WO2022142493A1 - 业务数据处理方法、装置、电子设备和存储介质 - Google Patents

业务数据处理方法、装置、电子设备和存储介质 Download PDF

Info

Publication number
WO2022142493A1
WO2022142493A1 PCT/CN2021/119158 CN2021119158W WO2022142493A1 WO 2022142493 A1 WO2022142493 A1 WO 2022142493A1 CN 2021119158 W CN2021119158 W CN 2021119158W WO 2022142493 A1 WO2022142493 A1 WO 2022142493A1
Authority
WO
WIPO (PCT)
Prior art keywords
target
data
sample
learning model
moment
Prior art date
Application number
PCT/CN2021/119158
Other languages
English (en)
French (fr)
Inventor
王若兰
刘洋
张钧波
郑宇�
Original Assignee
京东城市(北京)数字科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 京东城市(北京)数字科技有限公司 filed Critical 京东城市(北京)数字科技有限公司
Publication of WO2022142493A1 publication Critical patent/WO2022142493A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00Administration; Management
    • G06Q10/04Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/21Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00Machine learning
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00Administration; Management
    • G06Q10/06Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/067Enterprise or organisation modelling

Definitions

  • the present application relates to the field of computer technology, and in particular, to a business data processing method, apparatus, electronic device, and storage medium.
  • the business system obtains the crowd consumption portraits in the corresponding area from the crowd consumption portrait database determined based on a single data source, and then , obtain the people flow data corresponding to the area from the business system, and combine the people flow data and crowd consumption portraits for business processing.
  • the regional population portraits based on unilateral data sources are inaccurate, resulting in inaccurate business processing results.
  • the present application provides a business data processing method, apparatus, electronic device and storage medium.
  • an embodiment of the present application proposes a service data processing method, including: receiving a service data request, wherein the service data request includes an identifier of a target area and a target time; and obtaining first data according to the identifier of the target area
  • the target area on the terminal consumes portrait information of the first group at the target moment; cooperates with the second group of the target area on other first data terminals to consume the portrait information at the target moment, through the first target federated learning model Correcting the consumption profile information of the first group to obtain the target group consumption profile information of the target area at the target time; acquiring the target people flow data of the target area at the target time; according to the target people flow
  • the data and the consumption profile information of the target group are used to determine the consumer group characteristic distribution information of the target area at the target time; the service requested by the business data request is processed according to the consumer group characteristic distribution information.
  • the acquiring the target traffic data of the target area at the target time includes: acquiring the first traffic data of the target area on the second data terminal at the target time; cooperating with other The second traffic data of the target area at the target moment on the second data terminal, and the first traffic data is corrected through the second target federated learning model to obtain the target area at the target Time-to-moment target traffic data.
  • the obtaining the first traffic data of the target area on the second data terminal at the target time includes: obtaining the target area in each history within a first preset time period The people flow data at the target time; based on the people flow data at each historical target time, determine the first people flow data of the target area on the second data terminal at the target time.
  • the obtaining, according to the identification of the target area, the consumption portrait information of the first group of people in the target area on the first data terminal at the target time includes: according to the identification of the target area , obtain the crowd consumption profile information of the target area at each historical moment within the second preset duration; determine the target area on the first data terminal according to the crowd consumption profile information at each historical moment The first group of people at the target moment consumes portrait information.
  • the training process of the first target federated learning model includes: obtaining the consumption portrait information of the first sample population of the sample area on the first data terminal at the sample moment; A federated learning model is trained on the consumption portrait information of the second sample population of the sample area at the sample moment on a data end to generate the first target federated learning model, wherein the first sample population is selected Among the consumer portrait information and the second sample population consumption portrait information, the sample population consumption portrait information with the largest population consumption portrait information is used as the data to be corrected for the model, and the remaining sample population consumption portrait information is used as the feature data of the model.
  • a target federated learning model is used to establish a regression relationship between the feature data and the data to be corrected.
  • the training of a federated learning model is performed by cooperating with the consumption profile information of the second sample population of the sample area on the other first data terminal at the sample moment, so as to generate the first target
  • the federated learning model includes: based on the consumption profile information of the first group, controlling the local learning model on the first data terminal to perform training to obtain an intermediate result; based on the consumption profile information of the second group, controlling the other
  • the first data terminal trains its own local learning model to obtain intermediate results; obtains the intermediate results of each training output of the local learning models on each of the first data terminals, and uses the intermediate results of each output.
  • the result is sent to the coordinator for summarization; the global intermediate result sent by the coordinator for each summary is received; the model parameters of the local learning model are adjusted based on the global intermediate result and the next round of training is continued until the preset is met Conditionally stop training to obtain the first target federated learning model.
  • the training process of the second target federated learning model includes: obtaining the first sample traffic data of the sample area on the second data terminal at the sample moment; cooperating with other second data A federated learning model is trained on the second sample traffic data of the sample area on the terminal at the sample moment to generate the second target federated learning model, wherein the first sample traffic data and all traffic data are selected.
  • the sample human flow data with the largest human flow data in the second sample human flow data is used as the data to be corrected for the model, the remaining sample human flow data is used as the feature data of the model, and the second target federated learning model is used to establish all the data.
  • the training of a federated learning model is performed by cooperating with the second sample traffic data of the sample area on the other second data terminal at the sample moment, so as to generate the second target federation
  • the learning model includes: controlling the local learning model on the second data terminal to perform training based on the first sample human flow data to obtain an intermediate result; based on the second sample human flow data, controlling the other third
  • the second data terminal trains its own local learning model to obtain intermediate results; obtains the intermediate results of each training output of the local learning models on each of the second data terminals, and uses the intermediate results of each output Send to the coordinator for summarization; receive the global intermediate result sent by the coordinator for each summary; adjust the model parameters of the local learning model based on the global intermediate result and continue the next round of training until the preset conditions are met Stop training to obtain the second target federated learning model.
  • the business data processing method of the embodiment of the present application after receiving the business data request, according to the identifier of the target area in the business data request, obtain the consumption portrait information of the first group of people in the target area on the first data terminal at the target time, and Cooperate with the consumption profile information of the second group of the target area at the target time on other first data terminals, and correct the consumption profile information of the first group through the first target federated learning model to obtain the target group consumption profile of the target area at the target time.
  • the service requested by the service data request is accurately processed, and the accuracy of the service processing is improved.
  • a service data processing apparatus including: a receiving module for receiving a service data request, wherein the service data request includes an identifier of a target area and a target time; a first obtaining module, which uses According to the identification of the target area, obtain the first group consumption portrait information of the target area on the first data terminal at the target moment; the correction module is used for cooperating with the target area on other first data terminals in the target area.
  • the consumption profile information of the second group at the target time, and the consumption profile information of the first group is corrected through the first target federated learning model to obtain the target group consumption profile information of the target area at the target time;
  • the second an acquisition module for acquiring the target traffic data of the target area at the target time;
  • a determining module for determining the consumption of the target area at the target time according to the target traffic data and the consumption profile information of the target group Crowd characteristic distribution information;
  • a service processing module configured to process the service requested by the service data request according to the consumer crowd characteristic distribution information.
  • the second acquisition module includes: a first acquisition unit, configured to acquire the first traffic data of the target area on the second data terminal at the target time; a correction unit, used In cooperation with the second traffic data of the target area at the target moment on other second data terminals, the first traffic data is corrected by the second target federated learning model to obtain the target area at the target time.
  • the target traffic data at the target moment is: a first acquisition unit, configured to acquire the first traffic data of the target area on the second data terminal at the target time; a correction unit, used In cooperation with the second traffic data of the target area at the target moment on other second data terminals, the first traffic data is corrected by the second target federated learning model to obtain the target area at the target time.
  • the target traffic data at the target moment is
  • the first obtaining unit is specifically configured to: obtain the traffic data of the target area at each historical target moment within a first preset duration; based on each historical target The human flow data at the moment is used to determine the first human flow data of the target area on the second data terminal at the target moment.
  • the first obtaining module includes: a second obtaining unit, configured to obtain, according to the identifier of the target area, the target area at each historical moment within a second preset time period and the determining unit is configured to determine, according to the crowd consumption portrait information at each historical moment, the first crowd consumption portrait information of the target area on the first data terminal at the target moment.
  • the apparatus further includes a first training module, and the first training module includes: a third obtaining unit, configured to obtain the sample area on the first data terminal at the sample moment.
  • the first sample population consumes portrait information; the first training unit is used for training the federated learning model in conjunction with the second sample population consumption portrait information of the sample area on other first data terminals at the sample moment, to generate The first target federated learning model, wherein among the first sample group consumption profile information and the second sample group consumption profile information, the sample group consumption profile information with the largest group consumption profile information is selected as the data to be corrected for the model , and the remaining sample population consumption portrait information is used as feature data of the model, and the first target federated learning model is used to establish a regression relationship between the feature data and the data to be corrected.
  • the first training unit is specifically configured to: control the local learning model on the first data terminal to perform training based on the consumption profile information of the first group to obtain an intermediate result; Based on the consumption profile information of the second group, control the other first data terminals to train their own local learning models to obtain intermediate results; obtain each training of the local learning models on each of the first data terminals output intermediate results, and send the intermediate results output each time to the coordinator for aggregation; receive the global intermediate results sent by the coordinator for each aggregation; adjust the local learning model based on the global intermediate results and continue to the next round of training until the preset conditions are met, and stop training to obtain the first target federated learning model.
  • the apparatus further includes a second training module
  • the second training module includes: a fourth acquisition unit, configured to acquire the sample area on the second data terminal at the sample moment The first sample human flow data; the second training unit is used to perform federated learning model training with the second sample human flow data of the sample area on other second data terminals at the sample moment, so as to generate the first sample area.
  • Two-objective federated learning model wherein the sample human flow data with the largest human flow data among the first sample human flow data and the second sample human flow data is selected as the data to be corrected for the model, and the remaining sample human flow data is used as The feature data of the model, and the second target federated learning model is used to establish a regression relationship between the feature data and the data to be corrected.
  • the second training unit is specifically configured to: control the local learning model on the second data terminal to perform training based on the first sample traffic data, so as to obtain an intermediate result; Based on the second sample traffic data, control the other second data terminals to train their own local learning models to obtain intermediate results; obtain each training of the local learning models on each of the second data terminals output intermediate results, and send the intermediate results output each time to the coordinator for aggregation; receive the global intermediate results sent by the coordinator for each aggregation; adjust the local learning model based on the global intermediate results and continue the next round of training until the preset conditions are met, and stop training to obtain the second target federated learning model.
  • the business data processing apparatus of the embodiment of the present application after receiving the business data request, obtains the consumption portrait information of the first group of people in the target area on the first data terminal at the target time according to the identifier of the target area in the business data request, and Cooperate with the consumption profile information of the second group of the target area at the target time on other first data terminals, and correct the consumption profile information of the first group through the first target federated learning model to obtain the target group consumption profile of the target area at the target time. Then, combined with the consumption profile information of the target group and the target traffic data of the target area at the target time, determine the distribution information of the characteristics of the consumer group in the target area at the target time, and the data requested by the business data request according to the distribution information of the characteristics of the consumer group. business is processed. Therefore, based on the traffic data and the consumption profile information of the target group determined in combination with multiple data terminals, the service requested by the service data request is accurately processed, and the accuracy of the service processing is improved.
  • Another embodiment of the present application provides an electronic device, including: an electronic device, including: a memory, and a processor; the memory stores computer instructions, and when the computer instructions are executed by the processor, The service data processing method of the embodiment of the present application is implemented.
  • Another embodiment of the present application provides a non-transitory computer-readable storage medium storing computer instructions, where the computer instructions are used to cause the computer to execute the service data processing method disclosed in the embodiments of the present application.
  • Another embodiment of the present application provides a computer program product, which implements the service data processing method in the embodiment of the present application when an instruction processor in the computer program product is executed.
  • FIG. 1 is a schematic flowchart of a business data processing method according to an embodiment of the present application.
  • FIG. 2 is a schematic flowchart of training a first target federated learning model.
  • FIG. 3 is a schematic diagram of a refinement flow of step 202 .
  • FIG. 4 is a schematic flowchart of the second target federated learning model.
  • FIG. 5 is a schematic diagram of a refinement flow of step 402 .
  • FIG. 6 is a diagram showing an example of the relationship between various layers in the service device.
  • FIG. 7 is a schematic structural diagram of a service data processing apparatus according to an embodiment of the present application.
  • FIG. 8 is a schematic structural diagram of a service data processing apparatus according to another embodiment of the present application.
  • FIG. 9 is a block diagram of an electronic device according to an embodiment of the present application.
  • FIG. 1 is a schematic flowchart of a business data processing method according to an embodiment of the present application.
  • the executive body of the business data processing method provided in this embodiment is a business data processing apparatus, and the business data processing apparatus may be implemented by software and/or hardware.
  • the business data processing apparatus in this embodiment It may be configured in an electronic device, and the electronic device in this embodiment may include devices such as a terminal device and a server.
  • the business data processing method may include:
  • Step 101 Receive a service data request, wherein the service data request includes an identifier of a target area and a target time.
  • a service data request may be made to the service data processing apparatus based on a service client or a service website.
  • Step 102 Acquire, according to the identifier of the target area, the consumption portrait information of the first group of people of the target area on the first data terminal at the target moment.
  • the consumption profile information of the first group refers to the labelled profile abstracted by analyzing the corresponding user consumption behavior data recorded on the first data terminal.
  • the above-mentioned corresponding user consumption behavior data is the consumption behavior data of the user in the target area at the target time.
  • the above-mentioned acquisition of the consumption profile information of the first group of people in the target area on the first data terminal at the target time according to the identification of the target area can be achieved in various ways, and an example is as follows:
  • the data of the target area at each historical moment in the second preset time period can be obtained according to the identifier of the target area.
  • Crowd consumption portrait information according to the crowd consumption portrait information at each historical moment, determine the first crowd consumption portrait information of the target area on the first data terminal at the target moment. That is to say, the consumption profile information of the first group of people in the target area at the target time can be obtained from the average profile over a period of time in history.
  • the above-mentioned crowd consumption profile information may be based on the user's purchase behavior data recorded on the first data terminal, and may describe some consumption characteristics of the user according to the user's historical purchase behavior, such as: the user's consumption preference for commodities, the user's consumption level etc.
  • the first data terminal itself also stores the user's personal portrait, for example, the user's age, education background, whether there is a car or the like. Therefore, by aggregating the consumption characteristics and personal portraits of each user in the area, it is possible to form information on the group consumption portraits of the area at different times.
  • the first data terminal on the first data terminal at the target moment can be obtained based on the identification of the target area through the database of crowd consumption portrait information at each time in each area pre-stored on the first data terminal. Crowd consumption portrait information.
  • Step 103 in coordination with the consumption profile information of the second group of people in the target area on other first data terminals at the target time, correct the consumption profile information of the first group through the first target federated learning model, so as to obtain the target area of the target area at the target time. Crowd consumption portrait information.
  • the first target federated learning model is combined with the crowd consumption profile information on the first data terminal and other first data terminals for correction, so that the crowd consumption profile information obtained through multiple different data terminals can be accurately It depicts the crowd consumption portrait of the target area at the target moment. It is convenient for subsequent accurate provision of business processing based on the consumption profile of the target area at the target time.
  • the manner in which the other first data terminal determines the consumption profile information of the second group of people in the target area at the target time is the same as the way in which the above-mentioned first data terminal determines the first data terminal in the target area at the target time.
  • the way people consume portrait information is similar, and will not be repeated here.
  • first data terminal and the second data terminal in this embodiment are different data terminals, both of which can collect statistics on user consumption behavior data.
  • the business services provided by the first data terminal and the second data terminal may be the same.
  • the consumption profile information of the second group refers to the labelled profile abstracted by analyzing the corresponding user consumption behavior data recorded on the second data terminal.
  • the above-mentioned corresponding user consumption behavior data is the consumption behavior data of the user in the target area at the target time.
  • the above-mentioned first target federated learning model is obtained by training based on the crowd consumption portrait information of the first data terminal and other data terminals.
