点关联的第二WiFi信息的扫描状态；
样本生成模块，用于针对每个共现信息，基于所述共现信息以及与所述共现信息对应的标签，生成训练样本数据；所述共现信息对应的标签基于与所述共现信息对应的兴趣点所关联的预设线下行为信息确定；
模型训练模块，用于基于各个训练样本数据，对待训练的模型进行训练，得到训练好的模型；所述训练好的模型用于基于当前WiFi扫描信息确定对应的兴趣点定位信息。
在一种可能的实施方式中，每个兴趣点具有相应的兴趣点标识信息；所述信息生成模块具体用于：
针对每个兴趣点，基于所述兴趣点关联的第二WiFi信息以及所述每组中的第一WiFi信息，生成与所述兴趣点对应的扫描分布标识信息；
基于所述兴趣点的兴趣点标识信息以及与所述兴趣点对应的扫描分布标识信息，生成所述与所述兴趣点对应的共现信息。
在一种可能的实施方式中，所述第一WiFi信息包括第一WiFi以及所述第一WiFi的信号强度，所述第二WiFi信息包括第二WiFi以及所述第二WiFi的信号强度，每个兴趣点关联的第二WiFi之间存在分布顺序，所述分布顺序由所述各个第二WiFi的信号强度确定；所述信息生成模块具体用于：
针对每个兴趣点关联的每个第二WiFi，基于所述第二WiFi以及所述每组中的每个第一WiFi，生成子扫描分布标识信息；
将各个子扫描分布标识信息按照所述兴趣点关联的第二WiFi之间的分布顺序进行拼接，生成所述扫描分布标识信息。
在一种可能的实施方式中，所述子扫描分布标识信息对应一个扫描标志位以及多个信号强度标志位，不同信号强度标志位用于表征不同信号强度；所述信息生成模块具体用于：
针对每个第二WiFi，在存在与所述第二WiFi的媒体访问控制地址MAC地址相同的目标第一WiFi的情况下，确定所述扫描标志位的值为1；
从信号强度标志位中确定所述目标第一WiFi的信号强度对应的目标信号强度标志位，并将所述目标信号强度标志位的值确定为1，将除所述目标信号强度标志位之外的其他信号强度标志位的值确定为0；
基于所述扫描标志位的值以及所述信号强度标志位的值，生成所述子扫描分布标识信息。
在一种可能的实施方式中，所述共现信息的格式为字符串格式；所述样本生成模块具体用于：
针对每个共现信息，对所述字符串格式的共现信息进行特征处理，生成与所述共现信息对应的特征向量；
获取与所述共现信息对应的兴趣点所关联的预设线下行为信息，并基于所述预设线下行为信息确定所述标签对应的真值；
基于所述特征向量以及所述标签对应的真值，生成所述训练样本数据。
在一种可能的实施方式中，所述样本生成模块具体用于：
在发生预设线下行为的情况下，将所述标签对应的真值确定为1；或者，
在未发生预设线下行为的情况下，将所述标签对应的真值确定为0。
在一种可能的实施方式中，所述样本生成模块具体用于：
对字符串格式的各个共现信息按照预设编号方式分别进行编号，并基于所述各个共现信息分别对应的编号，构建特征字典；所述特征字典用于将所述字符串格式的共现信息转换成特征向量；
针对每个字符串格式的共现信息，基于所述特征字典对所述字符串格式的共现信息进行独热编码处理，生成与所述共现信息对应的特征向量。
在一种可能的实施方式中，所述样本生成模块具体用于：
确定所述特征字典中包含的编号的数量，并创建向量长度为所述数量的零向量；
针对每个共现信息，根据所述共现信息在所述特征字典中对应的编号，将所述零向量中与所述编号相同的目标索引位的值修改为1，生成与所述共现信息对应的特征向量。
在一种可能的实施方式中，所述模型训练模块具体用于：
将每个训练样本数据输入所述待训练的模型，得到与所述每个训练样本对应的预测结果；
基于所述每个训练样本对应的预测结果以及所述每个训练样本数据的标签，对所述待训练的模型的模型参数进行调整；
重复上述过程直到训练结果符合预设要求，得到所述训练好的模型。
本公开实施例提供了一种兴趣点定位装置，包括：
获取模块，用于获取当前设备的当前WiFi扫描信息以及所述当前设备所处区域对应的每个兴趣点对应的关联WiFi信息；
生成模块，用于针对每个兴趣点，基于所述当前WiFi扫描信息以及所述兴趣点对应的关联WiFi信息，生成当前共现信息；
定位模块，用于将各个当前共现信息输入至训练好的模型，得到所述当前设备的兴趣点定位信息；其中，所述训练好的模型通过权利要求1-9中任一项所述的模型训练方法得到。
本公开实施例还提供一种电子设备，包括：处理器、存储器和总线，所述存储器存储有所述处理器可执行的机器可读指令，当电子设备运行时，所述处理器与所述存储器之间通过总线通信，所述机器可读指令被所述处理器执行时执行上述任一种可能的实施方式中所述的模型训练方法的步骤或者上述兴趣点定位方法的步骤。
本公开实施例还提供一种计算机可读存储介质，该计算机可读存储介质上存储有计算机程序，该计算机程序被处理器运行时执行上述任一种可能的实施方式中所述的模型训练方法的步骤或者上述兴趣点定位方法的步骤。
附图说明
为了更清楚地说明本公开实施例的技术方案，下面将对实施例中所需要使用的附图作简单地介绍，此处的附图被并入说明书中并构成本说明书中的一部分，这些附图示出了符合本公开的实施例，并与说明书一起用于说明本公开的技术方案。应当理解，以下附图仅示出了本公开的某些实施例，因此不应被看作是对范围的限定，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他相关的附图。
图1示出了本公开实施例所提供的一种模型训练方法的流程图；
图2示出了本公开实施例所提供的一种共现信息的生成方法的流程图；
图3示出了本公开实施例所提供的一种子扫描分布标识信息的生成过程的示意图；
图4示出了本公开实施例所提供的一种训练数据样本的生成方法的流程图；
图5示出了本公开实施例所提供的一种共现信息的特征处理方法的流程图；
图6示出了本公开实施例所提供的一种兴趣点定位方法的流程图；
图7示出了本公开实施例所提供的一种模型训练装置的示意图；
图8示出了本公开实施例所提供的一种兴趣点定位装置的示意图；以及
图9示出了本公开实施例所提供的一种电子设备的示意图。
具体实施方式
