WO2021151305A1 - 基于缺失数据的样本分析方法、装置、电子设备及介质 - Google Patents

基于缺失数据的样本分析方法、装置、电子设备及介质 Download PDF

Info

Publication number
WO2021151305A1
WO2021151305A1 PCT/CN2020/119092 CN2020119092W WO2021151305A1 WO 2021151305 A1 WO2021151305 A1 WO 2021151305A1 CN 2020119092 W CN2020119092 W CN 2020119092W WO 2021151305 A1 WO2021151305 A1 WO 2021151305A1
Authority
WO
WIPO (PCT)
Prior art keywords
missing data
dimension
missing
saturation
data dimension
Prior art date
Application number
PCT/CN2020/119092
Other languages
English (en)
French (fr)
Inventor
阮晓雯
邓攀
徐亮
肖京
Original Assignee
平安科技(深圳)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 平安科技(深圳)有限公司 filed Critical 平安科技(深圳)有限公司
Publication of WO2021151305A1 publication Critical patent/WO2021151305A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/24Querying
    • G06F16/245Query processing
    • G06F16/2458Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
    • G06F16/2465Query processing support for facilitating data mining operations in structured databases
    • 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

Definitions

  • This application relates to big data technology, and in particular to a sample analysis method, device, electronic device, and computer-readable storage medium based on missing data.
  • the inventor realizes that whether it is the missing feature filling or directly discarding the missing sample, it will cause inaccuracy in sample analysis due to missing data.
  • a sample analysis method based on missing data includes:
  • the application also provides a sample analysis device based on missing data, the device including:
  • the calculation and selection module is used to obtain the missing data set and the corresponding label value, calculate the saturation of each missing data dimension in the missing data set, select the missing data dimension whose saturation is greater than the preset saturation, and generate a feature dimension list ;
  • the calculation and selection module is also used to calculate the correlation coefficient between each missing data dimension in the feature dimension list and the label value, and select the missing data dimension whose correlation coefficient is greater than a preset correlation coefficient to obtain the target missing Data dimension set;
  • the modeling module is used to model the target missing data dimension set by using a preset data missing insensitivity algorithm to generate a missing data insensitive model
  • the analysis module is used to perform data analysis on the sample data set to be analyzed by using the missing data insensitivity model to obtain the analysis result.
  • This application also provides an electronic device, which includes:
  • Memory storing at least one instruction
  • the processor executes the instructions stored in the memory to implement the following steps:
  • the present application also provides a computer-readable storage medium in which at least one instruction is stored, and the at least one instruction is executed by a processor in an electronic device to implement the following steps:
  • FIG. 1 is a schematic flowchart of a sample analysis method based on missing data provided by an embodiment of the application
  • step S3 is a schematic diagram of a detailed implementation flow of step S3 of the data filtering method provided in FIG. 1 in the first embodiment of the application;
  • FIG. 4 is a schematic diagram of modules of a sample analysis device based on missing data provided by an embodiment of the application;
  • FIG. 5 is a schematic diagram of the internal structure of an electronic device for implementing a sample analysis method based on missing data provided by an embodiment of the application;
  • the execution subject of the sample analysis method based on missing data provided in the embodiment of the present application includes but is not limited to at least one of the electronic devices that can be configured to execute the method provided in the embodiment of the present application, such as a server and a terminal.
  • the sample analysis method based on missing data can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform.
  • the server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc.
  • the sample analysis method based on missing data includes:
  • the missing data set is collected based on different business scenarios. For example, for questionnaire surveys, many users are unwilling to answer some private questions and will selectively answer part of the questionnaire. As a result, part of the missing data will be generated in the collection of questionnaires, and the missing data questionnaire will be obtained. According to the questionnaire responses of different users, multiple missing data questionnaires can be obtained, thereby forming a missing data set.
  • the label value refers to the predicted value corresponding to the missing data dimension included in the missing data set.
  • the corresponding label value may be disease, that is, the missing data dimension included in the disease questionnaire
  • the corresponding predictive value is disease.
  • the missing data dimension refers to a concept that characterizes the data characteristics of the missing data set.
  • the missing data dimensions include: age, gender, Weight and height etc.
  • the calculating the saturation of each missing data dimension in the missing data set includes:
  • 500 samples are collected, that is, the answers of 500 users.
  • the preset saturation calculation formula is as follows:
  • p refers to saturation
  • n refers to the number of illegal and/or non-empty samples
  • m refers to the number of samples in the sample set.
  • the preset saturation may be set to 10%, so the present application retains missing feature data with a saturation greater than 10%.
  • the processor before generating the feature dimension list, it may further include: verifying the selected missing data dimensions, and sorting the missing data dimensions that are successfully verified, Generate the feature dimension list.
  • the verification of the missing data dimension is to select the missing data dimension whose dependency between the missing data dimension and the label value is greater than the preset dependency, and to ensure the selected missing data dimension.
  • the reliability of the data dimension is to select the missing data dimension whose dependency between the missing data dimension and the label value is greater than the preset dependency, and to ensure the selected missing data dimension.
  • the currently known Kenall rank correlation coefficient algorithm is used to calculate the dependency of the selected missing data dimension label value.
  • the preset dependency is 0.38.
  • the following method is used to sort the missing data dimensions that have been successfully verified:
  • f * represents the weight of the missing data dimension
  • I represents the unit matrix
  • x i represents the i-th missing data dimension
  • m represents the offset of the missing data dimension
  • is the ranking factor
  • the missing data dimensions are sorted, that is, the larger the weight value, the higher the position of the missing data dimension in the feature dimension list.
  • the correlation coefficient can be understood as the contribution value of the missing data dimension to the label value.
  • the label value is disease.
  • the missing data dimensions include age, gender, weight, height, etc.
  • the contribution value of missing data dimensions such as age, gender, weight, and height to the disease, that is, the degree of correlation between the missing data dimensions such as age, gender, weight, and height, and the disease.
  • the embodiment of the present application uses the following method to calculate the correlation coefficient between each missing data dimension in the feature dimension list and the label value:
  • c(x,y) represents the correlation coefficient between the missing data dimension and the label value
  • COV(x,y) represents the covariance of the missing data dimension and the label value
  • Var[X] represents the variance of the missing data dimension
  • Var[Y ] Is the variance of the label value.
  • this application uses coronary heart disease as the label value, and the corresponding sample data includes cholesterol index, serum protein index, and glomerular index, etc., and age is the missing data dimension.
  • the corresponding sample data is the age range, including ages between 20 and 60 years old.
  • this application selects missing data dimensions whose correlation coefficient is greater than a preset correlation coefficient to obtain the target missing data dimension set, which can filter out some unimportant missing data dimensions in the feature dimension list, and speed up The subsequent model establishment speed that is not sensitive to missing data also ensures the reliability of subsequent model establishment.
  • the preset correlation coefficient is 0.39.
  • the above-mentioned target missing data dimension set may also be stored in a node of a blockchain.
  • the preset data missing insensitivity algorithm refers to a method that can automatically learn the splitting direction of sample data in the missing data dimension, that is, the missing samples are treated as a sparse matrix, and the value of the missing samples is not considered when the node is split.
  • the missing sample data will be divided into the left subtree and the right subtree to calculate the layer loss respectively, and finally the one with the smaller loss is selected.
  • the preset data missing insensitivity algorithm is the currently known XGBoost algorithm
  • the XGBoost algorithm is an improved decision tree algorithm.
  • the S3 includes:
  • the construction principle of the decision tree is: based on the input space where the target missing data dimension set is located, recursively divide each region in the input space into two subregions and determine each subregion. The output value on the region is used to construct a decision tree of the target missing data dimension set.
  • the negative gradient refers to the residual of each target missing data dimension in the target missing data dimension set, and the overall decision is enhanced by fitting the residual of the target missing data dimension Robustness and reliability of the tree.
  • the following method is used to calculate the negative gradient of each missing data dimension in the decision tree:
  • rim represents a negative gradient
  • L(y i ,f(x i ) represents the loss function
  • y i represents the predicted value of the sample data of the i-th missing data dimension
  • f(x i ) represents the true value of the sample data of the i-th missing data dimension Value
  • f(x) represents the area function in the decision tree
  • f m-1(x) represents the area fitting function in the decision tree.
  • the above-mentioned negative gradient is used to continuously train the decision tree until each missing data dimension has been trained on the decision tree, and the missing data is not sensitive to generation Model.
  • Modeling the target missing data dimension set based on the XGBoost algorithm can realize rapid and automatic modeling of input data and target values without domain knowledge, so as to be applied to a variety of different scenarios. This can avoid inaccurate analysis when constructing models due to missing data.
  • the use of the missing data insensitivity model to perform data analysis on the sample data set to be analyzed to obtain the analysis result includes:
  • the sample label value of the sample data set to be analyzed for example, if the sample data set to be analyzed is a user occupation, the sample label value may be salary;
  • the missing data insensitivity model to analyze the sample data set to be analyzed and the sample label value to obtain an analysis result
  • the analysis result can be understood as the correlation between the sample data set to be analyzed and the sample label value
  • the user’s occupation and salary are analyzed using the missing data missing insensitivity model, and the analysis result is that the user’s occupation and salary are closely related.
  • the missing data insensitivity model is used to analyze samples with missing data, and can also be applied to analyze samples without missing data.
  • the embodiment of the present application first obtains the missing data set and the corresponding label value, calculates the saturation of each missing data dimension in the missing data set, selects the missing data dimension whose saturation is greater than the preset saturation, and generates The feature dimension list can understand the contribution of each missing data dimension in the missing data set; secondly, the embodiment of this application calculates the correlation coefficient between each missing data dimension in the feature dimension list and the label value, and selects the The missing data dimension with the correlation coefficient greater than the preset correlation coefficient is obtained, and the target missing data dimension set is obtained, which can filter out some unimportant missing data dimensions in the feature dimension list, speeding up the subsequent establishment of models that are not sensitive to missing data, and also This ensures the reliability of subsequent model establishment; further, the embodiment of the application uses a preset data missing insensitivity algorithm to model the target missing data dimension set, which can realize the input data without domain knowledge. Perform fast and automatic modeling with target values to apply to a variety of different scenarios, so as to avoid inaccurate analysis when building models due to missing data.
  • FIG. 4 it is a functional block diagram of the sample analysis device based on missing data in this application.
  • the sample analysis device 100 based on missing data described in this application can be installed in an electronic device.
  • the sample analysis device based on missing data may include a calculation and selection module 101, a modeling module 102, and an analysis module 103.
  • the module described in the present invention can also be called a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can complete fixed functions, and are stored in the memory of the electronic device.
  • each module/unit is as follows:
  • the calculation and selection module 101 is used to obtain missing data sets and corresponding label values, calculate the saturation of each missing data dimension in the missing data set, select missing data dimensions whose saturation is greater than a preset saturation, and generate List of characteristic dimensions.
  • the missing data set is collected based on different business scenarios. For example, for questionnaire surveys, many users are unwilling to answer some private questions and will selectively answer part of the questionnaire. As a result, part of the missing data will be generated in the collection of questionnaires, and the missing data questionnaire will be obtained. According to the questionnaire responses of different users, multiple missing data questionnaires can be obtained, thereby forming a missing data set.
  • the label value refers to the predicted value corresponding to the missing data dimension included in the missing data set.
  • the corresponding label value may be disease, that is, the missing data dimension included in the disease questionnaire
  • the corresponding predictive value is disease.
  • the missing data dimension refers to a concept that characterizes the data characteristics of the missing data set.
  • the missing data dimensions include: age, gender, Weight and height etc.
  • said calculating the saturation of each missing data dimension in the missing data set includes:
  • Step A Obtain a sample set of the missing data dimension
  • 500 samples are collected, that is, the answers of 500 users.
  • Step B Identify whether there are illegal and/or non-empty samples in the sample set