  • the training process of the first target federated learning model will be described in subsequent embodiments.
  • Step 104 Acquire target traffic data of the target area at the target time.
  • the human flow data of the target area at each historical moment within a preset time period can be obtained, and then, based on the pre-trained human flow prediction model, the target human flow data of the target area at the target time can be determined .
  • the people flow data at each historical time (t-n,...,t-1) in this area can be obtained, and then, the obtained people flow data at each historical time can be Input to the pre-trained human flow prediction model to predict the human flow data of the target area at time t.
  • a possible implementation manner of the people flow data in the above-mentioned regions at each historical moment is: based on the user's movement trajectory data recorded by multiple telecom operator platforms, and aggregated based on the user's movement trajectory data, to Form the flow data of people in different areas and at different times.
  • the above-mentioned people flow prediction model is obtained by pre-training, wherein, in order to enable the people flow prediction model to accurately predict the people flow data of the target area at the corresponding target time.
  • the people flow prediction model can be trained based on the land area using the accurate people flow information. And after the training is completed, based on the trained traffic prediction model, the traffic data in the low-credibility area can be re-estimated to achieve the correction effect.
  • the target traffic flow data of the target area on the target area at the target moment is corrected by the second target federated learning model to obtain the target traffic flow data of the target area at the target moment.
  • the data of the first traffic data of the target area on the second data terminal at the target time can be obtained.
  • a possible implementation method is: the human flow data of the target area at each historical target moment within the first preset duration can be obtained; based on the human flow data at each historical target moment, it is determined that the target area on the second data terminal is The first person traffic data at the target moment.
  • the first preset duration is a preset duration.
  • the first preset duration may be one month, or one week.
  • the value of the first preset duration may be determined based on actual business requirements. This embodiment does not specifically limit this.
  • the traffic data of each historical target moment can be input into the traffic corresponding to the second data terminal
  • a prediction model is used to determine the first people flow data of the target area on the second data terminal at the target time through the people flow prediction model.
  • Step 105 according to the target people flow data and the consumption profile information of the target group, determine the distribution information of the characteristics of the consumer groups in the target area at the target time.
  • Step 106 Process the service requested by the service data request according to the distribution information of the characteristics of the consumer group.
  • an accurate shop location can be selected based on the distribution information of the characteristics of the consumer population in the target area at the target time, and the results of the shop location selection can be calculated. Provided to business data requesters.
  • the above-mentioned service data request is an advertisement placement service
  • the preference ratio of the product is combined with the preference ratio of the product for advertising.
  • the business data processing method of the embodiment of the present application after receiving the business data request, according to the identifier of the target area in the business data request, obtain the consumption portrait information of the first group of people in the target area on the first data terminal at the target time, and Cooperate with the consumption profile information of the second group of the target area at the target time on other first data terminals, and correct the consumption profile information of the first group through the first target federated learning model to obtain the target group consumption profile of the target area at the target time.
  • the service requested by the service data request is accurately processed, and the accuracy of the service processing is improved.
  • the first target federated learning model in order to enable the first target federated learning model to accurately correct the consumption profile information of the first population, can be trained by combining data on multiple first data terminals .
  • the process of training the first target federated learning model may include:
  • Step 201 acquiring the consumption portrait information of the first sample population of the sample area on the first data terminal at the sample moment.
  • the consumption profile information of the first sample population in the sample area at the sample moment may be collected in advance.
  • Step 202 train the federated learning model in conjunction with the consumption portrait information of the second sample population at the sample moment in the sample regions on other first data terminals, so as to generate a first target federated learning model.
  • the consumption portrait information of the sample population with the largest population consumption portrait information among the first sample population consumption portrait information and the second sample population consumption portrait information is selected as the data to be corrected for the model, and the remaining sample population consumption portrait information is used as the feature data of the model.
  • the first target federated learning model is used to establish the regression relationship between the feature data and the data to be corrected.
  • FIG. 3 a possible implementation of the above step 202 is shown in FIG. 3 . , which can include:
  • Step 301 based on the consumption profile information of the first group, control the local learning model on the first data terminal to perform training, so as to obtain an intermediate result.
  • Step 302 based on the consumption profile information of the second group, control other first data terminals to train their own local learning models to obtain intermediate results.
  • Step 303 Obtain the intermediate results of each training output of the local learning model on each first data terminal, and send the intermediate results of each output to the coordinator for summarization.
  • Step 304 Receive the global intermediate result aggregated each time sent by the coordinator.
  • Step 305 adjust the model parameters of the local learning model based on the global intermediate result and continue the next round of training, until the preset condition is met, the training is stopped to obtain the first target federated learning model.
  • the first target federated learning model can be accurately trained without the data on each data terminal being stored in the database, which facilitates subsequent training based on the first target federated learning model. Perform traffic corrections.
  • the second target The federated learning model is trained.
  • the training process of the second target federated learning model may include:
  • Step 401 Acquire first sample human flow data of the sample area on the second data terminal at the sample moment.
  • Step 402 train the federated learning model with the second sample traffic data of the sample area on the other second data terminal at the sample moment, so as to generate a second target federated learning model.
  • the sample human flow data with the largest human flow data in the first sample human flow data and the second sample human flow data is selected as the data to be corrected for the model, and the remaining sample human flow data is used as the feature data of the model.
  • the second target federated learning The model is used to establish the regression relationship between the feature data and the data to be corrected.
  • step 402 for:
  • Step 501 based on the first sample traffic data, control the local learning model on the second data terminal to perform training to obtain an intermediate result.
  • Step 502 based on the second sample traffic data, control other second data terminals to train their own local learning models to obtain intermediate results.
  • Step 503 Obtain the intermediate results of each training output of the local learning model on each second data terminal, and send the intermediate results of each output to the coordinator for summarization.
  • Step 504 Receive the aggregated global intermediate result sent by the coordinator each time.
  • Step 505 adjust the model parameters of the local learning model based on the global intermediate result and continue the next round of training, until the preset condition is met, the training is stopped to obtain the second target federated learning model.
  • the second target federated learning model is accurately trained when the data on multiple data terminals is not stored in the database, which facilitates the subsequent training based on the second target federated learning model. Accurately determine the target flow data of the corresponding area at the corresponding time.
  • the data terminal A in the figure processes the human flow trajectory data in combination with the human flow trajectory and the urban road network to obtain the regional population information, and through the data terminal B, performs the crowd portrait data processing on the population characteristics on the data terminal B to obtain the regional group image. information.
  • the data terminal C in the figure combined with the human flow trajectory and Point of Information (POI, Point of Information), carries out the human flow trajectory data processing to obtain the regional population information, and through the data terminal B, the population characteristics on the data terminal D are crowded. Image data processing to obtain regional group image information.
  • POI Point of Information
  • the region portrait on the local end can be corrected in combination with the region portrait on the other end.
  • the flow of people on the local end can be corrected in combination with the regional flow of people on other ends.
  • cross-domain crowd multi-dimensional real-time calculation can be used to obtain regional human flow portraits, regional population consumption portraits, regional energy portraits, and regional security portraits.
  • the application layer it can be applied based on the regional portrait information obtained by the knowledge layer.
  • the information in the knowledge layer can serve the application layer to guide applications such as store location selection, accurate advertisement placement, urban traffic planning, and fire station planning.
  • the knowledge layer can assist merchants in selecting store locations; Estimate the conversion rate of advertisements and the revenue of advertisements; based on the real-time regional human flow portrait, it can assist the government to plan traffic and recommend reasonable travel methods for users; according to the urban regional safety portrait, it can assist in planning the construction of urban fire stations to improve the urban safety factor, etc. .
  • an embodiment of the present application further provides a business data processing apparatus.
  • the implementation manner of the business data processing method is also applicable to the business data processing apparatus provided in this embodiment, and will not be described in detail in this embodiment.
  • FIG. 7 is a schematic structural diagram of a service data processing apparatus according to an embodiment of the present application.
  • the service data processing apparatus 700 includes:
  • the receiving module 701 is configured to receive a service data request, wherein the service data request includes an identifier of a target area and a target time.
  • the first obtaining module 702 is configured to obtain, according to the identifier of the target area, the consumption portrait information of the first group of people of the target area on the first data terminal at the target time.
  • the correction module 703 is used for cooperating with the consumption profile information of the second group of the target area on other first data terminals at the target time, and corrects the consumption profile information of the first group through the first target federated learning model, so as to obtain the target area in the target area.
  • the target group consumes portrait information at any time.
  • the second acquiring module 704 acquires the target traffic data of the target area at the target time.
  • the determining module 705 is configured to determine the characteristic distribution information of the consumer groups in the target area at the target time according to the target people flow data and the consumption profile information of the target group.
  • the service processing module 706 is configured to process the service requested by the service data request according to the consumer group characteristic distribution information.
  • the second obtaining module 704 may include:
  • the first obtaining unit 7041 is used to obtain the first traffic data of the target area on the second data terminal at the target moment;
  • the correction unit 7042 is used to coordinate the second traffic data of the target area on the other second data terminals at the target time, and correct the first traffic data through the second target federated learning model to obtain the target area at the target time.
  • Target traffic data is used to coordinate the second traffic data of the target area on the other second data terminals at the target time, and correct the first traffic data through the second target federated learning model to obtain the target area at the target time.
  • the first obtaining unit 7041 is specifically configured to: obtain the traffic data of the target area at each historical target moment within the first preset duration; based on the traffic data of each historical target moment , and determine the first traffic data of the target area on the second data terminal at the target time.
  • the above-mentioned first obtaining module 702 may include:
  • the second obtaining unit 7021 is configured to obtain, according to the identifier of the target area, the crowd consumption profile information of the target area at each historical moment within the second preset time period;
  • the determining unit 7022 is configured to determine, according to the crowd consumption portrait information at each historical moment, the first crowd consumption portrait information of the target area on the first data terminal at the target time.
  • the device further includes a first training module 707, and the first training module 707 may include:
  • a third obtaining unit 7071 configured to obtain the consumption portrait information of the first sample population of the sample area on the first data terminal at the sample moment;
  • the first training unit 7072 is used for training the federated learning model in conjunction with the consumption portrait information of the second sample population of the sample area on the other first data terminal at the sample moment, so as to generate the first target federated learning model, wherein the first target federated learning model is selected.
  • the consumption profile information of the sample population and the consumption profile information of the second sample population are the consumption profile information of the sample population with the largest consumption profile information as the data to be corrected for the model, and the consumption profile information of the remaining sample population is used as the feature data of the model.
  • the first target federation The learning model is used to establish the regression relationship between the feature data and the data to be corrected.
  • the above-mentioned first training unit 7072 is specifically configured to: control the local learning model on the first data terminal to perform training based on the consumption profile information of the first group, so as to obtain an intermediate result; Consume portrait information, control other first data terminals to train their own local learning models to obtain intermediate results; obtain the intermediate results of each training output of the local learning models on each first data terminal, and output each time
  • the intermediate results are sent to the coordinator for aggregation; the global intermediate results sent by the coordinator for each summary are received; the model parameters of the local learning model are adjusted based on the global intermediate results and the next round of training is continued until the preset conditions are met and the training is stopped.
  • the first objective federated learning model is specifically configured to: control the local learning model on the first data terminal to perform training based on the consumption profile information of the first group, so as to obtain an intermediate result; Consume portrait information, control other first data terminals to train their own local learning models to obtain intermediate results; obtain the intermediate results of each training output of the local learning models on each first data terminal,
  • the apparatus further includes a second training module 708, and the second training module 708 includes:
  • the fourth acquisition unit 7081 is used to acquire the first sample human flow data of the sample area on the second data terminal at the sample moment;
  • the second training unit 7082 is used for training the federated learning model with the second sample traffic data of the sample regions on other second data terminals at the sample moment, so as to generate a second target federated learning model, wherein the first sample is selected Among the personal traffic data and the second sample traffic data, the sample traffic data with the largest traffic data is used as the data to be corrected for the model, the remaining sample traffic data is used as the feature data of the model, and the second target federated learning model is used to establish the feature data. and the regression relationship between the data to be corrected.
  • the above-mentioned second training unit 7082 is specifically configured to: control the local learning model on the second data terminal to perform training based on the first sample traffic data, so as to obtain an intermediate result; based on the second sample People flow data, control other second data terminals to train their own local learning models to obtain intermediate results; obtain the intermediate results of each training output of the local learning models on each second data terminal, and output each time
  • the intermediate results are sent to the coordinator for aggregation; the global intermediate results sent by the coordinator for each summary are received; the model parameters of the local learning model are adjusted based on the global intermediate results and the next round of training is continued until the preset conditions are met and the training is stopped.
  • the second objective federated learning model is specifically configured to: control the local learning model on the second data terminal to perform training based on the first sample traffic data, so as to obtain an intermediate result; based on the second sample People flow data, control other second data terminals to train their own local learning models to obtain intermediate results; obtain the intermediate results of each training output of the local learning models on each
  • the business data processing apparatus of the embodiment of the present application after receiving the business data request, obtains the consumption portrait information of the first group of people in the target area on the first data terminal at the target time according to the identifier of the target area in the business data request, and Cooperate with the consumption profile information of the second group of the target area at the target moment on other first data terminals, and correct the consumption profile information of the first group through the first target federated learning model to obtain the target group consumption profile of the target area at the target moment.
  • the data requested by the business data request is determined. business is processed. Therefore, based on the traffic data and the consumption profile information of the target group determined in combination with multiple data terminals, the service requested by the service data request is accurately processed, and the accuracy of the service processing is improved.
  • the present application further provides an electronic device and a readable storage medium.
  • FIG. 9 it is a block diagram of an electronic device according to an embodiment of the present application.
  • the electronic device includes:
  • Memory 901 Memory 901 , processor 902 , and computer instructions stored on memory 901 and executable on processor 902 .
  • the electronic device also includes:
  • the communication interface 903 is used for communication between the memory 901 and the processor 902 .
  • Memory 901 for storing computer instructions executable on processor 902 .
  • the memory 901 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk memory.
  • the processor 902 is configured to implement the service data processing method of the above embodiment when executing the program.
  • the bus can be an Industry Standard Architecture (referred to as ISA) bus, a Peripheral Component (referred to as PCI) bus, or an Extended Industry Standard Architecture (referred to as EISA) bus or the like.
  • ISA Industry Standard Architecture
  • PCI Peripheral Component
  • EISA Extended Industry Standard Architecture
  • the bus can be divided into address bus, data bus, control bus and so on. For ease of presentation, only one thick line is used in FIG. 9, but it does not mean that there is only one bus or one type of bus.
  • the memory 901, the processor 902 and the communication interface 903 are integrated on a chip, the memory 901, the processor 902 and the communication interface 903 can communicate with each other through the internal interface.
  • the processor 902 may be a central processing unit (Central Processing Unit, referred to as CPU), or a specific integrated circuit (Application Specific Integrated Circuit, referred to as ASIC), or is configured to implement one or more of the embodiments of the present application integrated circuit.
  • CPU Central Processing Unit
  • ASIC Application Specific Integrated Circuit
  • the present application also provides a computer program product, which implements the business data processing method of the embodiment of the present application when an instruction processor in the computer program product is executed.
  • first and second are only used for descriptive purposes, and should not be construed as indicating or implying relative importance or implying the number of indicated technical features. Thus, a feature delimited with “first”, “second” may expressly or implicitly include at least one of that feature.
  • plurality means at least two, such as two, three, etc., unless expressly and specifically defined otherwise.
  • a "computer-readable medium” can be any device that can contain, store, communicate, propagate, or transport the program for use by or in connection with an instruction execution system, apparatus, or apparatus.
  • computer readable media include the following: electrical connections with one or more wiring (electronic devices), portable computer disk cartridges (magnetic devices), random access memory (RAM), Read Only Memory (ROM), Erasable Editable Read Only Memory (EPROM or Flash Memory), Fiber Optic Devices, and Portable Compact Disc Read Only Memory (CDROM).
  • the computer readable medium may even be paper or other suitable medium on which the program may be printed, as may be done, for example, by optically scanning the paper or other medium, followed by editing, interpretation, or other suitable means as necessary process to obtain the program electronically and then store it in computer memory.
  • each functional unit in each embodiment of the present application may be integrated into one processing module, or each unit may exist physically alone, or two or more units may be integrated into one module.
  • the above-mentioned integrated modules can be implemented in the form of hardware, and can also be implemented in the form of software function modules. If the integrated modules are implemented in the form of software functional modules and sold or used as independent products, they may also be stored in a computer-readable storage medium.
  • the above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disk, and the like.