为使本公开实施例的目的、技术方案和优点更加清楚，下面将结合本公开实施例中附图，对本公开实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本公开一部分实施例，而不是全部的实施例。通常在此处附图中描述和示出的本公开实施例的组件可以以各种不同的配置来布置和设计。因此，以下对在附图中提供的本公开的实施例的详细描述并非旨在限制要求保护的本公开的范围，而是仅仅表示本公开的选定实施例。基于本公开的实施例，本领域技术人员在没有做出创造性劳动的前提下所获得的所有其他实施例，都属于本公开保护的范围。
应注意到：相似的标号和字母在下面的附图中表示类似项，因此，一旦某一项在一个附图中被定义，则在随后的附图中不需要对其进行进一步定义和解释。
本文中术语“和/或”，仅仅是描述一种关联关系，表示可以存在三种关系，例如，A和/或B，可以表示：单独存在A，同时存在A和B，单独存在B这三种情况。另外，本文中术语“至少一种”表示多种中的任意一种或多种中的至少两种的任意组合，例如，包括A、B、C中的至少一种，可以表示包括从A、B和C构成的集合中选择的任意一个或多个元素。
可以理解的是，在使用本公开各实施例公开的技术方案之前，均应当依据相关法律法规通过恰当的方式对本公开所涉及个人信息的类型、使用范围、使用场景等告知用户并获得用户的授权。
经研究发现，常见的指纹定位技术中，通常是基于用户通过电子设备扫描到的WiFi信息与历史采集到的WiFi指纹信息之间的交叉信息，来确定用户当前所在的兴趣点，而交叉信息计算的准确程度会直接影响最终定位结果的可信度，因此，如何提升交叉信息的准确程度，进而提升定位结果的可信度，是当前亟待解决的问题。
在个性化信息推荐、信息扫描业务以及在线广告领域中，点击率(Click-through Rate，CTR)预估模型是重要技术之一，用来学习和预测用户的反馈信息，其中，用户的反馈信息可以是用户进行的点击、收藏或者购买等行为信息。CTR模型通过其具有的记忆能力实现上述信息推荐等功能，其中，记忆能力是指模型直接学习并利用历史数据中请求与历史指纹的“共现频率”的能力。一般来说，协同过滤模型以及逻辑回归模型等模型均具有较强的“记 忆能力”，由于这类模型的结构简单，历史数据往往可以直接影响推荐结果，也即，模型能够学习到历史数据的分布特点，并利用自身的记忆性进行结果的预测。
基于上述研究，本公开提供了一种模型训练方法、装置、电子设备以及存储介质，首先获取历史无线保真WiFi扫描信息以及历史WiFi分布信息；所述历史WiFi扫描信息包括多组WiFi扫描信息，每组WiFi扫描信息包括至少一个第一WiFi信息，所述历史WiFi分布信息包括每个兴趣点关联的第二WiFi信息；其次针对每个兴趣点，基于所述兴趣点关联的第二WiFi信息以及每组WiFi扫描信息中的至少一个第一WiFi信息，生成与所述兴趣点对应的共现信息；所述共现信息用于表征每组中的每个第一WiFi信息相对于所述兴趣点关联的第二WiFi信息的扫描状态；然后针对每个共现信息，基于所述共现信息以及与所述共现信息对应的标签，生成训练样本数据；所述共现信息对应的标签基于与所述共现信息对应的兴趣点所关联的预设线下行为信息确定；最后基于各个训练样本数据，对待训练的模型进行训练，得到训练好的模型；所述训练好的模型用于基于当前WiFi扫描信息确定对应的兴趣点定位信息。
本公开实施例中，基于历史WiFi分布信息以及历史WiFi扫描信息构建了对应的共现信息，由于共现信息表征了每个第一用户通过电子设备扫描到的WiFi信息相对于覆盖所述兴趣点的至少一个WiFi信息的扫描状态，因此，基于共现信息生成训练样本数据对模型进行训练，可以增强模型对于WiFi扫描信息与兴趣点之间的关联关系的记忆，并得到训练好的模型，进一步的，可以基于训练好的模型进行兴趣点预测，进而提升定位结果的可信度。
为便于对本实施例进行理解，首先对本公开实施例所提供的模型训练方法的执行主体进行详细介绍。本公开实施例所提供的模型训练方法的执行主体为电子设备。本实施方式中，所述电子设备为服务器，该服务器可以是独立的物理服务器，也可以是多个物理服务器构成的服务器集群或者分布式系统，还可以是提供云服务、云数据库、云计算、云存储、大数据和人工智能平台等基础云计算服务的云服务器。其他实施例中，该电子设备还可以为终端设备。其中，该终端设备可以为移动设备、用户终端、手持设备、计算设备及可穿戴设备等。此外，该模型训练方法还可以通过处理器调用存储器中 存储的计算机可读指令的方式来实现。
下面结合附图对本申请实施例所提供的模型训练方法进行详细说明。参见图1所示，为本公开实施例提供的模型训练方法的流程图，所述方法包括步骤S101～S104，其中：
S101，获取历史无线保真WiFi扫描信息以及历史WiFi分布信息；所述历史WiFi扫描信息包括多组WiFi扫描信息，每组WiFi扫描信息包括至少一个第一WiFi信息，所述历史WiFi分布信息包括每个兴趣点关联的第二WiFi信息。
其中，所述历史WiFi扫描信息可以是通过电子设备扫描到的WiFi信息，也即，可以是指通过不同电子设备扫描到的多组WiFi扫描信息，且由于WiFi信号具有相应的覆盖范围，因此，每组WiFi扫描信息包括至少一个第一WiFi信息。所述第一WiFi信息包括第一WiFi以及所述第一WiFi的信号强度，且每个第一WiFi对应一个兴趣点。
所述电子设备可以是终端设备，例如智能手机、平板电脑或者智能手表等设备，在此不做限定。应当理解，历史WiFi信息均为在历史时间(例如前一个月)内所采集到的WiFi信息。
所述历史WiFi分布信息包括每个兴趣点关联的第二WiFi信息，所述第二WiFi信息包括第二WiFi以及所述第二WiFi的信号强度，每个兴趣点关联的第二WiFi之间存在分布顺序，所述分布顺序由所述各个第二WiFi的信号强度确定。
其中，所述历史WiFi指纹分布信息可以是基于历史WiFi扫描信息构建出的在兴趣点(POI)维度下的WiFi分布信息，也即，历史WiFi指纹分布信息可以是根据针对各个兴趣点，在历史时间内扫描到的多组WiFi扫描信息所构建出的WiFi分布，并且，由于各个第二WiFi的信号强度大小不同，因此，针对每个兴趣点，与该兴趣点关联的第二WiFi之间存在相应的分布顺序。