  • step C after screening the illegal and/or non-empty samples, use a preset saturation calculation formula to calculate the saturation of the sample set after screening, to obtain the missing The saturation of the data dimension;
  • step D calculates the saturation of the sample set using a preset saturation calculation formula to obtain the saturation of the missing data dimension.
  • the preset saturation calculation formula is as follows:
  • p refers to saturation
  • n refers to the number of illegal and/or non-empty samples
  • m refers to the number of samples in the sample set.
  • the preset saturation may be set to 10%, so the present application retains missing feature data with a saturation greater than 10%.
  • the processor before generating the feature dimension list, it may further include: verifying the selected missing data dimensions, and sorting the missing data dimensions that are successfully verified, Generate the feature dimension list.
  • the verification of the missing data dimension is to select the missing data dimension whose dependency between the missing data dimension and the label value is greater than the preset dependency, and to ensure the selected missing data dimension.
  • the reliability of the data dimension is to select the missing data dimension whose dependency between the missing data dimension and the label value is greater than the preset dependency, and to ensure the selected missing data dimension.
  • the currently known Kenall rank correlation coefficient algorithm is used to calculate the dependency of the selected missing data dimension label value.
  • the preset dependency is 0.38.
  • the following method is used to sort the missing data dimensions that have been successfully verified:
  • f * represents the weight of the missing data dimension
  • I represents the unit matrix
  • x i represents the i-th missing data dimension
  • m represents the offset of the missing data dimension
  • is the ranking factor
  • the missing data dimensions are sorted, that is, the larger the weight value, the higher the position of the missing data dimension in the feature dimension list.
  • the calculation and selection module 101 is also used to calculate the correlation coefficient between each missing data dimension in the feature dimension list and the label value, and select the missing data dimension whose correlation coefficient is greater than a preset correlation coefficient to obtain the target Missing data dimension set.
  • the correlation coefficient can be understood as the contribution value of the missing data dimension to the label value.
  • the label value is disease.
  • the missing data dimensions include age, gender, weight, height, etc.
  • the contribution value of missing data dimensions such as age, gender, weight, and height to the disease, that is, the degree of correlation between the missing data dimensions such as age, gender, weight, and height, and the disease.
  • the embodiment of the present application uses the following method to calculate the correlation coefficient between each missing data dimension in the feature dimension list and the label value:
  • c(x,y) represents the correlation coefficient between the missing data dimension and the label value
  • COV(x,y) represents the covariance of the missing data dimension and the label value
  • Var[X] represents the variance of the missing data dimension
  • Var[Y ] Is the variance of the label value.
  • this application uses coronary heart disease as the label value, and the corresponding sample data includes cholesterol index, serum protein index, and glomerular index, etc., and age is the missing data dimension. Then the corresponding sample data is the age range, including ages between 20 and 60 years old. Using the above method to calculate the correlation coefficient between age and coronary disease, the correlation between age and coronary disease can be obtained, which can help users judge the coronary heart disease. Whether the disease is age-related.
  • this application selects missing data dimensions whose correlation coefficient is greater than a preset correlation coefficient to obtain the target missing data dimension set, which can filter out some unimportant missing data dimensions in the feature dimension list, and speed up The subsequent model establishment speed that is not sensitive to missing data also ensures the reliability of subsequent model establishment.
  • the preset correlation coefficient is 0.39.
  • the above-mentioned target missing data dimension set may also be stored in a node of a blockchain.
  • the modeling module 103 is configured to model the target missing data dimensional set using a preset data missing insensitivity algorithm to generate a missing data insensitive model.
  • the preset data missing insensitivity algorithm refers to a method that can automatically learn the splitting direction of the sample data in the missing data dimension, that is, the missing samples are treated as a sparse matrix, and the value of the missing samples is not considered when the node is split.
  • the missing sample data will be divided into the left subtree and the right subtree to calculate the layer loss respectively, and finally the one with the smaller loss is selected.
  • the preset data missing insensitivity algorithm is the currently known XGBoost algorithm
  • the XGBoost algorithm is an improved decision tree algorithm.
  • using a preset data missing insensitivity algorithm to model the target missing data dimension set, and generating a missing data insensitive model includes:
  • the construction principle of the decision tree is: based on the input space where the target missing data dimension set is located, recursively divide each region in the input space into two subregions and determine each subregion. The output value on the region is used to construct a decision tree of the target missing data dimension set.
  • the negative gradient refers to the residual of each target missing data dimension in the target missing data dimension set, and the overall decision is enhanced by fitting the residual of the target missing data dimension Robustness and reliability of the tree.
  • the following method is used to calculate the negative gradient of each missing data dimension in the decision tree:
  • rim represents a negative gradient
  • L(y i ,f(x i ) represents the loss function
  • y i represents the predicted value of the sample data of the i-th missing data dimension
  • f(x i ) represents the true value of the sample data of the i-th missing data dimension Value
  • f(x) represents the area function in the decision tree
  • f m-1(x) represents the area fitting function in the decision tree.
  • the above-mentioned negative gradient is used to continuously train the decision tree until each missing data dimension has been trained on the decision tree, and the missing data is not sensitive to generation Model.
  • Modeling the target missing data dimension set based on the XGBoost algorithm can realize rapid and automatic modeling of input data and target values without domain knowledge, so as to be applied to a variety of different scenarios. This can avoid inaccurate analysis when constructing models due to missing data.
  • the analysis module 104 is configured to use the missing data insensitivity model to perform data analysis on the sample data set to be analyzed to obtain an analysis result.
  • the use of the missing data insensitivity model to perform data analysis on the sample data set to be analyzed to obtain the analysis result includes:
  • the sample label value of the sample data set to be analyzed for example, if the sample data set to be analyzed is a user occupation, the sample label value may be salary;
  • the missing data insensitivity model to analyze the sample data set to be analyzed and the sample label value to obtain an analysis result
  • the analysis result can be understood as the correlation between the sample data set to be analyzed and the sample label value
  • the user’s occupation and salary are analyzed using the missing data missing insensitivity model, and the analysis result is that the user’s occupation and salary are closely related.
  • the missing data insensitivity model is used to analyze samples with missing data, and can also be applied to analyze samples without missing data.
  • the embodiment of the present application first obtains the missing data set and the corresponding label value, calculates the saturation of each missing data dimension in the missing data set, selects the missing data dimension whose saturation is greater than the preset saturation, and generates The feature dimension list can understand the contribution of each missing data dimension in the missing data set; secondly, the embodiment of this application calculates the correlation coefficient between each missing data dimension in the feature dimension list and the label value, and selects the The missing data dimension with the correlation coefficient greater than the preset correlation coefficient is obtained, and the target missing data dimension set is obtained, which can filter out some unimportant missing data dimensions in the feature dimension list, speeding up the subsequent establishment of models that are not sensitive to missing data, and also This ensures the reliability of subsequent model establishment; further, the embodiment of the application uses a preset data missing insensitivity algorithm to model the target missing data dimension set, which can realize the input data without domain knowledge. Perform fast and automatic modeling with target values to apply to a variety of different scenarios, so as to avoid inaccurate analysis when building models due to missing data.
  • FIG. 5 it is a schematic structural diagram of an electronic device implementing a sample analysis method based on missing data in the present application.
  • the electronic device 1 may include a processor 10, a memory 11, and a bus, and may also include a computer program stored in the memory 11 and running on the processor 10, such as a sample analysis program 12 based on missing data.
  • the memory 11 includes at least one type of readable storage medium.
  • the readable storage medium may be non-volatile or volatile.
  • the readable storage medium includes flash memory, mobile hard disk, and multimedia card.
  • Card-type memory for example: SD or DX memory, etc.
  • magnetic memory magnetic disk, optical disk, etc.
  • the memory 11 may be an internal storage unit of the electronic device 1 in some embodiments, for example, a mobile hard disk of the electronic device 1.
  • the memory 11 may also be an external storage device of the electronic device 1, such as a plug-in mobile hard disk, a smart media card (SMC), and a secure digital (Secure Digital) equipped on the electronic device 1.
  • SD Secure Digital
  • flash card Flash Card
  • the memory 11 may also include both an internal storage unit of the electronic device 1 and an external storage device.
  • the memory 11 can be used not only to store application software and various data installed in the electronic device 1, such as codes based on sample analysis of missing data, etc., but also to temporarily store data that has been output or will be output.
  • the processor 10 may be composed of integrated circuits in some embodiments, for example, may be composed of a single packaged integrated circuit, or may be composed of multiple integrated circuits with the same function or different functions, including one or more Combinations of central processing unit (CPU), microprocessor, digital processing chip, graphics processor, and various control chips, etc.
  • the processor 10 is the control unit of the electronic device, which uses various interfaces and lines to connect the various components of the entire electronic device, and runs or executes programs or modules stored in the memory 11 (such as executing Sample analysis based on missing data, etc.), and call data stored in the memory 11 to execute various functions of the electronic device 1 and process data.
  • the bus may be a peripheral component interconnect standard (PCI) bus or an extended industry standard architecture (EISA) bus, etc.
  • PCI peripheral component interconnect standard
  • EISA extended industry standard architecture
  • the bus can be divided into address bus, data bus, control bus and so on.
  • the bus is configured to implement connection and communication between the memory 11 and at least one processor 10 and the like.
  • FIG. 5 only shows an electronic device with components. Those skilled in the art can understand that the structure shown in FIG. 5 does not constitute a limitation on the electronic device 1, and may include fewer or more components than shown in the figure. Components, or a combination of certain components, or different component arrangements.
  • the electronic device 1 may also include a power source (such as a battery) for supplying power to various components.
  • the power source may be logically connected to the at least one processor 10 through a power management device, thereby controlling power
  • the device implements functions such as charge management, discharge management, and power consumption management.
  • the power supply may also include any components such as one or more DC or AC power supplies, recharging devices, power failure detection circuits, power converters or inverters, and power status indicators.
  • the electronic device 1 may also include various sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.
  • the electronic device 1 may also include a network interface.
  • the network interface may include a wired interface and/or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is usually used in the electronic device 1 Establish a communication connection with other electronic devices.
  • the electronic device 1 may also include a user interface.
  • the user interface may be a display (Display) and an input unit (such as a keyboard (Keyboard)).
  • the user interface may also be a standard wired interface or a wireless interface.
  • the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode, organic light-emitting diode) touch device, etc.
  • the display can also be appropriately called a display screen or a display unit, which is used to display the information processed in the electronic device 1 and to display a visualized user interface.
  • the sample analysis 12 based on missing data stored in the memory 11 in the electronic device 1 is a combination of multiple instructions. When running in the processor 10, it can realize:
  • the integrated module/unit of the electronic device 1 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a non-volatile computer readable storage medium.
  • the computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory) .
  • modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the objectives of the solutions of the embodiments.
  • the functional modules in the various embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit.
  • the above-mentioned integrated unit may be implemented in the form of hardware, or may be implemented in the form of hardware plus software functional modules.
  • the blockchain referred to in this application is a new application mode of computer technology such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm.
  • Blockchain essentially a decentralized database, is a series of data blocks associated with cryptographic methods. Each data block contains a batch of network transaction information for verification. The validity of the information (anti-counterfeiting) and the generation of the next block.
  • the blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Software Systems (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Databases & Information Systems (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Fuzzy Systems (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Probability & Statistics with Applications (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computational Linguistics (AREA)
  • Medical Informatics (AREA)
  • Computing Systems (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