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Strategic Management (AREA)
  • Human Resources & Organizations (AREA)
  • General Physics & Mathematics (AREA)
  • Economics (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Game Theory and Decision Science (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • General Business, Economics & Management (AREA)
  • Software Systems (AREA)
  • Tourism & Hospitality (AREA)
  • General Engineering & Computer Science (AREA)
  • Quality & Reliability (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Operations Research (AREA)
  • Development Economics (AREA)
  • Marketing (AREA)
  • Mathematical Physics (AREA)
  • Educational Administration (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Computing Systems (AREA)
  • Medical Informatics (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Image Analysis (AREA)

Abstract

一种业务数据处理方法、装置(700)、电子设备和存储介质,其中,方法包括:根据业务数据请求中的目标区域的标识,获取第一数据端上的目标区域在目标时刻的第一人群消费画像信息(102),并协同其他第一数据端上的目标区域在目标时刻的第二人群消费画像信息,通过第一目标联邦学习模型对第一人群消费画像信息进行修正,以得到目标区域在目标时刻的目标人群消费画像信息(103),结合目标人群消费画像信息以及目标区域在目标时刻的目标人流量数据,对业务数据请求所请求的业务进行处理。由此,基于人流量数据以及结合多个数据端所确定出的目标人群消费画像信息准确实现对业务数据请求所请求的业务进行准确处理。