示例性地，针对兴趣点A，其对应的第二WiFi可以包括W1、W2、W3，其中，W1、W2、W3相对于兴趣点A的距离不同，导致与该兴趣点A关联的各个第二WiFi的信号强度也是不同的，例如，W1、W2、W3之间的信号强度为W1>W2>W3，则各个第二WiFi之间的分布顺序可以是W1、W2、 W3。
本公开实施例中，所述兴趣点可以是指目标场所中的一个店铺，目标场所可以是商场、写字楼等，示例性的，兴趣点可以是商场中的一间店铺。其中，每个兴趣点具有相应的唯一的兴趣点标识信息，所述兴趣点标识信息可以是由数字、字母等构成的标识信息。
S102，针对每个兴趣点，基于所述兴趣点关联的第二WiFi信息以及每组WiFi扫描信息中的至少一个第一WiFi信息，生成与所述兴趣点对应的共现信息；所述共现信息用于表征每组中的每个第一WiFi信息相对于所述兴趣点关联的第二WiFi信息的扫描状态。
其中，所述扫描状态可以是指扫描到状态和未扫描到状态。
这里，可以针对每个兴趣点，将该兴趣点关联的第二WiFi信息与每组WiFi扫描信息中的至少一个第一WiFi信息进行比对，如此，可以确定所述至少一个第一WiFi信息中是否存在与该兴趣点关联的第二WiFi信息相同的第一WiFi信息，进而可以确定每个第一WiFi信息相对于所述兴趣点关联的第二WiFi信息的扫描状态。
可选的，针对步骤S102，在针对每个兴趣点，基于所述兴趣点关联的第二WiFi信息以及每组WiFi扫描信息中的至少一个第一WiFi信息，生成与所述兴趣点对应的共现信息时，请参见图2，可以包括以下步骤S1021～S1022：
S1021，针对每个兴趣点，基于所述兴趣点关联的第二WiFi信息以及所述每组中的第一WiFi信息，生成与所述兴趣点对应的扫描分布标识信息。
该步骤中，可以针对每个兴趣点，将该兴趣点关联的第二WiFi信息以及每组WiFi扫描信息中第一WiFi信息进行比对，从而生成与该兴趣点对应的扫描分布标识信息。
具体的，在针对每个兴趣点，基于所述兴趣点关联的第二WiFi信息以及所述每组中的第一WiFi信息，生成与所述兴趣点对应的扫描分布标识信息时，可以包括以下步骤(1)-(2)：
(1)针对每个兴趣点关联的每个第二WiFi，基于所述第二WiFi以及所述每组中的每个第一WiFi，生成子扫描分布标识信息。
其中，所述子扫描分布标识信息对应一个扫描标志位以及多个信号强度标志位，所述扫描标志位用于表征是否扫描到对应的第二WiFi。所述信号强 度标志位用于表征信号强度，其中，不同信号强度标志位对应不同的信号强度，本实施方式中，所述信号强度标志位包括8位，例如，-70dBm～-60dBm对应信号强度标志位0，-60dBm～-50dBm对应信号强度标志位1。
具体的，可以针对每个第二WiFi，根据每组第一WiFi信息中的第一WiFi是否与所述第二WiFi相同确定扫描标志位，以及基于所述第一WiFi信息中的每个第一WiFi的信号强度确定信号强度标志位，并基于扫描标志位和信号强度标志位，生成子扫描分布标识信息。具体的，可以包括以下步骤(a)～(c)：
(a)针对每个第二WiFi，在存在与所述第二WiFi的媒体访问控制地址MAC地址相同的目标第一WiFi的情况下，确定所述扫描标志位的值为1。
(b)从信号强度标志位中确定所述目标第一WiFi的信号强度对应的目标信号强度标志位，并将所述目标信号强度标志位的值确定为1，将除所述目标信号强度标志位之外的其他信号强度标志位的值确定为0。
(c)基于所述扫描标志位的值以及所述信号强度标志位的值，生成所述子扫描分布标识信息。
本实施方式中，在根据所述扫描标志位的值以及所述信号强度标志位的值，确定所述子扫描分布标识信息时，可以根据子扫描分布标识信息的生成公式进行确定，如公式(1)所示：
IDn＝Tagscan*(28+2I)         (1)
其中，IDn为子扫描分布标识信息，扫描标志位Tagscan用于指示第一WiFi信息中存在与所述第二WiFi的媒体访问控制地址MAC地址相同的目标第一WiFi，若存在，确定扫描标志位的值为1，若不存在，确定扫描标志位的值为0；I表示目标第一WiFi的信号强度落入信号分桶的索引值，所述信号分桶是指信号范围。
示例性的，请参见图3，为本公开实施例所提供的一种子扫描分布标识信息的生成过程的示意图。如图3中所示，兴趣点POI-1分布有5个第二WiFi(W1、W2、W3、W4、W5)，针对每个第二WiFi(以W1为例)，将W1与每组第一WiFi信息中的每个第一WiFi分别进行比对，若存在与W1的MAC地址相同的目标第一WiFi，则确定扫描标志位的值为1，然后，根据目标第一WiFi的信号强度的大小，判断其信号强度对应的目标信号强度标志 位，并将该目标信号强度标志位的值赋值为1，并将其他信号强度标志位的值赋值为0，如此，得到一个长度为9的二进制码100010000，进而将该二进制码转换为十进制整数，得到所述子扫描分布标识信息272。
类似的，可以针对W2可以确定对应的子扫描分布标识信息位000、W3对应的子扫描分布标识信息为272，W4对应的子扫描分布标识信息为258，W5对应的子扫描分布标识信息为288。
(2)将各个子扫描分布标识信息按照所述兴趣点关联的第二WiFi之间的分布顺序进行拼接，生成所述扫描分布标识信息。
这里，在得到各个子扫描分布标识信息后，即可按照各个第二WiFi的分布顺序进行拼接，如图3所示，得到的子扫描分布标识信息分别为272、000、272、258以及288，将其拼接得到扫描分布标识信息272000272258288。
S1022，基于所述兴趣点的兴趣点标识信息以及与所述兴趣点对应的扫描分布标识信息，生成所述与所述兴趣点对应的共现信息。
可以理解，在得到扫描分布标识信息后，即可基于所述兴趣点标识信息以及与所述兴趣点对应的扫描分布标识信息，生成与所述兴趣点对应的共现信息，具体的，可以将兴趣点标识信息以及与兴趣点对应的扫描分布标识信息进行拼接，生成共现信息，这里，共现信息的格式为字符串格式。
示例性的，兴趣点标识信息为22535659086281011，兴趣点对应的扫描分布标识信息为272000272258288，则共现信息即为：272000272258288_22535659086281011。