一种基于缺失数据的样本分析方法、装置、电子设备以及存储介质,涉及大数据技术,方法包括:获取缺失数据集及对应的标签值,计算所述缺失数据集中每一种缺失数据维度的饱和度,选取饱和度大于预设饱和度的缺失数据维度,生成特征维度列表(S1);计算所述特征维度列表中每一种缺失数据维度与所述标签值的相关系数,选取所述相关系数大于预设相关系数的缺失数据维度,得到目标缺失数据维度集(S2);利用预设数据缺失不敏感算法对所述目标缺失数据维度集进行建模,生成缺失数据不敏感模型(S3);利用所述缺失数据不敏感模型对待分析样本数据集进行数据分析,得到分析结果(S4)。此外,还涉及区块链技术,选取的缺失数据维度可存储于区块链中。通过该方法可以解决因缺失数据导致样本分析出现不准确的问题。

Description

基于缺失数据的样本分析方法、装置、电子设备及介质
本申请要求于2020年7月16日提交中国专利局、申请号为202010684956.3,发明名称为“基于缺失数据的样本分析方法、装置、电子设备及介质”的中国专利申请的优先权,其全部内容通过引用结合在本申请中。
技术领域
本申请涉及大数据技术,尤其涉及一种基于缺失数据的样本分析方法、装置、电子设备及计算机可读存储介质。
背景技术
当前真实世界数据挖掘的一个难点是数据缺失。比如对于基于网络或纸质问卷的数据,常常会有被调查者跳过特定问题的情况,造成回收问卷的回答不完整,此时这一调查样本中会出现特征缺失。
目前无论是对缺失特征进行填充,或者直接将这一缺失样本舍弃,均有各自的缺陷:其中,对于缺失特征填充来说,无法保证填充的数值是否能够真实地反映缺失数值;对于缺失样本舍弃来说,将缺失样本舍弃会造成信息的浪费。
因此,发明人意识到不管是缺失特征填充,或是直接将这一缺失样本舍弃,都会导致因缺失数据进行样本分析出现不准确的现象。
发明内容
本申请提供的一种基于缺失数据的样本分析方法,包括:
获取缺失数据集及对应的标签值,计算所述缺失数据集中每一种缺失数据维度的饱和度,选取饱和度大于预设饱和度的缺失数据维度,生成特征维度列表;
计算所述特征维度列表中每一种缺失数据维度与所述标签值的相关系数,选取所述相关系数大于预设相关系数的缺失数据维度,得到目标缺失数据维度集;
利用预设数据缺失不敏感算法对所述目标缺失数据维度集进行建模,生成缺失数据不敏感模型;
利用所述缺失数据不敏感模型对待分析样本数据集进行数据分析,得到分析结果。
本申请还提供一种基于缺失数据的样本分析装置,所述装置包括:
计算及选取模块,用于获取缺失数据集及对应的标签值,计算所述缺失数据集中每一种缺失数据维度的饱和度,选取饱和度大于预设饱和度的缺失数据维度,生成特征维度列表;
所述计算及选取模块,还用于计算所述特征维度列表中每一种缺失数据维度与所述标签值的相关系数,选取所述相关系数大于预设相关系数的缺失数据维度,得到目标缺失数据维度集;
建模模块,用于利用预设数据缺失不敏感算法对所述目标缺失数据维度集进行建模,生成缺失数据不敏感模型;
分析模块,用于利用所述缺失数据不敏感模型对待分析样本数据集进行数据分析,得到分析结果。
本申请还提供一种电子设备,所述电子设备包括:
存储器,存储至少一个指令;及
处理器,执行所述存储器中存储的指令以实现如下步骤:
获取缺失数据集及对应的标签值,计算所述缺失数据集中每一种缺失数据维度的饱和度,选取饱和度大于预设饱和度的缺失数据维度,生成特征维度列表;
计算所述特征维度列表中每一种缺失数据维度与所述标签值的相关系数,选取所述相关系数大于预设相关系数的缺失数据维度,得到目标缺失数据维度集;
利用预设数据缺失不敏感算法对所述目标缺失数据维度集进行建模,生成缺失数据不敏感模型;
利用所述缺失数据不敏感模型对待分析样本数据集进行数据分析,得到分析结果。
本申请还提供一种计算机可读存储介质,所述计算机可读存储介质中存储有至少一个指令,所述至少一个指令被电子设备中的处理器执行以实现如下步骤:
获取缺失数据集及对应的标签值,计算所述缺失数据集中每一种缺失数据维度的饱和度,选取饱和度大于预设饱和度的缺失数据维度,生成特征维度列表;
计算所述特征维度列表中每一种缺失数据维度与所述标签值的相关系数,选取所述相关系数大于预设相关系数的缺失数据维度,得到目标缺失数据维度集;
利用预设数据缺失不敏感算法对所述目标缺失数据维度集进行建模,生成缺失数据不敏感模型;
利用所述缺失数据不敏感模型对待分析样本数据集进行数据分析,得到分析结果。
附图说明
图1为本申请一实施例提供的基于缺失数据的样本分析方法的流程示意图;
图2为本申请第一实施例中图1提供的数据过滤方法步骤S1的详细实施流程示意图;
图3为本申请第一实施例中图1提供的数据过滤方法步骤S3的详细实施流程示意图;
图4为本申请一实施例提供的基于缺失数据的样本分析装置的模块示意图;
图5为本申请一实施例提供的实现基于缺失数据的样本分析方法的电子设备的内部结构示意图;
本申请目的的实现、功能特点及优点将结合实施例,参照附图做进一步说明。
具体实施方式
应当理解,此处所描述的具体实施例仅仅用以解释本申请,并不用于限定本申请。
本申请实施例提供的基于缺失数据的样本分析方法的执行主体包括但不限于服务端、终端等能够被配置为执行本申请实施例提供的该方法的电子设备中的至少一种。换言之,所述基于缺失数据的样本分析方法可以由安装在终端设备或服务端设备的软件或硬件来执行,所述软件可以是区块链平台。所述服务端包括但不限于:单台服务器、服务器集群、云端服务器或云端服务器集群等。
参照图1所示的本申请一实施例提供的基于缺失数据的样本分析方法的流程示意图。在本申请实施例中,所述基于缺失数据的样本分析方法包括:
S1、获取缺失数据集及对应的标签值,计算所述缺失数据集中每一种缺失数据维度的饱和度,选取饱和度大于预设饱和度的缺失数据维度,生成特征维度列表。
在本申请的至少一个实施例中,所述缺失数据集基于不同的业务场景采集得 到,例如对于问卷调查,许多用户并不愿意回答某些私密性的问题,会选择性的回答问卷中的部分问题,从而导致采集问卷会产生部分的缺失数据,得到缺失数据问卷,根据不同用户的问卷回答,可以得到多个缺失数据问卷,从而形成缺失数据集。
进一步地,所述标签值指的是所述缺失数据集中所包含缺失数据维度对应的预测值,例如关于疾病问卷调查中,其对应的标签值可以为疾病,即疾病问卷调查所包含缺失数据维度对应的预测值为疾病。
在本申请的一个可选实施例中,所述缺失数据维度指的是一种表征缺失数据集数据特征的概念,例如根据疾病问卷调查得到的缺失数据集中,缺失数据维度包括:年龄、性别、体重以及身高等。
较佳地,参阅图2所示,所述计算所述缺失数据集中每一种缺失数据维度的饱和度,包括:
S10、获取所述缺失数据维度的样本集;
例如,对于年龄这个缺失数据维度,采集了500个样本,即500个用户的回答。
S11、识别出所述样本集中是否存在非法和/或非空样本;
例如,查询500个年龄样本中是否存在没有填年龄的样本或者是否年龄样本填的并不是数值而是其它字符。
若存在非法和/或非空样本,执行S12、筛除所述非法和/或非空样本后,利用预设饱和度计算公式计算筛除后所述样本集的饱和度,得到所述缺失数据维度的饱和度;
若不存在非法和/或非空样本,执行S13、利用预设饱和度计算公式计算所述样本集的饱和度,得到所述缺失数据维度的饱和度。
一个可选实施例中,所述预设饱和度计算公式如下所示:
P=1-n/m*100%
其中,p指的是饱和度,n指的是非法和/或非空样本数量,m指的是样本集中样本数量。
优选地,本申请实施例中,所述预设饱和度可以设置为10%,因此本申请保留饱和度大于10%的缺失特征数据。
进一步地,本申请的另一实施例中,在生成所述特征维度列表之前,还可以包括:对选取的所述缺失数据维度进行校验,对校验成功的所述缺失数据维度进行排序,生成所述特征维度列表。