Description

业务数据处理方法、装置、电子设备和存储介质
相关申请的交叉引用
本申请要求京东城市(北京)数字科技有限公司公司于2020年12月29日提交的、发明名称为“业务数据处理方法、装置、电子设备和存储介质”的、中国专利申请号“202011597920.8”的优先权。
技术领域
本申请涉及计算机技术领域,尤其涉及业务数据处理方法、装置、电子设备和存储介质。
背景技术
相关技术中,在需要结合区域人流量数据和人群消费画像进行业务处理的场景中,通常业务***从基于单个数据源所确定出的人群消费画像数据库中,获取对应区域中的人群消费画像,然后,从业务***中获取区域对应的人流量数据,并结合人流量数据和人群消费画像进行业务处理。然而,这种基于单方面数据源刻画出的区域人群画像并不精准,从而导致所提供的业务处理结果并不准确。
发明内容
本申请提出一种业务数据处理方法、装置、电子设备和存储介质。
本申请一方面实施例提出了一种业务数据处理方法,包括:接收业务数据请求,其中,所述业务数据请求包括目标区域的标识以及目标时刻;根据所述目标区域的标识,获取第一数据端上的目标区域在所述目标时刻的第一人群消费画像信息;协同其他第一数据端上的所述目标区域在所述目标时刻的第二人群消费画像信息,通过第一目标联邦学习模型对所述第一人群消费画像信息进行修正,以得到所述目标区域在所述目标时刻的目标人群消费画像信息;获取所述目标区域在目标时刻的目标人流量数据;根据所述目标人流量数据和所述目标人群消费画像信息,确定所述目标区域在所述目标时刻的消费人群特征分布信息;根据所述消费人群特征分布信息对所述业务数据请求所请求的业务进行处理。
在本申请的一个实施例中,所述获取所述目标区域在目标时刻的目标人流量数据,包括:获取第二数据端上的所述目标区域在目标时刻的第一人流量数据;协同其他第二数据 端上的所述目标区域在所述目标时刻的第二人流量数据,通过第二目标联邦学习模型对所述第一人流量数据进行修正,以获取所述目标区域在所述目标时刻的目标人流量数据。
在本申请的一个实施例中,所述获取第二数据端上的所述目标区域在目标时刻的第一人流量数据,包括:获取在第一预设时长内所述目标区域在每个历史目标时刻的人流量数据;基于每个所述历史目标时刻的人流量数据,确定所述第二数据端上的所述目标区域在目标时刻的第一人流量数据。
在本申请的一个实施例中,所述根据所述目标区域的标识,获取第一数据端上的目标区域在所述目标时刻的第一人群消费画像信息,包括:根据所述目标区域的标识,获取在第二预设时长内所述目标区域在每个历史时刻的人群消费画像信息;根据每个所述历史时刻的人群消费画像信息,确定所述第一数据端上的所述目标区域在目标时刻的第一人群消费画像信息。
在本申请的一个实施例中,所述第一目标联邦学习模型的训练过程,包括:获取所述第一数据端上的样本区域在样本时刻的第一样本人群消费画像信息;协同其他第一数据端上的所述样本区域在所述样本时刻的第二样本人群消费画像信息进行联邦学习模型的训练,以生成所述第一目标联邦学习模型,其中,选取所述第一样本人群消费画像信息和所述第二样本人群消费画像信息中人群消费画像信息最大的样本人群消费画像信息作为模型的待修正数据,剩余的样本人群消费画像信息作为所述模型的特征数据,所述第一目标联邦学习模型用于建立所述特征数据和所述待修正数据之间的回归关系。
在本申请的一个实施例中,所述协同其他第一数据端上的所述样本区域在所述样本时刻的第二样本人群消费画像信息进行联邦学习模型的训练,以生成所述第一目标联邦学习模型,包括:基于所述第一人群消费画像信息,控制所述第一数据端上的本地学习模型进行训练,以得到中间结果;基于所述第二人群消费画像信息,控制所述其他第一数据端对其自身上的本地学习模型进行训练,以得到中间结果;获取各个所述第一数据端上的本地学习模型每次训练输出的中间结果,并将每次输出的所述中间结果发送给协调方进行汇总;接收所述协调方发送的每次汇总出的全局中间结果;基于所述全局中间结果调整所述本地学习模型的模型参数并继续下一轮训练,直至满足预设条件停止训练得到所述第一目标联邦学习模型。
在本申请的一个实施例中,所述第二目标联邦学习模型的训练过程,包括:获取所述第二数据端上的样本区域在样本时刻的第一样本人流量数据;协同其他第二数据端上的所述样本区域在所述样本时刻的第二样本人流量数据进行联邦学习模型的训练,以生成所述第二目标联邦学习模型,其中,选取所述第一样本人流量数据和所述第二样本人流量数据 中人流量数据最大的样本人流量数据作为模型的待修正数据,剩余的样本人流量数据作为所述模型的特征数据,所述第二目标联邦学习模型用于建立所述特征数据和所述待修正数据之间的回归关系。
在本申请的一个实施例中,所述协同其他第二数据端上的所述样本区域在所述样本时刻的第二样本人流量数据进行联邦学习模型的训练,以生成所述第二目标联邦学习模型,包括:基于所述第一样本人流量数据,控制所述第二数据端上的本地学习模型进行训练,以得到中间结果;基于所述第二样本人流量数据,控制所述其他第二数据端对其自身上的本地学习模型进行训练,以得到中间结果;获取各个所述第二数据端上的本地学习模型每次训练输出的中间结果,并将每次输出的所述中间结果发送给协调方进行汇总;接收所述协调方发送的每次汇总出的全局中间结果;基于所述全局中间结果调整所述本地学习模型的模型参数并继续下一轮训练,直至满足预设条件停止训练得到所述第二目标联邦学习模型。
本申请实施例的业务数据处理方法,在接收到业务数据请求后,根据业务数据请求中的目标区域的标识,获取第一数据端上的目标区域在目标时刻的第一人群消费画像信息,并协同其他第一数据端上的目标区域在目标时刻的第二人群消费画像信息,通过第一目标联邦学习模型对第一人群消费画像信息进行修正,以得到目标区域在目标时刻的目标人群消费画像信息,然后,结合目标人群消费画像信息以及该目标区域在目标时刻的目标人流量数据,确定目标区域在目标时刻的消费人群特征分布信息,以及根据消费人群特征分布信息对业务数据请求所请求的业务进行处理。由此,基于人流量数据以及结合多个数据端所确定出的目标人群消费画像信息准确实现了对业务数据请求所请求的业务进行准确处理,提高了业务处理的准确率。
本申请另一方面实施例提出了一种业务数据处理装置,包括:接收模块,用于接收业务数据请求,其中,所述业务数据请求包括目标区域的标识以及目标时刻;第一获取模块,用于根据所述目标区域的标识,获取第一数据端上的目标区域在所述目标时刻的第一人群消费画像信息;修正模块,用于协同其他第一数据端上的所述目标区域在所述目标时刻的第二人群消费画像信息,通过第一目标联邦学习模型对所述第一人群消费画像信息进行修正,以得到所述目标区域在所述目标时刻的目标人群消费画像信息;第二获取模块,获取所述目标区域在目标时刻的目标人流量数据;确定模块,用于根据所述目标人流量数据和所述目标人群消费画像信息,确定所述目标区域在所述目标时刻的消费人群特征分布信息;业务处理模块,用于根据所述消费人群特征分布信息对所述业务数据请求所请求的业务进行处理。
在本申请的一个实施例中,所述第二获取模块,包括:第一获取单元,用于获取第二数据端上的所述目标区域在目标时刻的第一人流量数据;修正单元,用于协同其他第二数据端上的所述目标区域在所述目标时刻的第二人流量数据,通过第二目标联邦学习模型对所述第一人流量数据进行修正,以获取所述目标区域在所述目标时刻的目标人流量数据。
在本申请的一个实施例中,所述第一获取单元,具体用于:获取在第一预设时长内所述目标区域在每个历史目标时刻的人流量数据;基于每个所述历史目标时刻的人流量数据,确定所述第二数据端上的所述目标区域在目标时刻的第一人流量数据。
在本申请的一个实施例中,所述第一获取模块,包括:第二获取单元,用于根据所述目标区域的标识,获取在第二预设时长内所述目标区域在每个历史时刻的人群消费画像信息;确定单元,用于根据每个所述历史时刻的人群消费画像信息,确定所述第一数据端上的所述目标区域在目标时刻的第一人群消费画像信息。
在本申请的一个实施例中,所述装置还包括第一训练模块,所述第一训练模块,包括:第三获取单元,用于获取所述第一数据端上的样本区域在样本时刻的第一样本人群消费画像信息;第一训练单元,用于协同其他第一数据端上的所述样本区域在所述样本时刻的第二样本人群消费画像信息进行联邦学习模型的训练,以生成所述第一目标联邦学习模型,其中,选取所述第一样本人群消费画像信息和所述第二样本人群消费画像信息中人群消费画像信息最大的样本人群消费画像信息作为模型的待修正数据,剩余的样本人群消费画像信息作为所述模型的特征数据,所述第一目标联邦学习模型用于建立所述特征数据和所述待修正数据之间的回归关系。
在本申请的一个实施例中,所述第一训练单元,具体用于:基于所述第一人群消费画像信息,控制所述第一数据端上的本地学习模型进行训练,以得到中间结果;基于所述第二人群消费画像信息,控制所述其他第一数据端对其自身上的本地学习模型进行训练,以得到中间结果;获取各个所述第一数据端上的本地学习模型每次训练输出的中间结果,并将每次输出的所述中间结果发送给协调方进行汇总;接收所述协调方发送的每次汇总出的全局中间结果;基于所述全局中间结果调整所述本地学习模型的模型参数并继续下一轮训练,直至满足预设条件停止训练得到所述第一目标联邦学习模型。
在本申请的一个实施例中,所述装置还包括第二训练模块,所述第二训练模块,包括:第四获取单元,用于获取所述第二数据端上的样本区域在样本时刻的第一样本人流量数据;第二训练单元,用于协同其他第二数据端上的所述样本区域在所述样本 时刻的第二样本人流量数据进行联邦学习模型的训练,以生成所述第二目标联邦学习模型,其中,选取所述第一样本人流量数据和所述第二样本人流量数据中人流量数据最大的样本人流量数据作为模型的待修正数据,剩余的样本人流量数据作为所述模型的特征数据,所述第二目标联邦学习模型用于建立所述特征数据和所述待修正数据之间的回归关系。
在本申请的一个实施例中,所述第二训练单元,具体用于:基于所述第一样本人流量数据,控制所述第二数据端上的本地学习模型进行训练,以得到中间结果;基于所述第二样本人流量数据,控制所述其他第二数据端对其自身上的本地学习模型进行训练,以得到中间结果;获取各个所述第二数据端上的本地学习模型每次训练输出的中间结果,并将每次输出的所述中间结果发送给协调方进行汇总;接收所述协调方发送的每次汇总出的全局中间结果;基于所述全局中间结果调整所述本地学习模型的模型参数并继续下一轮训练,直至满足预设条件停止训练得到所述第二目标联邦学习模型。
本申请实施例的业务数据处理装置,在接收到业务数据请求后,根据业务数据请求中的目标区域的标识,获取第一数据端上的目标区域在目标时刻的第一人群消费画像信息,并协同其他第一数据端上的目标区域在目标时刻的第二人群消费画像信息,通过第一目标联邦学习模型对第一人群消费画像信息进行修正,以得到目标区域在目标时刻的目标人群消费画像信息,然后,结合目标人群消费画像信息以及该目标区域在目标时刻的目标人流量数据,确定目标区域在目标时刻的消费人群特征分布信息,以及根据消费人群特征分布信息对业务数据请求所请求的业务进行处理。由此,基于人流量数据以及结合多个数据端所确定出的目标人群消费画像信息准确实现了对业务数据请求所请求的业务进行准确处理,提高了业务处理的准确率。
本申请另一方面实施例提出了一种电子设备,包括:一种电子设备,包括:存储器,处理器;所述存储器中存储有计算机指令,当所述计算机指令被所述处理器执行时,实现本申请实施例的业务数据处理方法。
本申请另一方面实施例提出了一种存储有计算机指令的非瞬时计算机可读存储介质,所述计算机指令用于使所述计算机执行本申请实施例公开的业务数据处理方法。
本申请另一方面实施例提出了一种计算机程序产品,当所述计算机程序产品中的指令处理器执行时实现本申请实施例中的业务数据处理方法。
上述可选方式所具有的其他效果将在下文中结合具体实施例加以说明。
附图说明
附图用于更好地理解本方案,不构成对本申请的限定。其中:
图1是根据本申请一个实施例的业务数据处理方法的流程示意图。
图2是训练第一目标联邦学习模型的流程示意图。
图3是步骤202的细化流程示意图。
图4是第二目标联邦学习模型的流程示意图。
图5是步骤402的细化流程示意图。