需要说明的是，上述示例是针对一个兴趣点以及一组第一WiFi扫描信息为例进行说明的，因此，在实际生成共现信息的过程中，针对每个兴趣点，可以基于多组第一WiFi扫描信息确定多个共现信息。
S103，针对每个共现信息，基于所述共现信息以及与所述共现信息对应的标签，生成训练样本数据；所述共现信息对应的标签基于与所述共现信息对应的兴趣点所关联的预设线下行为信息确定。
其中，所述预设线下行为信息可以是指是否发生预设线下行为，示例的，所述预设线下行为可以是指线下消费行为，例如优惠券核销等行为。
需要说明的是，在生成训练样本数据时，针对各个第一WiFi，若在任一个第一WiFi对应的兴趣点中均未发生预设线下行为，则对应的共现信息将 不被用于生成训练样本数据，例如，第一WiFi扫描信息中的第一WiFi包括W11、W12、…、W19，其分别对应兴趣点POI1、POI2、…、POI9，在兴趣点POI0、POI2、…、POI9均未发生线下行为，则对应的共现信息不会被用于生成训练样本数据。
可选的，针对步骤S103，在针对每个共现信息，基于所述共现信息以及与所述共现信息对应的标签，生成训练样本数据时，请参见图4，可以包括以下S1031～S1033：
S1031，针对每个共现信息，对所述字符串格式的共现信息进行特征处理，生成与所述共现信息对应的特征向量。
应当理解，由于共现信息为字符串格式，而为了便于后续步骤的模型训练，则需要将其转换成对应的特征向量。
可选的，在针对每个共现信息，对所述字符串格式的共现信息进行特征处理，生成与所述共现信息对应的特征向量时，请参见图5，包括以下S10311～S10312：
S10311，对字符串格式的各个共现信息按照预设编号方式分别进行编号，并基于所述各个共现信息分别对应的编号，构建特征字典；所述特征字典用于将所述字符串格式的共现信息转换成特征向量。
具体的，针对各个共现信息可以进行统一编号，使得每个共现信息唯一对应一个整数。示例性的，请参见表1，示出了共现信息与编号之间的对应关系。
表1
需要说明的是，在生成特征字典时，对于一些低频出现(例如出现次数 少于5次)的共现信息，将不会对其进行编号。
S10312，针对每个字符串格式的共现信息，基于所述特征字典对所述字符串格式的共现信息进行独热编码处理，生成与所述共现信息对应的特征向量。
其中，所述特征向量为高维度的稀疏特征向量。
在生成上述特征字典后，即可对字符串格式的共现信息进行编码处理，本公开实施例中，采用独热编码方式的方式，其中，独热编码又称为One-hot编码，其方法是使用N位状态寄存器来对N个状态进行编码，每个状态都有独立的寄存器位，并且在任意时刻只有一位寄存器位有效。
可选的，在基于所述特征字典对所述字符串格式的共现信息进行独热编码处理，生成与所述共现信息对应的特征向量时，可以首先确定所述特征字典中包含的编号的数量，并创建向量长度为所述数量的零向量，然后针对每个共现信息，根据所述共现信息在所述特征字典中对应的编号，将所述零向量中与所述编号相同的目标索引位的值修改为1，生成与所述共现信息对应的特征向量。
示例性的，请参见表2，示出了基于特征字典对共现信息进行特征转换所得到的特征向量的转换过程。
表2
基于表2可知，若特征字典中的编号数量为9个，则创建向量长度为9的零向量，针对每个共现信息(例如272288272258288_22535659086281035)，在特征字典中确定该共现信息对应的编号为2，则将零向量中索引位为2的寄存器位的值修改为1，生成与共现信息对应的特征向量001000000。
S1032，获取与所述共现信息对应的兴趣点所关联的预设线下行为信息，并基于所述预设线下行为信息确定所述标签对应的真值。
其中，在发生预设线下行为的情况下，将所述标签对应的真值确定为1，
在未发生预设线下行为的情况下，将所述标签对应的真值确定为0。
S1033，基于所述特征向量以及所述标签对应的真值，生成所述训练样本数据。
如此，在确定特征向量后，即可将前述实施例中所确定的与每个特征向量的真值进行标定，得到训练样本数据。
S104，基于各个训练样本数据，对待训练的模型进行训练，得到训练好的模型；所述训练好的模型用于基于当前WiFi扫描信息确定对应的兴趣点定位信息。
其中，所述待训练的模型可以是指逻辑回归模型，在其他实施方式中，所述待训练的模型还可以是其他模型，在此不做限定。
可以理解，在得到训练样本数据后，即可基于训练样本数据对待训练的模型进行训练(或者基于待训练的逻辑回归模型对所述训练样本数据进行拟合)，得到训练好的模型，这里，由于每个训练样本数据具有相应的标签，因此可以对待训练的模型进行有监督训练。
本公开实施例中，基于历史WiFi分布信息以及历史WiFi扫描信息构建了对应的共现信息，由于共现信息表征了扫描到的每组第一WiFi扫描信息相对于与所述兴趣点关联的至少一个第二WiFi信息的扫描状态，因此，基于共现信息生成训练样本数据对模型进行训练，可以增强模型对于WiFi扫描信息与兴趣点之间的关联关系的记忆，并得到训练好的模型，进一步的，可以基于训练好的模型进行兴趣点预测，进而提升定位结果的可信度。
可选的，在基于训练样本数据对待训练的模型进行训练，得到训练好的模型时，可以将各个训练样本数据输入到待训练的模型中，得到与每个训练样本样本对应的预测结果，然后，基于所述每个训练样本数据对应的预测结果以及所述每个训练样本数据的标签，对该模型的模型参数进行调整，具体的，可以预先设置损失函数，并计算每个训练样本数据对应的预测结果与每个训练样本数据的标签之间的损失值，基于损失值对模型的模型参数进行调整，如此，重复上述过程直到训练结果符合预设要求，得到所述训练好的模型。
请参见图6，为本公开实施例所提供的一种兴趣点定位方法，所述方法 包括S601～S603：
S601，获取当前设备的当前WiFi扫描信息以及所述当前设备所处区域对应的每个兴趣点对应的关联WiFi信息。
示例性的，可以通过当前电子设备获取当前WiFi扫描信息。所述当前WiFi扫描信息包括当前扫描到的至少一个WiFi以及每个WiFi的信号强度。