其中,本申请实施例中,所述对所述缺失数据维度的校验,即选取出所述缺失数据维度与标签值的依赖度大于预设依赖度的缺失数据维度,保障选取的所述缺失数据维度的可靠性。
一个优选实施例中,利用当前已知的Kenall秩相关系数算法计算所述选取的缺失数据维度标签值的依赖度,可选的,所述预设依赖度为0.38。
一个优选实施例中,利用下述方法对校验成功的缺失数据维度进行排序:
f *=(I-m) -1x i
Figure PCTCN2020119092-appb-000001
其中,f *表示缺失数据维度的权重,I表示单位矩阵,x i表示第i个缺失数据 维度,m表示缺失数据维度的偏置,φ为排序因子;
根据所述权重f *,对所述缺失数据维度进行排序,即权重值越大,所述缺失数据维度在特征维度列表位置越靠前。
基于所述缺失数据维度的排序,可以更加直观了解到每一种缺失数据维度与标签值的依赖程度大小。
S2、计算所述特征维度列表中每一种缺失数据维度与所述标签值的相关系数,选取所述相关系数大于预设相关系数的缺失数据维度,得到目标缺失数据维度集。
在本申请实施例中,所述相关系数可以理解为缺失数据维度对标签值的贡献值,例如上述疾病问卷调查,其标签值为疾病,缺失数据维度包括:年龄、性别、体重以及身高等,则年龄、性别、体重以及身高等缺失数据维度对疾病的贡献值,即年龄、性别、体重以及身高等缺失数据维度与疾病之间的相关程度。
较佳地,本申请实施例利用下述方法计算所述特征维度列表中每一种缺失数据维度与所述标签值的相关系数:
Figure PCTCN2020119092-appb-000002
其中,c(x,y)表示缺失数据维度与标签值的相关系数,COV(x,y)表示缺失数据维度与标签值的协方差,Var[X]表示缺失数据维度的方差,Var[Y]为标签值的方差。
进一步地,为了更好的理解所述相关系数,本申请以心冠疾病作为标签值,则其对应的样本数据包括:胆固醇指标、血清蛋白指标以及肾小球指标等,年龄作为缺失数据维度,则其对应的样本数据为年龄范围,包括20~60岁的年龄岁数,利用上述方法计算年龄与心冠疾病的相关系数,可以得到年龄与心冠疾病的相关性,从而可以帮助用户判断心冠疾病是否与年龄相关。
一个优选实施例中,本申请选取所述相关系数大于预设相关系数的缺失数据维度,得到所述目标缺失数据维度集,可以筛除所述特征维度列表中一些不重要的缺失数据维度,加快后续对缺失数据不敏感的模型建立速度,同时也保障了后续模型建立的可靠度,可选的,所述预设相关系数为0.39。
其中,需要强调的是,为进一步保证上述目标缺失数据维度集的私密和安全性,上述目标缺失数据维度集还可以存储于一区块链的节点中。
S3、利用预设数据缺失不敏感算法对所述目标缺失数据维度集进行建模,生成缺失数据不敏感模型。
所述预设数据缺失不敏感算法指的是一种可以自动学习出缺失数据维度中样本数据的分裂方向的方法,即将缺失样本当做稀疏矩阵来对待,在节点分裂时不考虑缺失样本的数值,而缺失样本数据会被分到左子树和右子树分别计层损失,最终选择损失较小的一个。
一个可选实施例中,所述预设数据缺失不敏感算法为当前已知XGBoost算法,所述XGBoost算法是一种改进的决策树算法。
详细地,参阅图5所示,所述S3包括:
S30、利用所述XGBoost算法构建所述目标缺失数据维度集的决策树;
在本申请实施例中,所述决策树的构建原理为:基于所述目标缺失数据维度集所在的输入空间中,递归的将所述输入空间中每个区域划分为两个子区域并决定每个子区域上的输出值,构建所述目标缺失数据维度集的决策树。
S31、计算所述决策树中每一种缺失数据维度的负梯度;
本申请较佳实施例中,所述负梯度指的是对所述目标缺失数据维度集中每一个目标缺失数据维度的残差,通过拟合所述目标缺失数据维度的残差,以增强整个决策树的鲁棒性以及可靠性。
一个优选实施例中,利用下述方法计算所述决策树中每一种缺失数据维度的负梯度:
Figure PCTCN2020119092-appb-000003
其中,r im表示负梯度,
Figure PCTCN2020119092-appb-000004
表示学习率,L(y i,f(x i)表示损失函数,y i表示第i个缺失数据维度的样本数据预测值,f(x i)表示第i个缺失数据维度的样本数据的真实值,f(x)表示决策树中的区域函数,f m-1(x)表示决策树中的区域拟合函数。
S32、根据所述负梯度,更新所述决策树,得到所述缺失数据不敏感模型。
在本申请至少一个实施例中,利用上述所得到的负梯度,不断的训练所述决策树,直至所述每个缺失数据维度对所述决策树都已进行训练,生成所述缺失数据不敏感模型。
基于所述XGBoost算法对所述目标缺失数据维度集进行建模,可以实现在不需要领域知识的前提下,对输入数据与目标值进行快速、自动的建模,以应用于多种不同场景,从而可以避免因数据缺失而导致构建模型时出现分析不准确的现象。
S4、利用所述缺失数据不敏感模型对待分析样本数据集进行数据分析,得到分析结果。
在本申请的至少一个实施例中,所述利用所述缺失数据不敏感模型对待分析样本数据集进行数据分析,得到分析结果,包括:
获取所述待分析样本数据集的样本标签值,例如,待分析样本数据集为用户职业,则其样本标签值可以为薪资;
利用所述缺失数据不敏感模型对所述待分析样本数据集及样本标签值进行分析,得到分析结果,其中,所述分析结果可以理解为所述待分析样本数据集与样本标签值的相关性,例如,利用所述缺失数据缺失不敏感模型对用户职业及薪资进行分析,得到分析结果为用户职业与薪资具有紧密的相关性。
进一步地,需要声明的是,在本申请中,所述缺失数据不敏感模型用于对存在缺失数据的样本进行分析,同时也可适用于对不存在缺失数据的样本进行分析。
综上所述,本申请实施例首先获取缺失数据集及对应的标签值,计算所述缺失数据集中每一种缺失数据维度的饱和度,选取饱和度大于预设饱和度的缺失数据维度,生成特征维度列表,可以了解缺失数据集中每一种缺失数据维度的贡献度;其次,本申请实施例计算所述特征维度列表中每一种缺失数据维度与所述标签值的相关系数,选取所述相关系数大于预设相关系数的缺失数据维度,得到目标缺失数据维度集,可以筛除所述特征维度列表中一些不重要的缺失数据维度,加快后续对缺失数据不敏感的模型建立速度,同时也保障了后续模型建立的可靠度;进一步地,本申请实施例利用预设数据缺失不敏感算法对所述目标缺失数据维度集进行建模,可以实现在不需要领域知识的前提下,对输入数据与目标值进行快速、自动的建模,以应用于多种不同场景,从而可以避免因数据缺失而导致构建模型时出现分析不准确的现象。
如图4所示,是本申请基于缺失数据的样本分析装置的功能模块图。
本申请所述基于缺失数据的样本分析装置100可以安装于电子设备中。根据实现的功能,所述基于缺失数据的样本分析装置可以包括计算及选取模块101、建模模块102以及分析模块103。本发所述模块也可以称之为单元,是指一种能够被电子设备处理器所执行,并且能够完成固定功能的一系列计算机程序段,其存储在电子设备的存储器中。
在本实施例中,关于各模块/单元的功能如下:
所述计算及选取模块101,用于获取缺失数据集及对应的标签值,计算所述缺失数据集中每一种缺失数据维度的饱和度,选取饱和度大于预设饱和度的缺失数据维度,生成特征维度列表。
在本申请的至少一个实施例中,所述缺失数据集基于不同的业务场景采集得到,例如对于问卷调查,许多用户并不愿意回答某些私密性的问题,会选择性的回答问卷中的部分问题,从而导致采集问卷会产生部分的缺失数据,得到缺失数据问卷,根据不同用户的问卷回答,可以得到多个缺失数据问卷,从而形成缺失数据集。