图6是业务装置中各层之间的关系示例图。
图7是根据本申请一个实施例的业务数据处理装置的结构示意图。
图8是根据本申请另一个实施例的业务数据处理装置的结构示意图。
图9是根据本申请一个实施例的电子设备的框图。
具体实施方式
下面详细描述本发明的实施例,实施例的示例在附图中示出,其中自始至终相同或类似的标号表示相同或类似的元件或具有相同或类似功能的元件。下面通过参考附图描述的实施例是示例性的,旨在用于解释本申请,而不能理解为对本申请的限制。
下面参考附图描述本申请实施例的业务数据处理方法、装置和电子设备。
图1是根据本申请一个实施例的业务数据处理方法的流程示意图。其中,需要说明的是,本实施例提供的业务数据处理方法的执行主体为业务数据处理装置,该业务数据处理装置可以由软件和/或硬件的方式实现,该实施例中的业务数据处理装置可以配置在电子设备中,本实施例中的电子设备可以包括终端设备和服务器等设备。
如图1所示,该业务数据处理方法可以包括:
步骤101,接收业务数据请求,其中,业务数据请求包括目标区域的标识以及目标时刻。
在本实施例中,在有业务需求时,可基于业务客户端或者业务网站向业务数据处理装置提出业务数据请求。
步骤102,根据目标区域的标识,获取第一数据端上的目标区域在目标时刻的第一人群消费画像信息。
其中,第一人群消费画像信息是指通过对第一数据端上所记录的对应用户消费行为数据进行分析,而抽象出的标签化画像。其中,上述对应用户消费行为数据为用户在目标区域在目标时刻上的消费行为数据。在本申请的一个实施例中,上述根据目标区域的标识,获取第一数据端上的目标区域在目标时刻的第一人群消费画像信息,可以通过多种方式实 现,举例说明如下:
作为一种可能的实现方式,为了可以准确确定出目标区域在目标时刻上的第一人群消费画像信息,可根据目标区域的标识,获取在第二预设时长内目标区域在每个历史时刻的人群消费画像信息,根据每个历史时刻的人群消费画像信息,确定第一数据端上的目标区域在目标时刻的第一人群消费画像信息。也就是说,目标区域在目标时刻上的第一人群消费画像信息可通过历史一段时间内的平均画像得出。
具体而言,上述人群消费画像信息可以是基于第一数据端上记录的用户购买行为数据,可以根据用户的历史购买行为刻画用户的一些消费特征,例如:用户对商品的消费偏好,用户的消费水平等。并且,第一数据端自身也储存的有用户的个人画像,例如,用户的年龄、学历、有无车等。因此,通过聚合区域内各个用户的消费特征及个人画像,可以形成区域在不同时刻上的人群消费画像信息。
作为另一种可能的实现方式,可通过第一数据端上预先保存的各个区域在各个时刻上的人群消费画像信息数据库,基于目标区域的标识,获取第一数据端上在目标时刻的第一人群消费画像信息。
步骤103,协同其他第一数据端上的目标区域在目标时刻的第二人群消费画像信息,通过第一目标联邦学习模型对第一人群消费画像信息进行修正,以得到目标区域在目标时刻的目标人群消费画像信息。
在本实施例中,通过第一目标联邦学习模型结合第一数据端以及其他第一数据端上的人群消费画像信息进行修正,从而可通过多个不同数据端所得到的人群消费画像信息,准确刻画出目标区域在目标时刻上的人群消费画像。方便后续基于该目标区域在目标时刻上的人群消费画像准确提供业务处理。
其中,需要说明的是,本实施例中其他第一数据端上确定目标区域在目标时刻上的第二人群消费画像信息的方式,与上述第一数据端确定目标区域在目标时刻上的第一人群消费画像信息的方式类似,此处不再赘述。
其中,本实施例中的第一数据端和第二数据端是不同的数据端,其均可以对用户消费行为数据进行统计。在一些实施例中,上述第一数据端和第二数据端其所提供的业务服务可以是相同的。
其中,第二人群消费画像信息是指通过对第二数据端上所记录的对应用户消费行为数据进行分析,而抽象出的标签化画像。其中,上述对应用户消费行为数据为用户在目标区域在目标时刻上的消费行为数据。
其中,需要说明的是,上述第一目标联邦学习模型是基于第一数据端以及其他数据端 的人群消费画像信息进行训练而得到的。关于第一目标联邦学习模型的训练过程将在后续实施例描述。
步骤104,获取目标区域在目标时刻的目标人流量数据。
其中,需要说明的是,在不同应用场景中,上述获取目标区域在目标时刻的目标人流量数据的方式不同,举例说明如下:
作为一种可能的实现方式,可获取目标区域在预设时长内各个历史时刻的人流量数据,然后,基于预先训练好的人流量预测模型,来确定目标区域在目标时刻上的目标人流量数据。
例如,在某一区域的某一时刻t,可获取该区域在各个历史时刻(t-n,…,t-1)上的人流量数据,然后,将所获取的各个历史时刻上的人流量数据,输入到预先训练好的人流量预测模型,以预测得到该目标区域在t时刻的人流量数据。
其中,上述各个区域在各个历史时刻上的人流量数据的一种可能实现方式为:可基于多个电信运营商平台所记录的用户的移动轨迹数据,并基于用户的移动轨迹数据聚合起来,以形成不同区域,不同时刻的人流量数据。
其中,上述人流量预测模型是预先训练得到的,其中,为了使得人流量预测模型可准确预测出目标区域在对应目标时刻上的人流量数据。在训练人流量预测模型时,可基于利用人流量信息准确的地块区域训练该人流量预测模型。并在训练完成后,可基于训练好的人流量预测模型重新预估可信度低的区域的人流量数据,以达到修正效果。
在本申请的一个实施例中,为了可以准确预测出目标区域在目标时刻的目标人流量数据,获取第二数据端上的目标区域在目标时刻的第一人流量数据;协同其他第二数据端上的目标区域在目标时刻的第二人流量数据,通过第二目标联邦学习模型对第一人流量数据进行修正,以获取目标区域在目标时刻的目标人流量数据。
在本申请的一个实施例中,为了可以准确获取第二数据端上的目标区域在目标时刻的第一人流量数据,获取第二数据端上的目标区域在目标时刻的第一人流量数据的一种可能实现方式为:可获取在第一预设时长内目标区域在每个历史目标时刻的人流量数据;基于每个历史目标时刻的人流量数据,确定第二数据端上的目标区域在目标时刻的第一人流量数据。
其中,第一预设时长是预先设置的时长,例如,第一预设时长可以为一个月,或者,一个星期,在实际应用中,可基于实际业务需求确定第一预设时长的取值,该实施例对此不作具体限定。
在本申请的一个实施例中,为了可以准确预测出目标区域在目标时刻上的第一人流量 数据,可将每个历史目标时刻的人流量数据输入到与第二数据端所对应的人流量预测模型,以通过该人流量预测模型确定第二数据端上的目标区域在目标时刻的第一人流量数据。
步骤105,根据目标人流量数据和目标人群消费画像信息,确定目标区域在目标时刻的消费人群特征分布信息。
步骤106,根据消费人群特征分布信息对业务数据请求所请求的业务进行处理。
在本申请的一个示例性的实施方式,在上述业务数据请求为商铺选址业务时,可基于该目标区域在目标时刻的消费人群特征分布信息进行准确的商铺选址,并将商铺选址结果提供给业务数据请求方。
在本申请的另一个示例性的实施方式,在上述业务数据请求为广告投放业务时,可基于该目标区域在目标时刻的消费人群特征信息,确定该目标区域在目标时对广告所对应的商品的偏好占比,并结合商品的偏好占比进行广告投放。
本申请实施例的业务数据处理方法,在接收到业务数据请求后,根据业务数据请求中的目标区域的标识,获取第一数据端上的目标区域在目标时刻的第一人群消费画像信息,并协同其他第一数据端上的目标区域在目标时刻的第二人群消费画像信息,通过第一目标联邦学习模型对第一人群消费画像信息进行修正,以得到目标区域在目标时刻的目标人群消费画像信息,然后,结合目标人群消费画像信息以及该目标区域在目标时刻的目标人流量数据,确定目标区域在目标时刻的消费人群特征分布信息,以及根据消费人群特征分布信息对业务数据请求所请求的业务进行处理。由此,基于人流量数据以及结合多个数据端所确定出的目标人群消费画像信息准确实现了对业务数据请求所请求的业务进行准确处理,提高了业务处理的准确率。
在本申请的一个实施例中,为了使得第一目标联邦学习模型可准确对第一人群消费画像信息进行修正,可联合多个第一数据端上的数据,对第一目标联邦学习模型进行训练。
如图2所示,训练第一目标联邦学习模型的过程,可以包括:
步骤201,获取第一数据端上的样本区域在样本时刻的第一样本人群消费画像信息。
其中,样本区域在样本时刻的第一样本人群消费画像信息可以预先收集。
步骤202,协同其他第一数据端上的样本区域在样本时刻的第二样本人群消费画像信息进行联邦学习模型的训练,以生成第一目标联邦学习模型。
其中,选取第一样本人群消费画像信息和第二样本人群消费画像信息中人群消费画像信息最大的样本人群消费画像信息作为模型的待修正数据,剩余的样本人群消费画像信息作为模型的特征数据,第一目标联邦学习模型用于建立特征数据和待修正数据之间的回归关系。
为了可以对各个数据端上提供的数据进行保护的同时,又可以训练出第一目标联邦学习模型,在本申请的一个实施例中,上述步骤202的一种可能实现方式,如图3所示,可以包括:
步骤301,基于第一人群消费画像信息,控制第一数据端上的本地学习模型进行训练,以得到中间结果。
步骤302,基于第二人群消费画像信息,控制其他第一数据端对其自身上的本地学习模型进行训练,以得到中间结果。
步骤303,获取各个第一数据端上的本地学习模型每次训练输出的中间结果,并将每次输出的中间结果发送给协调方进行汇总。
步骤304,接收协调方发送的每次汇总出的全局中间结果。
步骤305,基于全局中间结果调整本地学习模型的模型参数并继续下一轮训练,直至满足预设条件停止训练得到第一目标联邦学习模型。
在本实施例中,通过联邦学习的方式使得各个数据端上的数据在不出库的情况下,可准确训练出第一目标联邦学习模型,方便了后续基于训练出的第一目标联邦学习模型进行人流量修正。
在本申请的一个实施例中,为了通过第二目标联邦学习模型可以准确确定出对于区域在目标时刻上的人流量数据,可结合多个第二数据端上的样本人流量数据对第二目标联邦学习模型进行训练。如图4所示,第二目标联邦学习模型的训练过程可以包括:
步骤401,获取第二数据端上的样本区域在样本时刻的第一样本人流量数据。
步骤402,协同其他第二数据端上的样本区域在样本时刻的第二样本人流量数据进行联邦学习模型的训练,以生成第二目标联邦学习模型。
其中,选取第一样本人流量数据和第二样本人流量数据中人流量数据最大的样本人流量数据作为模型的待修正数据,剩余的样本人流量数据作为模型的特征数据,第二目标联邦学习模型用于建立特征数据和待修正数据之间的回归关系。
为了可以对各个数据端上提供的数据进行保护的同时,又可以训练出第一目标联邦学习模型,在本申请的一个实施例中,如图5所示,上述步骤402的一种可能实现方式为:
步骤501,基于第一样本人流量数据,控制第二数据端上的本地学习模型进行训练,以得到中间结果。
步骤502,基于第二样本人流量数据,控制其他第二数据端对其自身上的本地学习模型进行训练,以得到中间结果。
步骤503,获取各个第二数据端上的本地学习模型每次训练输出的中间结果,并将每 次输出的中间结果发送给协调方进行汇总。
步骤504,接收协调方发送的每次汇总出的全局中间结果。
步骤505,基于全局中间结果调整本地学习模型的模型参数并继续下一轮训练,直至满足预设条件停止训练得到第二目标联邦学习模型。
本实施例中,通过联合学习的方式,在多个数据端上的数据不出库的情况下,准确训练出了第二目标联邦学习模型,方便了后续基于训练出的第二目标联邦学习模型准确确定出对应区域在对应时刻上的目标人流量数据。
基于上述实施例的基础上,为了使得本领域的技术人员可以清楚了解本申请,下面结合图6对该实施例的业务数据处理方法进行示例性描述。如图6所示。
图中的数据端A,结合人流轨迹以及城市路网进行人流轨迹数据处理,以得到区域人口信息,并通过数据端B,对数据端B上的人口特征进行人群画像数据处理,以得到区域群像信息。
另外,图中的数据端C,结合人流轨迹以及兴趣点(POI,Point of Information)进行人流轨迹数据处理,以得到区域人口信息,并通过数据端B,对数据端D上的人口特征进行人群画像数据处理,以得到区域群像信息。
然后,在计算层,可结合其他端上的区域画像对本端上的区域画像进行修正。另外,可结合其他端上的区域人流量对本端上的人流量进行修正。