这里，可以根据当前设备的位置信息确定当前设备所处的区域，所述每个兴趣点对应的关联WiFi信息可以是预先构建好的。
S602，针对每个兴趣点，基于所述当前WiFi扫描信息以及所述兴趣点对应的关联WiFi信息，生成当前共现信息。
该步骤中，通过将当前WiFi扫描信息分别与每个兴趣点的兴趣点标识信息进行组合，并对组合后的信息进行独热编码(One-hot)，得到所述当前共现信息，其中，当前共现信息为向量形式。
S603，将各个当前共现信息输入至训练好的模型，得到所述当前设备的兴趣点定位信息。
其中，所述训练好的模型通过上述任一项所述的模型训练方法得到。
可以理解，在将各个当前共现信息输入至训练好的模型后，即可输出当前设备所处区域对应的各个兴趣点对应的概率，因此，在进行POI定位的过程中，只需要将上述得到的向量形式的当前共现信息输入到训练好的模型，模型即能够输出各个兴趣点对应的概率，也即，从而根据各个兴趣点对应的概率，便可以确定当前设备的兴趣点定位信息，具体的，可以将最大概率所对应的兴趣点确定为当前设备所在兴趣点。
在一些实施方式，在确定当前设备所位于的兴趣点之后，即可根据其所在的兴趣点，想使用当前设备的用户进行信息推荐(例如优惠券推荐、商品推荐、线上广告投送等)。
本领域技术人员可以理解，在具体实施方式的上述方法中，各步骤的撰写顺序并不意味着严格的执行顺序而对实施过程构成任何限定，各步骤的具体执行顺序应当以其功能和可能的内在逻辑确定。
基于同一发明构思，本公开实施例中还提供了与模型训练方法对应的模型训练装置，由于本公开实施例中的装置解决问题的原理与本公开实施例上述模型训练方法相似，因此装置的实施可以参见方法的实施，重复之处不再 赘述。
参照图7所示，为本公开实施例提供的一种模型训练装置700的示意图，所述装置包括：
信息获取模块701，用于获取历史无线保真WiFi扫描信息以及历史WiFi分布信息；所述历史WiFi扫描信息包括多组WiFi扫描信息，每组WiFi扫描信息包括至少一个第一WiFi信息，所述历史WiFi分布信息包括每个兴趣点关联的第二WiFi信息；
信息生成模块702，用于针对每个兴趣点，基于所述兴趣点关联的第二WiFi信息以及每组WiFi扫描信息中的至少一个第一WiFi信息，生成与所述兴趣点对应的共现信息；所述共现信息用于表征每组中的每个第一WiFi信息相对于所述兴趣点关联的第二WiFi信息的扫描状态；
样本生成模块703，用于针对每个共现信息，基于所述共现信息以及与所述共现信息对应的标签，生成训练样本数据；所述共现信息对应的标签基于与所述共现信息对应的兴趣点所关联的预设线下行为信息确定；
模型训练模块704，用于基于各个训练样本数据，对待训练的模型进行训练，得到训练好的模型；所述训练好的模型用于基于当前WiFi扫描信息确定对应的兴趣点定位信息。
在一种可能的实施方式中，每个兴趣点具有相应的兴趣点标识信息；；所述信息生成模块702具体用于：
针对每个兴趣点，基于所述兴趣点关联的第二WiFi信息以及所述每组中的第一WiFi信息，生成与所述兴趣点对应的扫描分布标识信息；
基于所述兴趣点的兴趣点标识信息以及与所述兴趣点对应的扫描分布标识信息，生成所述与所述兴趣点对应的共现信息。
在一种可能的实施方式中，所述第一WiFi信息包括第一WiFi以及所述第一WiFi的信号强度，所述第二WiFi信息包括第二WiFi以及所述第二WiFi的信号强度，每个兴趣点关联的第二WiFi之间存在分布顺序，所述分布顺序由所述各个第二WiFi的信号强度确定；所述信息生成模块702具体用于：
针对每个兴趣点关联的每个第二WiFi，基于所述第二WiFi以及所述每组中的每个第一WiFi，生成子扫描分布标识信息；
将各个子扫描分布标识信息按照所述兴趣点关联的第二WiFi之间的分 布顺序进行拼接，生成所述扫描分布标识信息。
在一种可能的实施方式中，所述子扫描分布标识信息对应一个扫描标志位以及多个信号强度标志位，不同信号强度标志位用于表征不同信号强度；所述信息生成模块702具体用于：
针对每个第二WiFi，在存在与所述第二WiFi的媒体访问控制地址MAC地址相同的目标第一WiFi的情况下，确定所述扫描标志位的值为1；
从信号强度标志位中确定所述目标第一WiFi的信号强度对应的目标信号强度标志位，并将所述目标信号强度标志位的值确定为1，将除所述目标信号强度标志位之外的其他信号强度标志位的值确定为0；
基于所述扫描标志位的值以及所述信号强度标志位的值，生成所述子扫描分布标识信息。
在一种可能的实施方式中，所述共现信息的格式为字符串格式；所述样本生成模块703具体用于：
针对每个共现信息，对所述字符串格式的共现信息进行特征处理，生成与所述共现信息对应的特征向量；
获取与所述共现信息对应的兴趣点所关联的预设线下行为信息，并基于所述预设线下行为信息确定所述标签对应的真值；
基于所述特征向量以及所述标签对应的真值，生成所述训练样本数据。
在一种可能的实施方式中，所述样本生成模块703具体用于：
在发生预设线下行为的情况下，将所述标签对应的真值确定为1；或者，
在未发生预设线下行为的情况下，将所述标签对应的真值确定为0。
在一种可能的实施方式中，所述样本生成模块703具体用于：
对字符串格式的各个共现信息按照预设编号方式分别进行编号，并基于所述各个共现信息分别对应的编号，构建特征字典；所述特征字典用于将所述字符串格式的共现信息转换成特征向量；
针对每个字符串格式的共现信息，基于所述特征字典对所述字符串格式的共现信息进行独热编码处理，生成与所述共现信息对应的特征向量。
在一种可能的实施方式中，所述样本生成模块703具体用于：
确定所述特征字典中包含的编号的数量，并创建向量长度为所述数量的零向量；
针对每个共现信息，根据所述共现信息在所述特征字典中对应的编号，将所述零向量中与所述编号相同的目标索引位的值修改为1，生成与所述共现信息对应的特征向量。
在一种可能的实施方式中，所述模型训练模块704具体用于：
将每个训练样本数据输入所述待训练的模型，得到与所述每个训练样本对应的预测结果；
基于所述每个训练样本对应的预测结果以及所述每个训练样本数据的标签，对所述待训练的模型的模型参数进行调整；