进一步地,所述标签值指的是所述缺失数据集中所包含缺失数据维度对应的预测值,例如关于疾病问卷调查中,其对应的标签值可以为疾病,即疾病问卷调查所包含缺失数据维度对应的预测值为疾病。
在本申请的一个可选实施例中,所述缺失数据维度指的是一种表征缺失数据集数据特征的概念,例如根据疾病问卷调查得到的缺失数据集中,缺失数据维度包括:年龄、性别、体重以及身高等。
较佳地,所述计算所述缺失数据集中每一种缺失数据维度的饱和度,包括:
步骤A、获取所述缺失数据维度的样本集;
例如,对于年龄这个缺失数据维度,采集了500个样本,即500个用户的回答。
步骤B、识别出所述样本集中是否存在非法和/或非空样本;
例如,查询500个年龄样本中是否存在没有填年龄的样本或者是否年龄样本填的并不是数值而是其它字符。
若存在非法和/或非空样本,执行步骤C、筛除所述非法和/或非空样本后,利用预设饱和度计算公式计算筛除后所述样本集的饱和度,得到所述缺失数据维度的饱和度;
若不存在非法和/或非空样本,执行步骤D、利用预设饱和度计算公式计算所述样本集的饱和度,得到所述缺失数据维度的饱和度。
一个可选实施例中,所述预设饱和度计算公式如下所示:
P=1-n/m*100%
其中,p指的是饱和度,n指的是非法和/或非空样本数量,m指的是样本集中样本数量。
优选地,本申请实施例中,所述预设饱和度可以设置为10%,因此本申请保留饱和度大于10%的缺失特征数据。
进一步地,本申请的另一实施例中,在生成所述特征维度列表之前,还可以包括:对选取的所述缺失数据维度进行校验,对校验成功的所述缺失数据维度进行排序,生成所述特征维度列表。
其中,本申请实施例中,所述对所述缺失数据维度的校验,即选取出所述缺失数据维度与标签值的依赖度大于预设依赖度的缺失数据维度,保障选取的所述缺失数据维度的可靠性。
一个优选实施例中,利用当前已知的Kenall秩相关系数算法计算所述选取的缺失数据维度标签值的依赖度,可选的,所述预设依赖度为0.38。
一个优选实施例中,利用下述方法对校验成功的缺失数据维度进行排序:
f *=(I-m) -1x i
Figure PCTCN2020119092-appb-000005
其中,f *表示缺失数据维度的权重,I表示单位矩阵,x i表示第i个缺失数据维度,m表示缺失数据维度的偏置,φ为排序因子;
根据所述权重f *,对所述缺失数据维度进行排序,即权重值越大,所述缺失数据维度在特征维度列表位置越靠前。
基于所述缺失数据维度的排序,可以更加直观了解到每一种缺失数据维度与标签值的依赖程度大小。
所述计算及选取模块101,还用于计算所述特征维度列表中每一种缺失数据维度与所述标签值的相关系数,选取所述相关系数大于预设相关系数的缺失数据维度,得到目标缺失数据维度集。
在本申请实施例中,所述相关系数可以理解为缺失数据维度对标签值的贡献值,例如上述疾病问卷调查,其标签值为疾病,缺失数据维度包括:年龄、性别、体重以及身高等,则年龄、性别、体重以及身高等缺失数据维度对疾病的贡献值,即年龄、性别、体重以及身高等缺失数据维度与疾病之间的相关程度。
较佳地,本申请实施例利用下述方法计算所述特征维度列表中每一种缺失数据维度与所述标签值的相关系数:
Figure PCTCN2020119092-appb-000006
其中,c(x,y)表示缺失数据维度与标签值的相关系数,COV(x,y)表示缺失数据维度与标签值的协方差,Var[X]表示缺失数据维度的方差,Var[Y]为标签值的方差。
进一步地,为了更好的理解所述相关系数,本申请以心冠疾病作为标签值,则其对应的样本数据包括:胆固醇指标、血清蛋白指标以及肾小球指标等,年龄作为缺失数据维度,则其对应的样本数据为年龄范围,包括20~60岁的年龄岁数,利用上述方法计算年龄与心冠疾病的相关系数,可以得到年龄与心冠疾病的相关性,从而可以帮助用户判断心冠疾病是否与年龄相关。
一个优选实施例中,本申请选取所述相关系数大于预设相关系数的缺失数据维度,得到所述目标缺失数据维度集,可以筛除所述特征维度列表中一些不重要的缺失数据维度,加快后续对缺失数据不敏感的模型建立速度,同时也保障了后续模型建立的可靠度,可选的,所述预设相关系数为0.39。
其中,需要强调的是,为进一步保证上述目标缺失数据维度集的私密和安全性,上述目标缺失数据维度集还可以存储于一区块链的节点中。
所述建模模块103,用于利用预设数据缺失不敏感算法对所述目标缺失数据维度集进行建模,生成缺失数据不敏感模型。
所述预设数据缺失不敏感算法指的是一种可以自动学习出缺失数据维度中样本数据的分裂方向的方法,即将缺失样本当做稀疏矩阵来对待,在节点分裂时不考虑缺失样本的数值,而缺失样本数据会被分到左子树和右子树分别计层损失,最终选择损失较小的一个。
一个可选实施例中,所述预设数据缺失不敏感算法为当前已知XGBoost算法,所述XGBoost算法是一种改进的决策树算法。
详细地,所述利用预设数据缺失不敏感算法对所述目标缺失数据维度集进行建模,生成缺失数据不敏感模型包括:
I、利用所述XGBoost算法构建所述目标缺失数据维度集的决策树;
在本申请实施例中,所述决策树的构建原理为:基于所述目标缺失数据维度集所在的输入空间中,递归的将所述输入空间中每个区域划分为两个子区域并决定每个子区域上的输出值,构建所述目标缺失数据维度集的决策树。
II、计算所述决策树中每一种缺失数据维度的负梯度;
本申请较佳实施例中,所述负梯度指的是对所述目标缺失数据维度集中每一个目标缺失数据维度的残差,通过拟合所述目标缺失数据维度的残差,以增强整个决策树的鲁棒性以及可靠性。
一个优选实施例中,利用下述方法计算所述决策树中每一种缺失数据维度的负梯度:
Figure PCTCN2020119092-appb-000007
其中,r im表示负梯度,
Figure PCTCN2020119092-appb-000008
表示学习率,L(y i,f(x i)表示损失函数,y i表示第i个缺失数据维度的样本数据预测值,f(x i)表示第i个缺失数据维度的样本数据的真实值,f(x)表示决策树中的区域函数,f m-1(x)表示决策树中的区域拟合函数。
III、根据所述负梯度,更新所述决策树,得到所述缺失数据不敏感模型。
在本申请至少一个实施例中,利用上述所得到的负梯度,不断的训练所述决策树,直至所述每个缺失数据维度对所述决策树都已进行训练,生成所述缺失数据不敏感模型。
基于所述XGBoost算法对所述目标缺失数据维度集进行建模,可以实现在不需要领域知识的前提下,对输入数据与目标值进行快速、自动的建模,以应用于多种不同场景,从而可以避免因数据缺失而导致构建模型时出现分析不准确的现象。
所述分析模块104,用于利用所述缺失数据不敏感模型对待分析样本数据集进行数据分析,得到分析结果。
在本申请的至少一个实施例中,所述利用所述缺失数据不敏感模型对待分析样本数据集进行数据分析,得到分析结果,包括:
获取所述待分析样本数据集的样本标签值,例如,待分析样本数据集为用户职业,则其样本标签值可以为薪资;
利用所述缺失数据不敏感模型对所述待分析样本数据集及样本标签值进行分析,得到分析结果,其中,所述分析结果可以理解为所述待分析样本数据集与样本标签值的相关性,例如,利用所述缺失数据缺失不敏感模型对用户职业及薪资进行分析,得到分析结果为用户职业与薪资具有紧密的相关性。
进一步地,需要声明的是,在本申请中,所述缺失数据不敏感模型用于对存在缺失数据的样本进行分析,同时也可适用于对不存在缺失数据的样本进行分析。