然后,可跨域人群多维度实时计算,以得到区域人流量画像、区域人群消费画像,区域能源画像,以及区域安全画像。
然后,在应用层,可基于知识层所得到区域画像信息进行应用。
知识层的信息可以服务于应用层,以指导商铺选址、广告精准投放、城市交通规划、消防站规划等应用。
具体地,结合知识层输出的区域人流量、人群消费、消费偏好等画像,可以协助商家进行商铺选址;根据区域人群消费画像及人流量数据,可以精准地获取某类商品的偏好人数,进而预估广告转化率及投放收益;基于实时的区域人流量画像,可以协助政府规划交通,推荐用户合理的出行方式;根据城市区域安全画像,可以协助规划城市消防站建设,以提升城市安全系数等。
与上述几种实施例提供的业务数据处理方法相对应,本申请的一种实施例还提供一种业务数据处理装置,由于本申请实施例提供的业务数据处理装置与上述几种实施例提供的业务数据处理方法相对应,因此在业务数据处理方法的实施方式也适用于本实施例提供的业务数据处理装置,在本实施例中不再详细描述。
图7是根据本申请一个实施例的业务数据处理装置的结构示意图。
如图7所示,该业务数据处理装置700包括:
接收模块701,用于接收业务数据请求,其中,业务数据请求包括目标区域的标识以及目标时刻。
第一获取模块702,用于根据目标区域的标识,获取第一数据端上的目标区域在目标时刻的第一人群消费画像信息。
修正模块703,用于协同其他第一数据端上的目标区域在目标时刻的第二人群消费画像信息,通过第一目标联邦学习模型对第一人群消费画像信息进行修正,以得到目标区域在目标时刻的目标人群消费画像信息。
第二获取模块704,获取目标区域在目标时刻的目标人流量数据。
确定模块705,用于根据目标人流量数据和目标人群消费画像信息,确定目标区域在目标时刻的消费人群特征分布信息。
业务处理模块706,用于根据消费人群特征分布信息对业务数据请求所请求的业务进行处理。
在本申请的一个实施例中,在图7所示的装置实施例的基础上,如图8所示,该第二获取模块704,可以包括:
第一获取单元7041,用于获取第二数据端上的目标区域在目标时刻的第一人流量数据;
修正单元7042,用于协同其他第二数据端上的目标区域在目标时刻的第二人流量数据,通过第二目标联邦学习模型对第一人流量数据进行修正,以获取目标区域在目标时刻的目标人流量数据。
在本申请的一个实施例中,第一获取单元7041,具体用于:获取在第一预设时长内目标区域在每个历史目标时刻的人流量数据;基于每个历史目标时刻的人流量数据,确定第二数据端上的目标区域在目标时刻的第一人流量数据。
在本申请的一个实施例中,上述第一获取模块702,可以包括:
第二获取单元7021,用于根据目标区域的标识,获取在第二预设时长内目标区域在每个历史时刻的人群消费画像信息;
确定单元7022,用于根据每个历史时刻的人群消费画像信息,确定第一数据端上的目标区域在目标时刻的第一人群消费画像信息。
在本申请的一个实施例中,如图8所示,该装置还包括第一训练模块707,第一训练 模块707,可以包括:
第三获取单元7071,用于获取第一数据端上的样本区域在样本时刻的第一样本人群消费画像信息;
第一训练单元7072,用于协同其他第一数据端上的样本区域在样本时刻的第二样本人群消费画像信息进行联邦学习模型的训练,以生成第一目标联邦学习模型,其中,选取第一样本人群消费画像信息和第二样本人群消费画像信息中人群消费画像信息最大的样本人群消费画像信息作为模型的待修正数据,剩余的样本人群消费画像信息作为模型的特征数据,第一目标联邦学习模型用于建立特征数据和待修正数据之间的回归关系。
在本申请的一个实施例中,上述第一训练单元7072,具体用于:基于第一人群消费画像信息,控制第一数据端上的本地学习模型进行训练,以得到中间结果;基于第二人群消费画像信息,控制其他第一数据端对其自身上的本地学习模型进行训练,以得到中间结果;获取各个第一数据端上的本地学习模型每次训练输出的中间结果,并将每次输出的中间结果发送给协调方进行汇总;接收协调方发送的每次汇总出的全局中间结果;基于全局中间结果调整本地学习模型的模型参数并继续下一轮训练,直至满足预设条件停止训练得到第一目标联邦学习模型。
在本申请的一个实施例中,如图8所示,该装置还包括第二训练模块708,第二训练模块708,包括:
第四获取单元7081,用于获取第二数据端上的样本区域在样本时刻的第一样本人流量数据;
第二训练单元7082,用于协同其他第二数据端上的样本区域在样本时刻的第二样本人流量数据进行联邦学习模型的训练,以生成第二目标联邦学习模型,其中,选取第一样本人流量数据和第二样本人流量数据中人流量数据最大的样本人流量数据作为模型的待修正数据,剩余的样本人流量数据作为模型的特征数据,第二目标联邦学习模型用于建立特征数据和待修正数据之间的回归关系。
在本申请的一个实施例中,上述第二训练单元7082,具体用于:基于第一样本人流量数据,控制第二数据端上的本地学习模型进行训练,以得到中间结果;基于第二样本人流量数据,控制其他第二数据端对其自身上的本地学习模型进行训练,以得到中间结果;获取各个第二数据端上的本地学习模型每次训练输出的中间结果,并将每次输出的中间结果发送给协调方进行汇总;接收协调方发送的每次汇总出的全局中间结果;基于全局中间结果调整本地学习模型的模型参数并继续下一轮训练,直至满足预设条件停止训练得到第二 目标联邦学习模型。
本申请实施例的业务数据处理装置,在接收到业务数据请求后,根据业务数据请求中的目标区域的标识,获取第一数据端上的目标区域在目标时刻的第一人群消费画像信息,并协同其他第一数据端上的目标区域在目标时刻的第二人群消费画像信息,通过第一目标联邦学习模型对第一人群消费画像信息进行修正,以得到目标区域在目标时刻的目标人群消费画像信息,然后,结合目标人群消费画像信息以及该目标区域在目标时刻的目标人流量数据,确定目标区域在目标时刻的消费人群特征分布信息,以及根据消费人群特征分布信息对业务数据请求所请求的业务进行处理。由此,基于人流量数据以及结合多个数据端所确定出的目标人群消费画像信息准确实现了对业务数据请求所请求的业务进行准确处理,提高了业务处理的准确率。
根据本申请的实施例,本申请还提供了一种电子设备和一种可读存储介质。
如图9所示,是根据本申请一个实施例的电子设备的框图。
如图9所示,该电子设备该电子设备包括:
存储器901、处理器902及存储在存储器901上并可在处理器902上运行的计算机指令。
处理器902执行指令时实现上述实施例中提供的业务数据处理方法。
进一步地,电子设备还包括:
通信接口903,用于存储器901和处理器902之间的通信。
存储器901,用于存放可在处理器902上运行的计算机指令。
存储器901可能包含高速RAM存储器,也可能还包括非易失性存储器(non-volatile memory),例如至少一个磁盘存储器。
处理器902,用于执行程序时实现上述实施例的业务数据处理方法。
如果存储器901、处理器902和通信接口903独立实现,则通信接口903、存储器901和处理器902可以通过总线相互连接并完成相互间的通信。总线可以是工业标准体系结构(Industry Standard Architecture,简称为ISA)总线、外部设备互连(Peripheral Component,简称为PCI)总线或扩展工业标准体系结构(Extended Industry Standard Architecture,简称为EISA)总线等。总线可以分为地址总线、数据总线、控制总线等。为便于表示,图9中仅用一条粗线表示,但并不表示仅有一根总线或一种类型的总线。
可选的,在具体实现上,如果存储器901、处理器902及通信接口903,集成在一块芯片上实现,则存储器901、处理器902及通信接口903可以通过内部接口完成相互间的通信。
处理器902可能是一个中央处理器(Central Processing Unit,简称为CPU),或者是特定集成电路(Application Specific Integrated Circuit,简称为ASIC),或者是被配置成实施本申请实施例的一个或多个集成电路。
本申请还提出一种计算机程序产品,当计算机程序产品中的指令处理器执行时实现本申请实施例的业务数据处理方法。
在本说明书的描述中,参考术语“一个实施例”、“一些实施例”、“示例”、“具体示例”、或“一些示例”等的描述意指结合该实施例或示例描述的具体特征、结构、材料或者特点包含于本申请的至少一个实施例或示例中。在本说明书中,对上述术语的示意性表述不必须针对的是相同的实施例或示例。而且,描述的具体特征、结构、材料或者特点可以在任一个或多个实施例或示例中以合适的方式结合。此外,在不相互矛盾的情况下,本领域的技术人员可以将本说明书中描述的不同实施例或示例以及不同实施例或示例的特征进行结合和组合。
此外,术语“第一”、“第二”仅用于描述目的,而不能理解为指示或暗示相对重要性或者隐含指明所指示的技术特征的数量。由此,限定有“第一”、“第二”的特征可以明示或者隐含地包括至少一个该特征。在本申请的描述中,“多个”的含义是至少两个,例如两个,三个等,除非另有明确具体的限定。
流程图中或在此以其他方式描述的任何过程或方法描述可以被理解为,表示包括一个或更多个用于实现定制逻辑功能或过程的步骤的可执行指令的代码的模块、片段或部分,并且本申请的优选实施方式的范围包括另外的实现,其中可以不按所示出或讨论的顺序,包括根据所涉及的功能按基本同时的方式或按相反的顺序,来执行功能,这应被本申请的实施例所属技术领域的技术人员所理解。
在流程图中表示或在此以其他方式描述的逻辑和/或步骤,例如,可以被认为是用于实现逻辑功能的可执行指令的定序列表,可以具体实现在任何计算机可读介质中,以供指令执行***、装置或设备(如基于计算机的***、包括处理器的***或其他可以从指令执行***、装置或设备取指令并执行指令的***)使用,或结合这些指令执行***、装置或设备而使用。就本说明书而言,"计算机可读介质"可以是任何可以包含、存储、通信、传播或传输程序以供指令执行***、装置或设备或结合这些指令执行***、装置或设备而使用的装置。计算机可读介质的更具体的示例(非穷尽性列表)包括以下:具有一个或多个布线的电连接部(电子装置),便携式计算机盘盒(磁装置),随机存取存储器(RAM),只读存储器(ROM),可擦除可编辑只读存储器(EPROM或闪速存储器),光纤装置,以及便携式光盘只读存储器(CDROM)。另外,计算机可读介质甚至可以是可在其上打印程序 的纸或其他合适的介质,因为可以例如通过对纸或其他介质进行光学扫描,接着进行编辑、解译或必要时以其他合适方式进行处理来以电子方式获得程序,然后将其存储在计算机存储器中。
应当理解,本申请的各部分可以用硬件、软件、固件或它们的组合来实现。在上述实施方式中,多个步骤或方法可以用存储在存储器中且由合适的指令执行***执行的软件或固件来实现。如,如果用硬件来实现和在另一实施方式中一样,可用本领域公知的下列技术中的任一项或他们的组合来实现:具有用于对数据信号实现逻辑功能的逻辑门电路的离散逻辑电路,具有合适的组合逻辑门电路的专用集成电路,可编程门阵列(PGA),现场可编程门阵列(FPGA)等。
本技术领域的普通技术人员可以理解实现上述实施例方法携带的全部或部分步骤是可以通过程序来指令相关的硬件完成,的程序可以存储于一种计算机可读存储介质中,该程序在执行时,包括方法实施例的步骤之一或其组合。
此外,在本申请各个实施例中的各功能单元可以集成在一个处理模块中,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模块中。上述集成的模块既可以采用硬件的形式实现,也可以采用软件功能模块的形式实现。集成的模块如果以软件功能模块的形式实现并作为独立的产品销售或使用时,也可以存储在一个计算机可读取存储介质中。
上述提到的存储介质可以是只读存储器,磁盘或光盘等。尽管上面已经示出和描述了本申请的实施例,可以理解的是,上述实施例是示例性的,不能理解为对本申请的限制,本领域的普通技术人员在本申请的范围内可以对上述实施例进行变化、修改、替换和变型。