重复上述过程直到训练结果符合预设要求，得到所述训练好的模型。
请参见图8，为本公开实施例提供的一种兴趣点定位装置，所述兴趣点定位装置包括：
获取模块801，用于获取当前设备的当前WiFi扫描信息以及所述当前设备所处区域对应的每个兴趣点对应的关联WiFi信息；
生成模块802，用于针对每个兴趣点，基于所述当前WiFi扫描信息以及所述兴趣点对应的关联WiFi信息，生成当前共现信息；
定位模块803，用于将各个当前共现信息输入至训练好的模型，得到所述当前设备的兴趣点定位信息；其中，所述训练好的模型通过权利要求1-9中任一项所述的模型训练方法得到。
关于装置中的各模块的处理流程、以及各模块之间的交互流程的描述可以参照上述方法实施例中的相关说明，这里不再详述。
基于同一技术构思，本公开实施例还提供了一种电子设备。参照图9所示，为本公开实施例提供的电子设备900的结构示意图，包括处理器901、存储器902、和总线903。其中，存储器902用于存储执行指令，包括内存9021和外部存储器9022；这里的内存9021也称内存储器，用于暂时存放处理器901中的运算数据，以及与硬盘等外部存储器9022交换的数据，处理器901通过内存9021与外部存储器9022进行数据交换。
本申请实施例中，存储器902具体用于存储执行本申请方案的应用程序代码，并由处理器901来控制执行。也即，当电子设备900运行时，处理器901与存储器902之间通过总线903通信，使得处理器901执行存储器902中存储的应用程序代码，进而执行前述任一实施例中所述的方法。
其中，存储器902可以是，但不限于，随机存取存储器(Random Access Memory，RAM)，只读存储器(Read Only Memory，ROM)，可编程只读存储器(Programmable Read－Only Memory，PROM)，可擦除只读存储器(Erasable Programmable Read－Only Memory，EPROM)，电可擦除只读存储器(Electric Erasable Programmable Read－Only Memory，EEPROM)等。
处理器901可能是一种集成电路芯片，具有信号的处理能力。上述的处理器可以是通用处理器，包括中央处理器(Central Processing Unit，CPU)、网络处理器(Network Processor，NP)等；还可以是数字信号处理器(DSP)、专用集成电路(ASIC)、现场可编程门阵列(FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。可以实现或者执行本发明实施例中的公开的各方法、步骤及逻辑框图。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。
可以理解的是，本申请实施例示意的结构并不构成对电子设备900的具体限定。在本申请另一些实施例中，电子设备900可以包括比图示更多或更少的部件，或者组合某些部件，或者拆分某些部件，或者不同的部件布置。图示的部件可以以硬件，软件或软件和硬件的组合实现。
本公开实施例还提供一种计算机可读存储介质，该计算机可读存储介质上存储有计算机程序，该计算机程序被处理器运行时执行上述方法实施例中的模型训练的步骤。其中，该存储介质可以是易失性或非易失的计算机可读取存储介质。
本公开实施例还提供一种计算机程序产品，该计算机程序产品承载有程序代码，所述程序代码包括的指令可用于执行上述方法实施例中的模型训练的步骤，具体可参见上述方法实施例，在此不再赘述。
其中，上述计算机程序产品可以具体通过硬件、软件或其结合的方式实现。在一个可选实施例中，所述计算机程序产品具体体现为计算机存储介质，在另一个可选实施例中，计算机程序产品具体体现为软件产品，例如软件开发包(Software Development Kit，SDK)等等。
所属领域的技术人员可以清楚地了解到，为描述的方便和简洁，上述描述的系统和装置的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘述。在本公开所提供的几个实施例中，应该理解到，所揭露的系 统、装置和方法，可以通过其它的方式实现。以上所描述的装置实施例仅仅是示意性的，例如，所述单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，又例如，多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些通信接口，装置或单元的间接耦合或通信连接，可以是电性，机械或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本公开各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。
所述功能如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个处理器可执行的非易失的计算机可读取存储介质中。基于这样的理解，本公开的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以使得一台电子设备(可以是个人计算机，服务器，或者网络设备等)执行本公开各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器、随机存取存储器、磁碟或者光盘等各种可以存储程序代码的介质。
最后应说明的是：以上所述实施例，仅为本公开的具体实施方式，用以说明本公开的技术方案，而非对其限制，本公开的保护范围并不局限于此，尽管参照前述实施例对本公开进行了详细的说明，本领域的普通技术人员应当理解：任何熟悉本技术领域的技术人员在本公开揭露的技术范围内，其依然可以对前述实施例所记载的技术方案进行修改或可轻易想到变化，或者对其中部分技术特征进行等同替换；而这些修改、变化或者替换，并不使相应技术方案的本质脱离本公开实施例技术方案的精神和范围，都应涵盖在本公开的保护范围之内。因此，本公开的保护范围应所述以权利要求的保护范围为准。