综上所述,本申请实施例首先获取缺失数据集及对应的标签值,计算所述缺失数据集中每一种缺失数据维度的饱和度,选取饱和度大于预设饱和度的缺失数据维度,生成特征维度列表,可以了解缺失数据集中每一种缺失数据维度的贡献度;其次,本申请实施例计算所述特征维度列表中每一种缺失数据维度与所述标签值的相关系数,选取所述相关系数大于预设相关系数的缺失数据维度,得到目标缺失数据维度集,可以筛除所述特征维度列表中一些不重要的缺失数据维度,加快后续对缺失数据不敏感的模型建立速度,同时也保障了后续模型建立的可靠度;进一步地,本申请实施例利用预设数据缺失不敏感算法对所述目标缺失数据维度集进行建模,可以实现在不需要领域知识的前提下,对输入数据与目标值进行快速、自动的建模,以应用于多种不同场景,从而可以避免因数据缺失而导致构建模型时出现分析不准确的现象。
如图5所示,是本申请实现基于缺失数据的样本分析方法的电子设备的结构示意图。
所述电子设备1可以包括处理器10、存储器11和总线,还可以包括存储在所述存储器11中并可在所述处理器10上运行的计算机程序,如基于缺失数据的样本分析程序12。
其中,所述存储器11至少包括一种类型的可读存储介质,所述可读存储介质可以是非易失性,也可以是易失性,所述可读存储介质包括闪存、移动硬盘、多媒体卡、卡型存储器(例如:SD或DX存储器等)、磁性存储器、磁盘、光盘等。所述存储器11在一些实施例中可以是电子设备1的内部存储单元,例如该电子设备1的移动硬盘。所述存储器11在另一些实施例中也可以是电子设备1的外部存储设备,例如电子设备1上配备的插接式移动硬盘、智能存储卡(Smart Media Card,SMC)、安全数字(Secure Digital,SD)卡、闪存卡(Flash Card)等。进一步地,所述存储器11还可以既包括电子设备1的内部存储单元也包括外部存储设备。所述存储器11不仅可以用于存储安装于电子设备1的应用软件及各类数据,例如基于缺失数据的样本分析的代码等,还可以用于暂时地存储已经输出或者将要输出的数据。
所述处理器10在一些实施例中可以由集成电路组成,例如可以由单个封装的集成电路所组成,也可以是由多个相同功能或不同功能封装的集成电路所组成,包括一个或者多个中央处理器(Central Processing unit,CPU)、微处理器、数字处理芯片、图形处理器及各种控制芯片的组合等。所述处理器10是所述电子设备的控制核心(Control Unit),利用各种接口和线路连接整个电子设备的各个部件,通过运行或执行存储在所述存储器11内的程序或者模块(例如执行基于缺失数据的样本分析等),以及调用存储在所述存储器11内的数据,以执行电子设备1的各种功能和处理数据。
所述总线可以是外设部件互连标准(peripheral component interconnect,简称PCI)总线或扩展工业标准结构(extended industry standard architecture,简称EISA)总线等。该总线可以分为地址总线、数据总线、控制总线等。所述总线被设置为实现所述存储器11以及至少一个处理器10等之间的连接通信。
图5仅示出了具有部件的电子设备,本领域技术人员可以理解的是,图5示出的结构并不构成对所述电子设备1的限定,可以包括比图示更少或者更多的部件,或者组合某些部件,或者不同的部件布置。
例如,尽管未示出,所述电子设备1还可以包括给各个部件供电的电源(比如电池),优选地,电源可以通过电源管理装置与所述至少一个处理器10逻辑相 连,从而通过电源管理装置实现充电管理、放电管理、以及功耗管理等功能。电源还可以包括一个或一个以上的直流或交流电源、再充电装置、电源故障检测电路、电源转换器或者逆变器、电源状态指示器等任意组件。所述电子设备1还可以包括多种传感器、蓝牙模块、Wi-Fi模块等,在此不再赘述。
进一步地,所述电子设备1还可以包括网络接口,可选地,所述网络接口可以包括有线接口和/或无线接口(如WI-FI接口、蓝牙接口等),通常用于在该电子设备1与其他电子设备之间建立通信连接。
可选地,该电子设备1还可以包括用户接口,用户接口可以是显示器(Display)、输入单元(比如键盘(Keyboard)),可选地,用户接口还可以是标准的有线接口、无线接口。可选地,在一些实施例中,显示器可以是LED显示器、液晶显示器、触控式液晶显示器以及OLED(Organic Light-Emitting Diode,有机发光二极管)触摸器等。其中,显示器也可以适当的称为显示屏或显示单元,用于显示在电子设备1中处理的信息以及用于显示可视化的用户界面。
应该了解,所述实施例仅为说明之用,在专利申请范围上并不受此结构的限制。
所述电子设备1中的所述存储器11存储的基于缺失数据的样本分析12是多个指令的组合,在所述处理器10中运行时,可以实现:
获取缺失数据集及对应的标签值,计算所述缺失数据集中每一种缺失数据维度的饱和度,选取饱和度大于预设饱和度的缺失数据维度,生成特征维度列表;
计算所述特征维度列表中每一种缺失数据维度与所述标签值的相关系数,选取所述相关系数大于预设相关系数的缺失数据维度,得到目标缺失数据维度集;
利用预设数据缺失不敏感算法对所述目标缺失数据维度集进行建模,生成缺失数据不敏感模型;
利用所述缺失数据不敏感模型对待分析样本数据集进行数据分析,得到分析结果。
具体地,所述处理器10对上述指令的具体实现方法可参考图1对应实施例中相关步骤的描述,在此不赘述。
进一步地,所述电子设备1集成的模块/单元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,可以存储在一个非易失性计算机可读取存储介质中。所述计算机可读介质可以包括:能够携带所述计算机程序代码的任何实体或装置、记录介质、U盘、移动硬盘、磁碟、光盘、计算机存储器、只读存储器(ROM,Read-Only Memory)。
在本申请所提供的几个实施例中,应该理解到,所揭露的设备,装置和方法,可以通过其它的方式实现。例如,以上所描述的装置实施例仅仅是示意性的,例如,所述模块的划分,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。
所述作为分离部件说明的模块可以是或者也可以不是物理上分开的,作为模块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模块来实现本实施例方案的目的。
另外,在本申请各个实施例中的各功能模块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。上述集成的单元既可以采用硬件的形式实现,也可以采用硬件加软件功能模块的形式实现。
对于本领域技术人员而言,显然本申请不限于上述示范性实施例的细节,而且在不背离本申请的精神或基本特征的情况下,能够以其他的具体形式实现本申请。
因此,无论从哪一点来看,均应将实施例看作是示范性的,而且是非限制性的,本申请的范围由所附权利要求而不是上述说明限定,因此旨在将落在权利要求的等同要件的含义和范围内的所有变化涵括在本申请内。不应将权利要求中的任何附关联图标记视为限制所涉及的权利要求。
本申请所指区块链是分布式数据存储、点对点传输、共识机制、加密算法等计算机技术的新型应用模式。区块链(Blockchain),本质上是一个去中心化的数据库,是一串使用密码学方法相关联产生的数据块,每一个数据块中包含了一批次网络交易的信息,用于验证其信息的有效性(防伪)和生成下一个区块。区块链可以包括区块链底层平台、平台产品服务层以及应用服务层等。
此外,显然“包括”一词不排除其他单元或步骤,单数不排除复数。系统权利要求中陈述的多个单元或装置也可以由一个单元或装置通过软件或者硬件来实现。第二等词语用来表示名称,而并不表示任何特定的顺序。
最后应说明的是,以上实施例仅用以说明本申请的技术方案而非限制,尽管参照较佳实施例对本申请进行了详细说明,本领域的普通技术人员应当理解,可以对本申请的技术方案进行修改或等同替换,而不脱离本申请技术方案的精神和范围。