Claims (18)

  1. 一种业务数据处理方法,其特征在于,包括:
    接收业务数据请求,其中,所述业务数据请求包括目标区域的标识以及目标时刻;
    根据所述目标区域的标识,获取第一数据端上的目标区域在所述目标时刻的第一人群消费画像信息;
    协同其他第一数据端上的所述目标区域在所述目标时刻的第二人群消费画像信息,通过第一目标联邦学习模型对所述第一人群消费画像信息进行修正,以得到所述目标区域在所述目标时刻的目标人群消费画像信息;
    获取所述目标区域在目标时刻的目标人流量数据;
    根据所述目标人流量数据和所述目标人群消费画像信息,确定所述目标区域在所述目标时刻的消费人群特征分布信息;
    根据所述消费人群特征分布信息对所述业务数据请求所请求的业务进行处理。
  2. 如权利要求1所述的方法,其特征在于,所述获取所述目标区域在目标时刻的目标人流量数据,包括:
    获取第二数据端上的所述目标区域在目标时刻的第一人流量数据;
    协同其他第二数据端上的所述目标区域在所述目标时刻的第二人流量数据,通过第二目标联邦学习模型对所述第一人流量数据进行修正,以获取所述目标区域在所述目标时刻的目标人流量数据。
  3. 如权利要求2所述的方法,其特征在于,所述获取第二数据端上的所述目标区域在目标时刻的第一人流量数据,包括:
    获取在第一预设时长内所述目标区域在每个历史目标时刻的人流量数据;
    基于每个所述历史目标时刻的人流量数据,确定所述第二数据端上的所述目标区域在目标时刻的第一人流量数据。
  4. 根据权利要求1所述的方法,其特征在于,所述根据所述目标区域的标识,获取第一数据端上的目标区域在所述目标时刻的第一人群消费画像信息,包括:
    根据所述目标区域的标识,获取在第二预设时长内所述目标区域在每个历史时刻的人群消费画像信息;
    根据每个所述历史时刻的人群消费画像信息,确定所述第一数据端上的所述目标区域在目标时刻的第一人群消费画像信息。
  5. 根据权利要求1所述的方法,其特征在于,所述第一目标联邦学习模型的训练过程,包括:
    获取所述第一数据端上的样本区域在样本时刻的第一样本人群消费画像信息;
    协同其他第一数据端上的所述样本区域在所述样本时刻的第二样本人群消费画像信息进行联邦学习模型的训练,以生成所述第一目标联邦学习模型,其中,选取所述第一样本人群消费画像信息和所述第二样本人群消费画像信息中人群消费画像信息最大的样本人群消费画像信息作为模型的待修正数据,剩余的样本人群消费画像信息作为所述模型的特征数据,所述第一目标联邦学习模型用于建立所述特征数据和所述待修正数据之间的回归关系。
  6. 根据权利要求5所述的方法,其特征在于,所述协同其他第一数据端上的所述样本区域在所述样本时刻的第二样本人群消费画像信息进行联邦学习模型的训练,以生成所述第一目标联邦学习模型,包括:
    基于所述第一人群消费画像信息,控制所述第一数据端上的本地学习模型进行训练,以得到中间结果;
    基于所述第二人群消费画像信息,控制所述其他第一数据端对其自身上的本地学习模型进行训练,以得到中间结果;
    获取各个所述第一数据端上的本地学习模型每次训练输出的中间结果,并将每次输出的所述中间结果发送给协调方进行汇总;
    接收所述协调方发送的每次汇总出的全局中间结果;
    基于所述全局中间结果调整所述本地学习模型的模型参数并继续下一轮训练,直至满足预设条件停止训练得到所述第一目标联邦学习模型。
  7. 根据权利要求2所述的方法,其特征在于,所述第二目标联邦学习模型的训练过程,包括:
    获取所述第二数据端上的样本区域在样本时刻的第一样本人流量数据;
    协同其他第二数据端上的所述样本区域在所述样本时刻的第二样本人流量数据进行联邦学习模型的训练,以生成所述第二目标联邦学习模型,其中,选取所述第一样本人流量 数据和所述第二样本人流量数据中人流量数据最大的样本人流量数据作为模型的待修正数据,剩余的样本人流量数据作为所述模型的特征数据,所述第二目标联邦学习模型用于建立所述特征数据和所述待修正数据之间的回归关系。
  8. 根据权利要求7所述的方法,其特征在于,所述协同其他第二数据端上的所述样本区域在所述样本时刻的第二样本人流量数据进行联邦学习模型的训练,以生成所述第二目标联邦学习模型,包括:
    基于所述第一样本人流量数据,控制所述第二数据端上的本地学习模型进行训练,以得到中间结果;
    基于所述第二样本人流量数据,控制所述其他第二数据端对其自身上的本地学习模型进行训练,以得到中间结果;
    获取各个所述第二数据端上的本地学习模型每次训练输出的中间结果,并将每次输出的所述中间结果发送给协调方进行汇总;
    接收所述协调方发送的每次汇总出的全局中间结果;
    基于所述全局中间结果调整所述本地学习模型的模型参数并继续下一轮训练,直至满足预设条件停止训练得到所述第二目标联邦学习模型。
  9. 一种业务数据处理装置,其特征在于,包括:
    接收模块,用于接收业务数据请求,其中,所述业务数据请求包括目标区域的标识以及目标时刻;
    第一获取模块,用于根据所述目标区域的标识,获取第一数据端上的目标区域在所述目标时刻的第一人群消费画像信息;
    修正模块,用于协同其他第一数据端上的所述目标区域在所述目标时刻的第二人群消费画像信息,通过第一目标联邦学习模型对所述第一人群消费画像信息进行修正,以得到所述目标区域在所述目标时刻的目标人群消费画像信息;
    第二获取模块,获取所述目标区域在目标时刻的目标人流量数据;
    确定模块,用于根据所述目标人流量数据和所述目标人群消费画像信息,确定所述目标区域在所述目标时刻的消费人群特征分布信息;
    业务处理模块,用于根据所述消费人群特征分布信息对所述业务数据请求所请求的业务进行处理。
  10. 如权利要求9所述的装置,其特征在于,所述第二获取模块,包括:
    第一获取单元,用于获取第二数据端上的所述目标区域在目标时刻的第一人流量数据;
    修正单元,用于协同其他第二数据端上的所述目标区域在所述目标时刻的第二人流量数据,通过第二目标联邦学习模型对所述第一人流量数据进行修正,以获取所述目标区域在所述目标时刻的目标人流量数据。
  11. 如权利要求10所述的装置,其特征在于,所述第一获取单元,具体用于:
    获取在第一预设时长内所述目标区域在每个历史目标时刻的人流量数据;
    基于每个所述历史目标时刻的人流量数据,确定所述第二数据端上的所述目标区域在目标时刻的第一人流量数据。
  12. 根据权利要求9所述的装置,其特征在于,所述第一获取模块,包括:
    第二获取单元,用于根据所述目标区域的标识,获取在第二预设时长内所述目标区域在每个历史时刻的人群消费画像信息;
    确定单元,用于根据每个所述历史时刻的人群消费画像信息,确定所述第一数据端上的所述目标区域在目标时刻的第一人群消费画像信息。
  13. 根据权利要求9所述的装置,其特征在于,所述装置还包括第一训练模块,所述第一训练模块,包括:
    第三获取单元,用于获取所述第一数据端上的样本区域在样本时刻的第一样本人群消费画像信息;
    第一训练单元,用于协同其他第一数据端上的所述样本区域在所述样本时刻的第二样本人群消费画像信息进行联邦学习模型的训练,以生成所述第一目标联邦学习模型,其中,选取所述第一样本人群消费画像信息和所述第二样本人群消费画像信息中人群消费画像信息最大的样本人群消费画像信息作为模型的待修正数据,剩余的样本人群消费画像信息作为所述模型的特征数据,所述第一目标联邦学习模型用于建立所述特征数据和所述待修正数据之间的回归关系。
  14. 根据权利要求13所述的装置,其特征在于,所述第一训练单元,具体用于:
    基于所述第一人群消费画像信息,控制所述第一数据端上的本地学习模型进行训练,以得到中间结果;
    基于所述第二人群消费画像信息,控制所述其他第一数据端对其自身上的本地学习模型进行训练,以得到中间结果;
    获取各个所述第一数据端上的本地学习模型每次训练输出的中间结果,并将每次输出的所述中间结果发送给协调方进行汇总;
    接收所述协调方发送的每次汇总出的全局中间结果;
    基于所述全局中间结果调整所述本地学习模型的模型参数并继续下一轮训练,直至满足预设条件停止训练得到所述第一目标联邦学习模型。
  15. 根据权利要求10所述的装置,其特征在于,所述装置还包括第二训练模块,所述第二训练模块,包括:
    第四获取单元,用于获取所述第二数据端上的样本区域在样本时刻的第一样本人流量数据;
    第二训练单元,用于协同其他第二数据端上的所述样本区域在所述样本时刻的第二样本人流量数据进行联邦学习模型的训练,以生成所述第二目标联邦学习模型,其中,选取所述第一样本人流量数据和所述第二样本人流量数据中人流量数据最大的样本人流量数据作为模型的待修正数据,剩余的样本人流量数据作为所述模型的特征数据,所述第二目标联邦学习模型用于建立所述特征数据和所述待修正数据之间的回归关系。
  16. 根据权利要求15所述的装置,其特征在于,所述第二训练单元,具体用于:
    基于所述第一样本人流量数据,控制所述第二数据端上的本地学习模型进行训练,以得到中间结果;
    基于所述第二样本人流量数据,控制所述其他第二数据端对其自身上的本地学习模型进行训练,以得到中间结果;
    获取各个所述第二数据端上的本地学习模型每次训练输出的中间结果,并将每次输出的所述中间结果发送给协调方进行汇总;
    接收所述协调方发送的每次汇总出的全局中间结果;
    基于所述全局中间结果调整所述本地学习模型的模型参数并继续下一轮训练,直至满足预设条件停止训练得到所述第二目标联邦学习模型。
  17. 一种电子设备,包括:存储器,处理器;所述存储器中存储有计算机指令,当所述计算机指令被所述处理器执行时,实现如权利要求1-8中任一项所述的业务数据 处理方法。
  18. 一种存储有计算机指令的非瞬时计算机可读存储介质,其特征在于,所述计算机指令用于使所述计算机执行权利要求1-8中任一项所述的业务数据处理方法。
PCT/CN2021/119158 2020-12-29 2021-09-17 业务数据处理方法、装置、电子设备和存储介质 WO2022142493A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202011597920.8 2020-12-29
CN202011597920.8A CN113723652A (zh) 2020-12-29 2020-12-29 业务数据处理方法、装置、电子设备和存储介质

Publications (1)

Publication Number Publication Date
WO2022142493A1 true WO2022142493A1 (zh) 2022-07-07

Family

ID=78672417

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/119158 WO2022142493A1 (zh) 2020-12-29 2021-09-17 业务数据处理方法、装置、电子设备和存储介质

Country Status (2)

Country Link
CN (1) CN113723652A (zh)
WO (1) WO2022142493A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115587853A (zh) * 2022-10-24 2023-01-10 广州易海创腾信息科技有限公司 电子商务与大数据结合的业务画像分析方法及***
CN116796860A (zh) * 2023-08-24 2023-09-22 腾讯科技(深圳)有限公司 联邦学习方法、装置、电子设备及存储介质
CN117271905A (zh) * 2023-11-21 2023-12-22 杭州小策科技有限公司 基于人群画像的侧向需求分析方法及***

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114282943A (zh) * 2021-12-23 2022-04-05 智道网联科技(北京)有限公司 流量数据的处理方法、处理***、处理装置和电子设备

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2006293610A (ja) * 2005-04-08 2006-10-26 North Plan:Kk 広告料金決定装置及び方法、並びにプログラム
CN109598537A (zh) * 2018-11-06 2019-04-09 阿里巴巴集团控股有限公司 户外广告投放方法及装置
CN110322274A (zh) * 2019-05-30 2019-10-11 深圳壹账通智能科技有限公司 基于数据分析的人群画像生成方法、装置和计算机设备
CN110428276A (zh) * 2019-06-21 2019-11-08 口碑(上海)信息技术有限公司 选取线下广告投放区域的方法、装置和存储介质
CN111695009A (zh) * 2019-03-11 2020-09-22 浙江莲荷科技有限公司 信息展示方法及装置
CN111768241A (zh) * 2020-06-30 2020-10-13 深圳前海微众银行股份有限公司 交互资格等级确定方法、装置、设备及存储介质
CN112035519A (zh) * 2020-08-28 2020-12-04 中国平安人寿保险股份有限公司 用户画像方法、装置、计算机可读存储介质及终端设备

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109829020B (zh) * 2018-12-20 2023-04-07 平安科技(深圳)有限公司 地点资源数据推送方法、装置、计算机设备和存储介质
CN111125512B (zh) * 2019-11-18 2021-04-27 口碑(上海)信息技术有限公司 业务推荐处理方法、装置及***
CN110909102B (zh) * 2019-11-22 2023-02-28 腾讯科技(深圳)有限公司 室内热力图展示方法、装置和计算机可读存储介质
CN111666460A (zh) * 2020-05-27 2020-09-15 中国平安财产保险股份有限公司 基于隐私保护的用户画像生成方法、装置及存储介质

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2006293610A (ja) * 2005-04-08 2006-10-26 North Plan:Kk 広告料金決定装置及び方法、並びにプログラム
CN109598537A (zh) * 2018-11-06 2019-04-09 阿里巴巴集团控股有限公司 户外广告投放方法及装置
CN111695009A (zh) * 2019-03-11 2020-09-22 浙江莲荷科技有限公司 信息展示方法及装置
CN110322274A (zh) * 2019-05-30 2019-10-11 深圳壹账通智能科技有限公司 基于数据分析的人群画像生成方法、装置和计算机设备
CN110428276A (zh) * 2019-06-21 2019-11-08 口碑(上海)信息技术有限公司 选取线下广告投放区域的方法、装置和存储介质
CN111768241A (zh) * 2020-06-30 2020-10-13 深圳前海微众银行股份有限公司 交互资格等级确定方法、装置、设备及存储介质
CN112035519A (zh) * 2020-08-28 2020-12-04 中国平安人寿保险股份有限公司 用户画像方法、装置、计算机可读存储介质及终端设备

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115587853A (zh) * 2022-10-24 2023-01-10 广州易海创腾信息科技有限公司 电子商务与大数据结合的业务画像分析方法及***
CN115587853B (zh) * 2022-10-24 2024-04-05 广州易海创腾信息科技有限公司 电子商务与大数据结合的业务画像分析方法及***
CN116796860A (zh) * 2023-08-24 2023-09-22 腾讯科技(深圳)有限公司 联邦学习方法、装置、电子设备及存储介质
CN116796860B (zh) * 2023-08-24 2023-12-12 腾讯科技(深圳)有限公司 联邦学习方法、装置、电子设备及存储介质
CN117271905A (zh) * 2023-11-21 2023-12-22 杭州小策科技有限公司 基于人群画像的侧向需求分析方法及***
CN117271905B (zh) * 2023-11-21 2024-02-09 杭州小策科技有限公司 基于人群画像的侧向需求分析方法及***

Also Published As

Publication number Publication date
CN113723652A (zh) 2021-11-30

Similar Documents

Publication Publication Date Title
WO2022142493A1 (zh) 业务数据处理方法、装置、电子设备和存储介质
US11627065B2 (en) Selective sensor polling
CN110085225B (zh) 语音交互方法、装置、智能机器人及计算机可读存储介质
US11127020B2 (en) Generating an activity inference model from contextual data
CN110472647B (zh) 基于人工智能的辅助面试方法、装置及存储介质
Peer et al. Train commuters’ scheduling preferences: Evidence from a large-scale peak avoidance experiment
US11056008B2 (en) Real-time service provider progress monitoring
US20130317873A1 (en) Meeting attendance planner
CN106447114A (zh) 提供叫车服务的方法和装置
WO2017143773A1 (zh) 一种众包学习方法及装置
US11663497B2 (en) Facilitating changes to online computing environment by assessing impacts of actions using a knowledge base representation
JP2023505122A (ja) クラウドソーシングされたオンデマンドaiデータの注釈、収集、処理
US20140030681A1 (en) Activity-oriented Studying Method in an Online-to-offline Manner
US20200356781A1 (en) Method and apparatus for performing categorised matching of videos, and selection engine
CN110573837A (zh) 导航方法、装置、存储介质以及服务器
CN112115372B (zh) 一种停车场的推荐方法及装置
CN110648161A (zh) 商业选址的效益评估方法、装置及存储介质
Yang et al. A novel arterial travel time distribution estimation model and its application to energy/emissions estimation
Thodi et al. An analytical approach to real-time bus signal priority system for isolated intersections
Tanaka et al. Experimental study of effects of travel time distribution information on dynamic route choice behavior
CN106844378A (zh) 一种应答方式确定方法和装置
Jin et al. Metro passengers’ route choice model and its application considering perceived transfer threshold
CN111882112A (zh) 一种预测到达时间的方法和***
US10713598B2 (en) Anticipating user dissatisfaction via machine learning
Sun et al. On the tradeoff between sensitivity and specificity in bus bunching prediction

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: 21913239

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21913239

Country of ref document: EP

Kind code of ref document: A1