Claims (14)

  1. 一种模型训练方法，包括：
    获取历史无线保真WiFi扫描信息以及历史WiFi分布信息；所述历史WiFi扫描信息包括多组WiFi扫描信息，每组WiFi扫描信息包括至少一个第一WiFi信息，所述历史WiFi分布信息包括每个兴趣点关联的第二WiFi信息；
    针对每个兴趣点，基于所述兴趣点关联的第二WiFi信息以及每组WiFi扫描信息中的至少一个第一WiFi信息，生成与所述兴趣点对应的共现信息；所述共现信息用于表征每组中的每个第一WiFi信息相对于所述兴趣点关联的第二WiFi信息的扫描状态；
    针对每个共现信息，基于所述共现信息以及与所述共现信息对应的标签，生成训练样本数据；所述共现信息对应的标签基于与所述共现信息对应的兴趣点所关联的预设线下行为信息确定；
    基于各个训练样本数据，对待训练的模型进行训练，得到训练好的模型；所述训练好的模型用于基于当前WiFi扫描信息确定对应的兴趣点定位信息。
  2. 根据权利要求1所述的方法，其中，每个兴趣点具有相应的兴趣点标识信息；所述基于所述兴趣点关联的第二WiFi信息以及每组WiFi扫描信息中的至少一个第一WiFi信息，生成与所述兴趣点对应的共现信息，包括：
    针对每个兴趣点，基于所述兴趣点关联的第二WiFi信息以及所述每组中的第一WiFi信息，生成与所述兴趣点对应的扫描分布标识信息；
    基于所述兴趣点的兴趣点标识信息以及与所述兴趣点对应的扫描分布标识信息，生成所述与所述兴趣点对应的共现信息。
  3. 根据权利要求2所述的方法，其中，所述第一WiFi信息包括第一WiFi以及所述第一WiFi的信号强度，所述第二WiFi信息包括第二WiFi以及所述第二WiFi的信号强度，每个兴趣点关联的第二WiFi之间存在分布顺序，所述分布顺序由所述各个第二WiFi的信号强度确定；
    所述针对每个兴趣点，基于所述兴趣点关联的第二WiFi信息以及所述每组中的第一WiFi信息，生成与所述兴趣点对应的扫描分布标识信息，包括：
    针对每个兴趣点关联的每个第二WiFi，基于所述第二WiFi以及所述每组中的每个第一WiFi，生成子扫描分布标识信息；
    将各个子扫描分布标识信息按照所述兴趣点关联的第二WiFi之间的分布顺序进行拼接，生成所述扫描分布标识信息。
  4. 根据权利要求3所述的方法，其中，所述子扫描分布标识信息对应一个扫描标志位以及多个信号强度标志位，不同信号强度标志位用于表征不同信号强度；所述针对每个兴趣点关联的每个第二WiFi，基于所述第二WiFi以及所述每组中的每个第一WiFi，生成子扫描分布标识信息，包括：
    针对每个第二WiFi，在存在与所述第二WiFi的媒体访问控制地址MAC地址相同的目标第一WiFi的情况下，确定所述扫描标志位的值为1；
    从信号强度标志位中确定所述目标第一WiFi的信号强度对应的目标信号强度标志位，并将所述目标信号强度标志位的值确定为1，将除所述目标信号强度标志位之外的其他信号强度标志位的值确定为0；
    基于所述扫描标志位的值以及所述信号强度标志位的值，生成所述子扫描分布标识信息。
  5. 根据权利要求1所述的方法，其中，所述共现信息的格式为字符串格式；所述针对每个共现信息，基于所述共现信息以及与所述共现信息对应的标签，生成训练样本数据，包括：
    针对每个共现信息，对所述字符串格式的共现信息进行特征处理，生成与所述共现信息对应的特征向量；
    获取与所述共现信息对应的兴趣点所关联的预设线下行为信息，并基于所述预设线下行为信息确定所述标签对应的真值；
    基于所述特征向量以及所述标签对应的真值，生成所述训练样本数据。
  6. 根据权利要求5所述的方法，其中，所述基于所述预设线下行为信息确定所述标签对应的真值，包括：
    在发生预设线下行为的情况下，将所述标签对应的真值确定为1；或者，
    在未发生预设线下行为的情况下，将所述标签对应的真值确定为0。
  7. 根据权利要求5所述的方法，其中，所述针对每个共现信息，对所述字符串格式的共现信息进行特征处理，生成与所述共现信息对应的特征向量，包括：
    对字符串格式的各个共现信息按照预设编号方式分别进行编号，并基于所述各个共现信息分别对应的编号，构建特征字典；所述特征字典用于将所述字符串格式的共现信息转换成特征向量；
    针对每个字符串格式的共现信息，基于所述特征字典对所述字符串格式的共现信息进行独热编码处理，生成与所述共现信息对应的特征向量。
  8. 根据权利要求7所述的方法，其中，所述基于所述特征字典对所述字符串格式的共现信息进行独热编码处理，生成与所述共现信息对应的特征向量，包括：
    确定所述特征字典中包含的编号的数量，并创建向量长度为所述数量的零向量；
    针对每个共现信息，根据所述共现信息在所述特征字典中对应的编号，将所述零向量中与所述编号相同的目标索引位的值修改为1，生成与所述共现信息对应的特征向量。
  9. 根据权利要求1所述的方法，其中，所述基于各个训练样本数据，对待训练的模型进行训练，得到训练好的模型，包括：
    将每个训练样本数据输入所述待训练的模型，得到与所述每个训练样本对应的预测结果；
    基于所述每个训练样本对应的预测结果以及所述每个训练样本数据的标签，对所述待训练的模型的模型参数进行调整；
    重复上述过程直到训练结果符合预设要求，得到所述训练好的模型。
  10. 一种兴趣点定位方法，包括：
    获取当前设备的当前WiFi扫描信息以及所述当前设备所处区域对应的每个兴趣点对应的关联WiFi信息；
    针对每个兴趣点，基于所述当前WiFi扫描信息以及所述兴趣点对应的关联WiFi信息，生成当前共现信息；
    将各个当前共现信息输入至训练好的模型，得到所述当前设备的兴趣点定位信息；其中，所述训练好的模型通过权利要求1-9中任一项所述的模型训练方法得到。
  11. 一种模型训练装置，包括：
    信息获取模块，配置为获取历史无线保真WiFi扫描信息以及历史WiFi 分布信息；所述历史WiFi扫描信息包括多组WiFi扫描信息，每组WiFi扫描信息包括至少一个第一WiFi信息，所述历史WiFi分布信息包括每个兴趣点关联的第二WiFi信息；
    信息生成模块，配置为针对每个兴趣点，基于所述兴趣点关联的第二WiFi信息以及每组WiFi扫描信息中的至少一个第一WiFi信息，生成与所述兴趣点对应的共现信息；所述共现信息用于表征每组中的每个第一WiFi信息相对于所述兴趣点关联的第二WiFi信息的扫描状态；
    样本生成模块，配置为针对每个共现信息，基于所述共现信息以及与所述共现信息对应的标签，生成训练样本数据；所述共现信息对应的标签基于与所述共现信息对应的兴趣点所关联的预设线下行为信息确定；
    模型训练模块，配置为基于各个训练样本数据，对待训练的模型进行训练，得到训练好的模型；所述训练好的模型用于基于当前WiFi扫描信息确定对应的兴趣点定位信息。
  12. 一种兴趣点定位装置，其中包括：
    获取模块，配置为获取当前设备的当前WiFi扫描信息以及所述当前设备所处区域对应的每个兴趣点对应的关联WiFi信息；
    生成模块，配置为针对每个兴趣点，基于所述当前WiFi扫描信息以及所述兴趣点对应的关联WiFi信息，生成当前共现信息；
    定位模块，配置为将各个当前共现信息输入至训练好的模型，得到所述当前设备的兴趣点定位信息；其中，所述训练好的模型通过权利要求1-9中任一项所述的模型训练方法得到。
  13. 一种电子设备，包括：处理器、存储器和总线，所述存储器存储有所述处理器可执行的机器可读请求，当电子设备运行时，所述处理器与所述存储器之间通过总线通信，所述机器可读请求被所述处理器执行时执行如权利要求1至9任意一项所述的模型训练方法或者权利要求10所述的兴趣点定位方法。
  14. 一种计算机可读存储介质，其中，该计算机可读存储介质上存储有计算机程序，该计算机程序被处理器运行时执行如权利要求1至9任意一项所述的模型训练方法或者权利要求10所述的兴趣点定位方法。
PCT/CN2024/096051 2023-06-08 2024-05-29 模型训练方法、装置、电子设备及存储介质 Ceased WO2024251004A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202310678757.5A CN116663677B (zh) 2023-06-08 2023-06-08 模型训练方法、装置、电子设备及存储介质
CN202310678757.5 2023-06-08