Claims (20)

  1. 一种基于缺失数据的样本分析方法,其中,所述方法包括:
    获取缺失数据集及对应的标签值,计算所述缺失数据集中每一种缺失数据维度的饱和度,选取饱和度大于预设饱和度的缺失数据维度,生成特征维度列表;
    计算所述特征维度列表中每一种缺失数据维度与所述标签值的相关系数,选取所述相关系数大于预设相关系数的缺失数据维度,得到目标缺失数据维度集;
    利用预设数据缺失不敏感算法对所述目标缺失数据维度集进行建模,生成缺失数据不敏感模型;
    利用所述缺失数据不敏感模型对待分析样本数据集进行数据分析,得到分析结果。
  2. 如权利要求1所述的基于缺失数据的样本分析方法,其中,所述计算所述缺失数据集中每一种缺失数据维度的饱和度,包括:
    获取所述缺失数据维度的样本集;
    识别出所述样本集中是否存在非法和/或非空样本;
    若存在非法和/或非空样本,则筛除所述非法和/或非空样本后,利用预设饱和度计算公式计算筛除后所述样本集的饱和度,得到所述缺失数据维度的饱和度;
    若不存在非法和/或非空样本,则利用预设饱和度计算公式计算所述样本集的饱和度,得到所述缺失数据维度的饱和度。
  3. 如权利要求1所述的基于缺失数据的样本分析方法,其中,所述生成特征维度列表之前,该方法还包括:对选取的缺失数据维度进行校验,对校验成功的缺失数据维度进行排序,生成所述特征维度列表。
  4. 如权利要求3所述的基于缺失数据的样本分析方法,其中,所述对校验成功的所述缺失数据维度进行排序,包括:
    利用下述方法计算所述缺失数据维度的权重f *
    f *=(I-m) -1x i
    Figure PCTCN2020119092-appb-100001
    其中,I表示单位矩阵,x i表示第i个缺失数据维度,m表示缺失数据维度的偏置,φ为排序因子;
    根据所述权重大小,对所述缺失数据维度进行排序。
  5. 如权利要求1所述的基于缺失数据的样本分析方法,其中,所述计算所述特征维度列表中每一种缺失数据维度与所述标签值的相关系数,包括:
    利用下述方法计算所述特征维度列表中每一种缺失数据维度与所述标签值的相关系数:
    Figure PCTCN2020119092-appb-100002
    其中,c(x,y)表示缺失数据维度与标签值的相关系数,COV(x,y)表示缺失数据维度与标签值的协方差,Var[X]表示缺失数据维度的方差,Var[Y]为标签值的方差。
  6. 如权利要求1至5中任意一项所述的基于缺失数据的样本分析方法,其中,所述利用预设数据缺失不敏感算法对所述目标缺失数据维度集进行建模,生成缺失数据不敏感模型,包括:
    利用所述预设数据缺失不敏感算法构建所述目标缺失数据维度集的决策树;
    计算所述决策树中每一种缺失数据维度的负梯度;
    根据所述负梯度,更新所述决策树,得到所述缺失数据不敏感模型。
  7. 如权利要求6所述的基于缺失数据的样本分析方法,其中,所述计算所述决策树中每一种缺失数据维度的负梯度,包括:
    利用下述方法计算所述决策树中每一种缺失数据维度的负梯度:
    Figure PCTCN2020119092-appb-100003
    其中,r im表示负梯度,
    Figure PCTCN2020119092-appb-100004
    表示学习率,L(y i,f(x i)表示损失函数,y i表示第i个缺失数据维度的样本数据预测值,f(x i)表示第i个缺失数据维度的样本数据的真实值,f(x)表示决策树中的区域函数,f m-1(x)表示决策树中的区域拟合函数。
  8. 一种基于缺失数据的样本分析装置,其中,所述装置包括:
    计算及选取模块,用于获取缺失数据集及对应的标签值,计算所述缺失数据集中每一种缺失数据维度的饱和度,选取饱和度大于预设饱和度的缺失数据维度,生成特征维度列表;
    所述计算及选取模块,还用于计算所述特征维度列表中每一种缺失数据维度与所述标签值的相关系数,选取所述相关系数大于预设相关系数的缺失数据维度,得到目标缺失数据维度集;
    建模模块,用于利用预设数据缺失不敏感算法对所述目标缺失数据维度集进行建模,生成缺失数据不敏感模型;
    分析模块,用于利用所述缺失数据不敏感模型对待分析样本数据集进行数据分析,得到分析结果。
  9. 一种电子设备,其中,所述电子设备包括:
    至少一个处理器;以及,
    与所述至少一个处理器通信连接的存储器;其中,
    所述存储器存储有可被所述至少一个处理器执行的指令,所述指令被所述至少一个处理器执行,以使所述至少一个处理器能够执行如下步骤:
    获取缺失数据集及对应的标签值,计算所述缺失数据集中每一种缺失数据维度的饱和度,选取饱和度大于预设饱和度的缺失数据维度,生成特征维度列表;
    计算所述特征维度列表中每一种缺失数据维度与所述标签值的相关系数,选取所述相关系数大于预设相关系数的缺失数据维度,得到目标缺失数据维度集;
    利用预设数据缺失不敏感算法对所述目标缺失数据维度集进行建模,生成缺失数据不敏感模型;
    利用所述缺失数据不敏感模型对待分析样本数据集进行数据分析,得到分析结果。
  10. 如权利要求9所述的电子设备,其中,所述计算所述缺失数据集中每一种缺失数据维度的饱和度,包括:
    获取所述缺失数据维度的样本集;
    识别出所述样本集中是否存在非法和/或非空样本;
    若存在非法和/或非空样本,则筛除所述非法和/或非空样本后,利用预设饱 和度计算公式计算筛除后所述样本集的饱和度,得到所述缺失数据维度的饱和度;
    若不存在非法和/或非空样本,则利用预设饱和度计算公式计算所述样本集的饱和度,得到所述缺失数据维度的饱和度。
  11. 如权利要求9所述的电子设备,其中,所述生成特征维度列表之前,该方法还包括:对选取的缺失数据维度进行校验,对校验成功的缺失数据维度进行排序,生成所述特征维度列表。
  12. 如权利要求11所述的电子设备,其中,所述对校验成功的所述缺失数据维度进行排序,包括:
    利用下述方法计算所述缺失数据维度的权重f *
    f *=(I-m) -1x i
    Figure PCTCN2020119092-appb-100005
    其中,I表示单位矩阵,x i表示第i个缺失数据维度,m表示缺失数据维度的偏置,φ为排序因子;
    根据所述权重大小,对所述缺失数据维度进行排序。
  13. 如权利要求9所述的电子设备,其中,所述计算所述特征维度列表中每一种缺失数据维度与所述标签值的相关系数,包括:
    利用下述方法计算所述特征维度列表中每一种缺失数据维度与所述标签值的相关系数:
    Figure PCTCN2020119092-appb-100006
    其中,c(x,y)表示缺失数据维度与标签值的相关系数,COV(x,y)表示缺失数据维度与标签值的协方差,Var[X]表示缺失数据维度的方差,Var[Y]为标签值的方差。
  14. 如权利要求9至13中任意一项所述的电子设备,其中,所述利用预设数据缺失不敏感算法对所述目标缺失数据维度集进行建模,生成缺失数据不敏感模型,包括:
    利用所述预设数据缺失不敏感算法构建所述目标缺失数据维度集的决策树;
    计算所述决策树中每一种缺失数据维度的负梯度;
    根据所述负梯度,更新所述决策树,得到所述缺失数据不敏感模型。
  15. 如权利要求14所述的电子设备,其中,所述计算所述决策树中每一种缺失数据维度的负梯度,包括:
    利用下述方法计算所述决策树中每一种缺失数据维度的负梯度:
    Figure PCTCN2020119092-appb-100007
    其中,r im表示负梯度,
    Figure PCTCN2020119092-appb-100008
    表示学习率,L(y i,f(x i)表示损失函数,y i表示第i个缺失数据维度的样本数据预测值,f(x i)表示第i个缺失数据维度的样本数据的真实值,f(x)表示决策树中的区域函数,f m-1(x)表示决策树中的区域拟合函数。
  16. 一种计算机可读存储介质,存储有计算机程序,其中,所述计算机程序 被处理器执行时实现如下步骤:
    获取缺失数据集及对应的标签值,计算所述缺失数据集中每一种缺失数据维度的饱和度,选取饱和度大于预设饱和度的缺失数据维度,生成特征维度列表;
    计算所述特征维度列表中每一种缺失数据维度与所述标签值的相关系数,选取所述相关系数大于预设相关系数的缺失数据维度,得到目标缺失数据维度集;
    利用预设数据缺失不敏感算法对所述目标缺失数据维度集进行建模,生成缺失数据不敏感模型;
    利用所述缺失数据不敏感模型对待分析样本数据集进行数据分析,得到分析结果。
  17. 如权利要求16所述的计算机可读存储介质,其中,所述计算所述缺失数据集中每一种缺失数据维度的饱和度,包括:
    获取所述缺失数据维度的样本集;
    识别出所述样本集中是否存在非法和/或非空样本;
    若存在非法和/或非空样本,则筛除所述非法和/或非空样本后,利用预设饱和度计算公式计算筛除后所述样本集的饱和度,得到所述缺失数据维度的饱和度;
    若不存在非法和/或非空样本,则利用预设饱和度计算公式计算所述样本集的饱和度,得到所述缺失数据维度的饱和度.
  18. 如权利要求16所述的计算机可读存储介质,其中,所述生成特征维度列表之前,该方法还包括:对选取的缺失数据维度进行校验,对校验成功的缺失数据维度进行排序,生成所述特征维度列表。
  19. 如权利要求18所述的计算机可读存储介质,其中,所述对校验成功的所述缺失数据维度进行排序,包括:
    利用下述方法计算所述缺失数据维度的权重f *
    f *=(I-m) -1x i
    Figure PCTCN2020119092-appb-100009
    其中,I表示单位矩阵,x i表示第i个缺失数据维度,m表示缺失数据维度的偏置,φ为排序因子;
    根据所述权重大小,对所述缺失数据维度进行排序。
  20. 如权利要求16所述的计算机可读存储介质,其中,所述计算所述特征维度列表中每一种缺失数据维度与所述标签值的相关系数,包括:
    利用下述方法计算所述特征维度列表中每一种缺失数据维度与所述标签值的相关系数:
    Figure PCTCN2020119092-appb-100010
    其中,c(x,y)表示缺失数据维度与标签值的相关系数,COV(x,y)表示缺失数据维度与标签值的协方差,Var[X]表示缺失数据维度的方差,Var[Y]为标签值的方差。
PCT/CN2020/119092 2020-07-16 2020-09-29 基于缺失数据的样本分析方法、装置、电子设备及介质 WO2021151305A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010684956.3 2020-07-16
CN202010684956.3A CN111831708A (zh) 2020-07-16 2020-07-16 基于缺失数据的样本分析方法、装置、电子设备及介质

Publications (1)

Publication Number Publication Date
WO2021151305A1 true WO2021151305A1 (zh) 2021-08-05

Family

ID=72923408

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/119092 WO2021151305A1 (zh) 2020-07-16 2020-09-29 基于缺失数据的样本分析方法、装置、电子设备及介质

Country Status (2)

Country Link
CN (1) CN111831708A (zh)
WO (1) WO2021151305A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116227929A (zh) * 2023-03-07 2023-06-06 广州爱浦路网络技术有限公司 通信数据的分析方法、装置、设备及存储介质
CN117591920A (zh) * 2024-01-17 2024-02-23 北京科百宏业科技有限公司 历史数据的补充方法、装置、电子设备和可读存储介质

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114186182A (zh) * 2021-11-02 2022-03-15 联想(北京)有限公司 一种处理方法、装置、设备及存储介质

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8255412B2 (en) * 2008-12-17 2012-08-28 Microsoft Corporation Boosting algorithm for ranking model adaptation
CN108717867A (zh) * 2018-05-02 2018-10-30 中国科学技术大学苏州研究院 基于梯度迭代树的疾病预测模型建立方法及装置
CN109948851A (zh) * 2019-03-20 2019-06-28 北京百度网讯科技有限公司 用于预测事件发生概率的方法和装置
CN110427974A (zh) * 2019-07-09 2019-11-08 华中科技大学 一种基于广义支持向量机的液压部件健康状态检测方法
CN110909963A (zh) * 2018-09-14 2020-03-24 中国软件与技术服务股份有限公司 一种信用评分卡模型训练方法及纳税人非正常风险评估方法
CN110956303A (zh) * 2019-10-12 2020-04-03 未鲲(上海)科技服务有限公司 信息预测方法、装置、终端及可读存储介质
CN111292298A (zh) * 2020-01-21 2020-06-16 无锡祥生医疗科技股份有限公司 乳腺癌病理分型确定方法、装置和存储介质

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8255412B2 (en) * 2008-12-17 2012-08-28 Microsoft Corporation Boosting algorithm for ranking model adaptation
CN108717867A (zh) * 2018-05-02 2018-10-30 中国科学技术大学苏州研究院 基于梯度迭代树的疾病预测模型建立方法及装置
CN110909963A (zh) * 2018-09-14 2020-03-24 中国软件与技术服务股份有限公司 一种信用评分卡模型训练方法及纳税人非正常风险评估方法
CN109948851A (zh) * 2019-03-20 2019-06-28 北京百度网讯科技有限公司 用于预测事件发生概率的方法和装置
CN110427974A (zh) * 2019-07-09 2019-11-08 华中科技大学 一种基于广义支持向量机的液压部件健康状态检测方法
CN110956303A (zh) * 2019-10-12 2020-04-03 未鲲(上海)科技服务有限公司 信息预测方法、装置、终端及可读存储介质
CN111292298A (zh) * 2020-01-21 2020-06-16 无锡祥生医疗科技股份有限公司 乳腺癌病理分型确定方法、装置和存储介质

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
CHONGQING SHUFENG TECHNOLOGY CO., LTD.: "Seven dimensionality reduction solutions in the field of big data analysis", NEWS, 7 June 2018 (2018-06-07), CN, pages 1 - 5, XP009529455, Retrieved from the Internet <URL:http://www.softft.com/news/155.html> *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116227929A (zh) * 2023-03-07 2023-06-06 广州爱浦路网络技术有限公司 通信数据的分析方法、装置、设备及存储介质
CN116227929B (zh) * 2023-03-07 2024-03-19 广州爱浦路网络技术有限公司 通信数据的分析方法、装置、设备及存储介质
CN117591920A (zh) * 2024-01-17 2024-02-23 北京科百宏业科技有限公司 历史数据的补充方法、装置、电子设备和可读存储介质

Also Published As

Publication number Publication date
CN111831708A (zh) 2020-10-27

Similar Documents

Publication Publication Date Title
WO2021151305A1 (zh) 基于缺失数据的样本分析方法、装置、电子设备及介质
WO2022141861A1 (zh) 情感分类方法、装置、电子设备及存储介质
WO2021151306A1 (zh) 智能问答语料分析方法、装置、电子设备及可读存储介质
WO2022160449A1 (zh) 文本分类方法、装置、电子设备及存储介质
CN112231586A (zh) 基于迁移学习的课程推荐方法、装置、设备及介质
CN112380859A (zh) 舆情信息的推荐方法、装置、电子设备及计算机存储介质
CN106600052A (zh) 一种基于时空轨迹的用户属性与社会网络检测系统
CN115002200B (zh) 基于用户画像的消息推送方法、装置、设备及存储介质
WO2023159755A1 (zh) 虚假新闻检测方法、装置、设备及存储介质
CN112288337B (zh) 行为推荐方法、装置、设备及介质
CN113887941B (zh) 业务流程生成方法、装置、电子设备及介质
CN114663198A (zh) 基于用户画像的产品推荐方法、装置、设备及存储介质
CN114840531B (zh) 基于血缘关系的数据模型重构方法、装置、设备及介质
CN112906377A (zh) 基于实体限制的问答方法、装置、电子设备及存储介质
CN114612194A (zh) 产品推荐方法、装置、电子设备及存储介质
CN114398560B (zh) 基于web平台的营销界面设置方法、装置、设备及介质
CN110321546A (zh) 账号识别、显示方法、装置、服务器、终端及存储介质
CN114862140A (zh) 基于行为分析的潜力评估方法、装置、设备及存储介质
CN112860851B (zh) 基于根因分析的课程推荐方法、装置、设备及介质
CN112069782B (zh) 文档模板生成方法、装置、电子设备及存储介质
CN112948705A (zh) 一种基于政策大数据的智能匹配方法、装置及介质
CN113591881B (zh) 基于模型融合的意图识别方法、装置、电子设备及介质
CN113435746B (zh) 用户工作量的评分方法、装置、电子设备及存储介质
CN114780688A (zh) 基于规则匹配的文本质检方法、装置、设备及存储介质
CN114722280A (zh) 基于用户画像的课程推荐方法、装置、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: 20916376

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: 20916376

Country of ref document: EP

Kind code of ref document: A1