Publications (1)

Publication Number Publication Date
WO2024251004A1 true WO2024251004A1 (zh) 2024-12-12

Family

ID=87727636

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2024/096051 Ceased WO2024251004A1 (zh) 2023-06-08 2024-05-29 模型训练方法、装置、电子设备及存储介质

Country Status (2)

Country Link
CN (1) CN116663677B (zh)
WO (1) WO2024251004A1 (zh)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116663677B (zh) * 2023-06-08 2025-03-14 抖音视界有限公司 模型训练方法、装置、电子设备及存储介质

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103200673A (zh) * 2013-03-07 2013-07-10 浙江大学 无线终端定位方法和定位装置
CN110781256A (zh) * 2019-08-30 2020-02-11 腾讯大地通途（北京）科技有限公司 基于发送位置数据确定与Wi-Fi相匹配的POI的方法及装置
CN111954175A (zh) * 2020-08-25 2020-11-17 腾讯科技（深圳）有限公司 一种兴趣点到访判别方法和相关装置
CN112399555A (zh) * 2020-10-20 2021-02-23 北京嘀嘀无限科技发展有限公司 一种位置定位的方法、装置、可读存储介质和电子设备
CN112804634A (zh) * 2020-12-31 2021-05-14 北京嘀嘀无限科技发展有限公司 Wi-Fi信号的处理方法、装置、设备及存储介质
CN116193362A (zh) * 2023-03-02 2023-05-30 北京抖音智图科技有限公司 用于兴趣点信息维护的方法、装置、设备和存储介质
CN116663677A (zh) * 2023-06-08 2023-08-29 抖音视界有限公司 模型训练方法、装置、电子设备及存储介质

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CA2773302A1 (en) * 2011-04-05 2012-10-05 Her Majesty The Queen In Right Of Canada, As Represented By The Ministerof Industry, Through The Communications Research Centre Canada Cognitive wi-fi radio network
CN107729459A (zh) * 2017-09-30 2018-02-23 百度在线网络技术（北京）有限公司 地图兴趣点失效挖掘方法、装置、设备和计算机可读存储介质
CN110782284A (zh) * 2019-10-24 2020-02-11 腾讯科技（深圳）有限公司 信息推送方法、装置以及可读存储介质
CN111417066B (zh) * 2020-03-05 2021-01-05 滴图(北京)科技有限公司 定位方法、装置、电子设备和计算机可读存储介质
CN114861017A (zh) * 2022-04-02 2022-08-05 深圳依时货拉拉科技有限公司 兴趣点推荐方法、装置、计算机设备和存储介质

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103200673A (zh) * 2013-03-07 2013-07-10 浙江大学 无线终端定位方法和定位装置
CN110781256A (zh) * 2019-08-30 2020-02-11 腾讯大地通途（北京）科技有限公司 基于发送位置数据确定与Wi-Fi相匹配的POI的方法及装置
CN111954175A (zh) * 2020-08-25 2020-11-17 腾讯科技（深圳）有限公司 一种兴趣点到访判别方法和相关装置
CN112399555A (zh) * 2020-10-20 2021-02-23 北京嘀嘀无限科技发展有限公司 一种位置定位的方法、装置、可读存储介质和电子设备
CN112804634A (zh) * 2020-12-31 2021-05-14 北京嘀嘀无限科技发展有限公司 Wi-Fi信号的处理方法、装置、设备及存储介质
CN116193362A (zh) * 2023-03-02 2023-05-30 北京抖音智图科技有限公司 用于兴趣点信息维护的方法、装置、设备和存储介质
CN116663677A (zh) * 2023-06-08 2023-08-29 抖音视界有限公司 模型训练方法、装置、电子设备及存储介质

Also Published As

Publication number Publication date
CN116663677A (zh) 2023-08-29
CN116663677B (zh) 2025-03-14

Similar Documents

Publication Publication Date Title
CN108280115B (zh) 识别用户关系的方法及装置
CN114547267B (zh) 智能问答模型的生成方法、装置、计算设备和存储介质
CN111026858B (zh) 基于项目推荐模型的项目信息处理方法及装置
CN110309275B (zh) 一种对话生成的方法和装置
WO2022188534A1 (zh) 信息推送的方法和装置
CN110858226B (zh) 对话管理方法和装置
CN112084752B (zh) 基于自然语言的语句标注方法、装置、设备及存储介质
CN108805565B (zh) 基于区块链的承诺存在性证明方法、设备及可读存储介质
CN108682421B (zh) 一种语音识别方法、终端设备及计算机可读存储介质
CN114780701A (zh) 自动问答匹配方法、装置、计算机设备及存储介质
CN113515703A (zh) 信息推荐方法、装置、电子设备及可读存储介质
CN112906361A (zh) 文本数据的标注方法和装置、电子设备和存储介质
CN118377869A (zh) 一种大语言模型进行自校正事实增强的处理方法和装置
WO2024251004A1 (zh) 模型训练方法、装置、电子设备及存储介质
CN115408508A (zh) 对话推荐与模型训练方法、装置、电子设备和存储介质
CN118535565A (zh) 前端表单数据验证方法及装置
CN114970535B (zh) 意图识别方法、系统、装置及存储介质
CN111523295A (zh) 数据渲染方法、装置、介质及电子设备
CN116204602A (zh) 字码映射方法、装置、存储介质以及终端
CN115619142A (zh) 匹配数据的方法、装置、设备和计算机可读介质
CN118820312A (zh) 模型训练方法、信息提示方法及相关设备
CN116108810A (zh) 文本数据增强方法及装置
CN115687040A (zh) 访问时长确定方法、装置、设备、介质和程序产品
CN107688948A (zh) 理赔数据处理方法、装置、计算机设备和存储介质
CN115329767B (zh) 抽取文本实体的方法、装置、电子设备及存储介质

Legal Events

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

Ref document number: 24818537

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE