Method and device for determining accuracy of prediction information
By obtaining and converting multi-dimensional information vector sets, using vector sets in the target storage for matching and analysis, the problem of data analysis in the prior art cannot be realized without leaking the original data, and the data security and analysis efficiency are achieved promote.
Patent Information
- Application Number
- CN202411170253.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-23
- Publication Date
- 2024-11-26
AI Technical Summary
The prior art is difficult to implement data analysis without leaking original data, resulting in the inability to determine the accuracy of prediction information.
By obtaining and converting multi-dimensional information vector sets, using the vector sets in the target storage for matching and analysis, calculating the accuracy of prediction information, ensuring that the original data is safe and not leaked.
While ensuring data security, the accuracy of data analysis and prediction information is achieved, improving data security and analysis efficiency.
Smart Images

Figure CN119025565A_ABST
Abstract
Description
Technical Field
[0001] The embodiments of the present application relate to the field of data processing, and specifically, to a method and device for determining the accuracy of prediction information. Background Art
[0002] Protecting privacy and security is critical for both individuals and organizations. To ensure privacy and security, people often take measures such as strengthening password security, regularly updating software, using encrypted communication tools, and sharing personal information with caution. Companies are also constantly strengthening privacy protection measures to ensure the security and privacy of user data. Although privacy and security are critical, in some cases, individuals and organizations may need to share information data to help others conduct data analysis. In this case, how to ensure the security of original data and realize relevant data analysis without leaking the original information is an urgent problem that needs to be solved.
[0003] There is currently no effective solution to the above problems. Summary of the invention
[0004] The embodiments of the present application provide a method and device for determining the accuracy of prediction information, so as to at least solve the problem in the related art that it is impossible to perform data analysis without leaking the original data, thereby determining the accuracy of prediction information.
[0005] According to an embodiment of the present application, a method for determining the accuracy of prediction information is provided, comprising: obtaining a first vector set, wherein the first vector set includes vectors of N dimensions, and the vectors of the N dimensions are vectors obtained by performing vector transformation on information of the N dimensions; determining a second vector set that does not match the first vector set in a target storage, and obtaining a first number of second vector sets, wherein the second vector set includes vectors of M dimensions, and the vectors of the M dimensions are vectors obtained by performing vector transformation on information of the M dimensions, and M is greater than N; obtaining prediction information of a target object based on the information of the N dimensions, and performing vector transformation on the prediction information to obtain a prediction vector, and determining the vectors of the N dimensions and the prediction vector as a third vector set, wherein the prediction information is information of the first dimension, and the M dimensions include the first dimension; determining a fourth vector set that does not match the third vector set in the target storage, and obtaining a second number of fourth vector sets, wherein the fourth vector set includes the vectors of the M dimensions; and determining the accuracy of the prediction information according to the first number and the second number.
[0006] In an exemplary embodiment, before determining in the target storage a second vector set that does not match the first vector set, the method includes: acquiring sample information of multiple samples to obtain multiple sample information sets, wherein each of the sample information sets includes information of a sample in M dimensions, and the information in the M dimensions includes information of multiple media types; performing vector conversion on each information in the multiple sample information sets according to the media type to obtain multiple sample vector sets, wherein the sample vector sets include vectors of the M dimensions, and the vectors of the M dimensions are vectors obtained by performing vector conversion on each information in the corresponding sample information sets, and the vector conversion method for sample information of different media types is different; and storing the multiple sample vector sets in the target storage.
[0007] In an exemplary embodiment, vector conversion is performed on each information in a plurality of sample information sets according to the media type to obtain a plurality of sample vector sets, including: performing the following operations on each sample information in the sample information set, and taking the sample information subjected to the following operations as the current sample information: when the media type of the current sample information is a text type, performing vector conversion on the current sample information according to a text vector conversion method to obtain a current sample vector of the current sample information; when the type of the current sample information is an image type, performing vector conversion on the current sample information according to an image vector conversion method to obtain a current sample vector of the current sample information; when the media type of the current sample information is an audio type, performing vector conversion on the current sample information according to an audio vector conversion method to obtain a current sample vector of the current sample information; when the type of the current sample information is a video type, performing vector conversion on the current sample information according to a video vector conversion method to obtain a current sample vector of the current sample information.
[0008] In an exemplary embodiment, storing the multiple sample vector sets in the target storage includes: performing the following operations on each sample vector set in the multiple sample vector sets, using the sample vector set that performs the following operations as a first current sample vector set, wherein the first current sample vector set is a sample vector set obtained by performing the vector conversion on sample information of the current sample: determining the sample type of the current sample as the current sample type, wherein the multiple samples include multiple sample types; determining a current hypercube model that does not match the current sample type in the target storage, wherein the target storage contains multiple hypercube models, the multiple hypercube models include the current hypercube model, each of the hypercube models is used to store a sample vector set of the same sample type, and each point on the hypercube model is used to store a sample vector set; storing each sample vector in the first current sample vector set to the current hypercube model.
[0009] In an exemplary embodiment, determining a second vector set that does not match the first vector set in the target storage includes: determining the type of the target object as a target type; determining a target hypercube model that does not match the target type from a plurality of the hypercube models in the target storage; traversing each node in the target hypercube model, and determining a sample vector set stored on each node whose similarity to the first vector set is greater than or equal to a preset similarity threshold as the second vector set.
[0010] In an exemplary embodiment, before determining the sample vector set stored on the nodes whose similarity with the first vector set is greater than or equal to a preset similarity threshold in each of the nodes as the second vector set, the method further includes: calculating the similarity between the first vector set and the sample vector set stored on each node in the target hypercube model in the following manner, including: traversing each node in the target hypercube model, taking the currently traversed node as the current node, and taking the sample vector set stored on the current node as the second current sample vector set; determining the vectors of N dimensions that match the vectors of N dimensions of the first vector set from the vectors of M dimensions in the second current sample vector set to obtain a second current sample vector subset; determining the vectors of N dimensions of the first vector set that match the vectors of N dimensions of the second current sample vector set to obtain a second current sample vector subset; The similarity between the vectors of N dimensions in the subset is determined to obtain N similarities; the sum of the N similarity values is determined as the similarity between the first vector set and the second current sample vector set stored on the current node; wherein, the similarity between the vectors of N dimensions in the first vector set and the vectors of N dimensions in the subset of the second current sample vector is determined to obtain N similarities, including: performing the following operations on each vector of the N dimensions in the first vector set, and the vectors performing the following operations are used as the first current vector, and the first current vector is the vector of the current dimension; determining the vector of the current dimension in the subset of the second current sample vector to obtain the second current vector, and calculating the similarity between the first current vector and the second current vector to obtain the current similarity, wherein the N similarities include the current similarity.
[0011] In an exemplary embodiment, determining the accuracy of the prediction information based on the first number and the second number includes: determining the ratio of the second number to the first number as the accuracy of the prediction information; after determining the accuracy of the prediction information, the method further includes: when the accuracy ratio of the prediction information is greater than or equal to a preset threshold, determining the prediction information as the diagnosis result of the target object, and outputting the prediction information, wherein the information of N dimensions corresponding to the vectors of the N dimensions in the first vector set is the detection information of the target object; when the accuracy ratio of the prediction information is less than the preset threshold, outputting a prompt message, wherein the prompt message is used to indicate that the prediction information is erroneous information.
[0012] According to another embodiment of the present application, a device for determining the accuracy of prediction information is provided, including: a first acquisition module, used to acquire a first vector set, wherein the first vector set includes vectors of N dimensions, and the vectors of the N dimensions are vectors obtained by vector conversion of information of the N dimensions; a first determination module, used to determine a second vector set that does not match the first vector set in the target storage, and obtain a first number of second vector sets, wherein the second vector set includes vectors of M dimensions, and the vectors of the M dimensions are vectors obtained by vector conversion of information of the M dimensions, and M is greater than N; a second acquisition module, used to Acquire prediction information of the target object based on the information of the N dimensions, perform vector conversion on the prediction information to obtain a prediction vector, determine the vectors of the N dimensions and the prediction vector as a third vector set, wherein the prediction information is information of the first dimension, and the M dimensions include the first dimension; a second determination module is used to determine a fourth vector set that does not match the third vector set in the target storage to obtain a second number of fourth vector sets, wherein the fourth vector set includes the vectors of the M dimensions; a third determination module is used to determine the accuracy of the prediction information based on the first number and the second number.
[0013] According to another embodiment of the present application, a computer-readable storage medium is provided, in which a computer program is stored, wherein the computer program is configured to execute the steps of any of the above method embodiments when running.
[0014] According to another embodiment of the present application, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.
[0015] According to another embodiment of the present application, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in any of the above method embodiments are implemented.
[0016] Through the present application, after the prediction information of the target object is determined based on the information of N dimensions represented by the vectors of N dimensions in the first vector set, the prediction information is vector-converted to obtain the prediction vector, and the third vector set is determined based on the prediction vector and the first vector set, and the accuracy of the prediction information is determined based on the first number of vector sets that do not match the first vector set and the second number that do not match the third vector set determined in the target storage, so that the accuracy of the prediction information can be determined by using only the relevant information through the vector set in the target storage while ensuring the security of the original data and preventing it from being leaked. Therefore, the problem in the related art that data analysis cannot be realized without leaking the original data, so as to determine the accuracy of the prediction information, can be solved, and the effect of improving data security can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] Figure 1 It is a hardware structure block diagram of a server device according to a method for determining the accuracy of prediction information according to an embodiment of the present application;
[0018] Figure 2 is a flow chart of a method for determining the accuracy of prediction information according to an embodiment of the present application;
[0019] Figure 3 is a sample information storage flow chart according to an embodiment of the present application;
[0020] Figure 4 is a schematic structural diagram of a three-dimensional hypercube model according to an embodiment of the present application;
[0021] Figure 5 It is a schematic diagram of the overall process according to an embodiment of the present application;
[0022] Figure 6 It is a structural block diagram of a device for determining the accuracy of prediction information according to an embodiment of the present application. DETAILED DESCRIPTION
[0023] In order to help better understand the embodiments of the present application, some abbreviations are explained below:
[0024] VSS Vector: Similarity Search, vector similarity search.
[0025] SQL: Structured Query Language.
[0026] VSS-SQL: Vector Similarity Search & Structured Query Language, a structured query language for vector similarity search.
[0027] MM: Multiple Modality, multimodality refers to the collaborative reasoning of multiple heterogeneous modal data.
[0028] HC:Hyper Cube, hypercube means using three or more dimensions to describe an object, with each dimension being perpendicular to each other.
[0029] PC: Privacy compute, privacy computing refers to a set of technologies that implement data analysis and computing while protecting the data itself from external leakage.
[0030] CPU: Central Processing Unit.
[0031] Privacy computing refers to a collection of technologies that implement data analysis and computing while protecting the data from external disclosure, so as to achieve the goal of "available but invisible" data; and to achieve data sharing, value transformation and release while fully protecting data and privacy security. At present, mainstream privacy computing technologies are mainly divided into three directions: the first category is privacy computing technology based on cryptography represented by multi-party secure computing; the second category is technology derived from the integration of artificial intelligence and privacy protection technologies represented by federated learning; the third category is privacy computing technology based on trusted hardware represented by trusted execution environment. Different technologies are often used in combination to complete the calculation and analysis of data while ensuring the security and privacy of the original data.
[0032] At present, mainstream privacy computing technologies are mainly divided into three categories:
[0033] 1. Multi-party secure computation (MPC): This is a technology that allows multiple parties to jointly compute a target function without a trusted third party, and ensures that each party only obtains its own calculation results and cannot infer the input data of any other party through the interactive data during the calculation process. Unless the function itself can infer the input of other parties from its own input and the obtained output, the input data of other parties is confidential to any party. Common cryptographic technologies used in MPC include obfuscated circuits, unencrypted transmission, secret sharing, homomorphic encryption, etc. By running a secure multi-party computation protocol, multiple parties can collaborate and jointly perform calculations on the data without exposing the plaintext of the data.
[0034] 2. Federated Learning (FL): Federated learning achieves multi-party joint machine learning training by circulating and processing intermediate encrypted data without leaving the local raw data. Participants generally include data parties, algorithm parties, coordination parties, calculation parties, result parties, task initiators, etc. According to the different distribution of the data involved in the calculation among the data parties, federated learning can be divided into horizontal federated learning, vertical federated learning and federated transfer learning.
[0035] 3. Trusted Execution Environment (TEE): A secure area is built in the CPU through software and hardware methods to ensure that the programs and data loaded inside are protected in terms of confidentiality and integrity. TEE provides an isolated execution environment for running sensitive operations or applications, thereby protecting the privacy and security of data. This technology is very useful in scenarios that require high security guarantees, such as financial transactions, electronic voting, etc.
[0036] Existing privacy computing technologies are mainly based on cryptography, machine learning, and secure hardware. Although they can solve privacy protection and secure computing problems in specific scenarios in their respective fields, they all have a common dilemma, which is high threshold, high cost, difficulty in universalization, and difficulty in commercialization. Among them, privacy computing technology based on cryptography still has some subjective leakage risks in the distribution, storage, and use of secret keys; privacy computing technology based on machine learning still has the problem of poor generation model effect when facing the difficult control of training data quality; privacy computing technology based on secure hardware still has the problem of inflexible expansion when facing the computing links that require distributed processing. How to find a low-threshold, low-cost, universal, and commercial computing method under the premise of ensuring the security of privacy data, and maximize the value of data through data sharing, is a core key technical issue that needs to be considered and solved at present.
[0037] Therefore, this application proposes a method for determining the accuracy of prediction information. While ensuring the security of the original data (localization, non-circulation), a highly secure, shareable multi-dimensional vector group data warehouse is formed, so that the statistical analysis results that meet the similarity matching requirements can be output based on the input search condition data that meets the dimension definition, thereby helping to conduct relevant data analysis and determine the accuracy of the prediction information. It provides a low-threshold, low-cost, universal, and commercial privacy computing feasible solution for application scenarios such as medical health, financial risk control, and intelligent recommendations.
[0038] The embodiments of the present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.
[0039] It should be noted that the terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.
[0040] The method embodiments provided in the embodiments of the present application can be executed in a server device or a similar computing device. Taking running on a server device as an example, Figure 1 1 is a hardware structure block diagram of a server device of a method for determining the accuracy of prediction information in an embodiment of the present application. Figure 1 As shown, the server device may include one or more ( Figure 1The server device may include a processor 102 (only one is shown in the figure) (the processor 102 may include but is not limited to a microprocessor MCU or a programmable logic device FPGA and other processing devices) and a memory 104 for storing data, wherein the server device may also include a transmission device 106 and an input / output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1 The structure shown is for illustration only and does not limit the structure of the above-mentioned server device. Figure 1 More or fewer components than shown, or with different Figure 1 Different configurations are shown.
[0041] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the method for determining the accuracy of prediction information in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, to implement the above method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the server device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
[0042] The transmission device 106 is used to receive or send data through a network. The specific example of the above network may include a wireless network provided by a communication provider of the server device. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.
[0043] In this embodiment, a method is provided. Figure 2 is a flow chart of a method for determining the accuracy of prediction information according to an embodiment of the present application. Figure 2 As shown, the process includes the following steps:
[0044] Step S202, obtaining a first vector set, wherein the first vector set includes vectors of N dimensions, and the vectors of N dimensions are vectors obtained by performing vector conversion on information of N dimensions;
[0045] The above-mentioned first vector set can be a filtering condition set, wherein each vector represents a filtering condition of one dimension, and the vector is obtained by vector conversion of each filtering condition (i.e., information); obtaining filtering conditions (i.e., information) of N dimensions, performing vector conversion on the filtering conditions of N dimensions respectively, and obtaining the first vector set.
[0046] Step S204, determining in the target storage a second vector set that does not match the first vector set, and obtaining a first number of second vector sets, wherein the second vector set includes vectors of M dimensions, and the vectors of the M dimensions are vectors obtained by performing vector conversion on information of the M dimensions, and M is greater than N;
[0047] The above-mentioned target storage may be a data warehouse, which is a database system used to centrally store and manage a large amount of structured and unstructured data. By integrating data from different sources and performing processes such as cleaning, conversion and loading, it is convenient for users to easily access and analyze these data; the above-mentioned second vector set may be data stored in the data warehouse, and the vector set includes multiple vectors, each of which is obtained by vector conversion through information of one dimension, that is, the second vector set includes information of multiple dimensions; based on the first vector set, a first number of second vector sets that do not match the first vector set are determined in the target storage.
[0048] As an optional implementation, before determining in the target storage a second vector set that does not match the first vector set, the method includes: acquiring sample information of multiple samples to obtain multiple sample information sets, wherein each of the sample information sets includes information of a sample in M dimensions, and the information in the M dimensions includes information of multiple media types; performing vector conversion on each information in the multiple sample information sets according to the media type to obtain multiple sample vector sets, wherein the sample vector sets include vectors of the M dimensions, and the vectors of the M dimensions are vectors obtained by performing vector conversion on each information in the corresponding sample information sets, and the vector conversion method for sample information of different media types is different; and storing the multiple sample vector sets in the target storage.
[0049] The above sample information can be a multi-dimensional feature of the sample, and the sample information can be information of multiple media types, such as text, audio, picture, and video; the above vector conversion process includes normalization, vectorization, etc., wherein the information vector conversion methods of different media types are also different, which can be performed through multimodal vectorization tools, such as BERT, CLIP, etc.; obtain sample information of multiple samples, obtain a sample information set of each sample, perform vector conversion on each sample information set, and store the sample vector sets obtained after vector conversion in the target storage respectively. Storing sample information through a data warehouse effectively improves the quality of sample information, ensures the accuracy and consistency of sample information, avoids conflicts and inconsistencies of sample information of different media types and from different sources, and is better than storing samples in a unified location for easy management and access. At the same time, when using the stored sample information, it can improve the efficiency of processing and query, reduce the time of information access and analysis, and facilitate data mining, business analysis and decision making, and reduce costs.
[0050] Figure 3 is a sample information storage flow chart according to an embodiment of the present application, such as Figure 3 As shown, sample information of multiple samples is obtained to obtain multiple sample information sets, wherein each sample corresponds to a sample information set, and the sample information set includes information of multiple dimensions. For example, in the field of traditional Chinese medicine diagnosis, each sample can be each patient, and an information set of each patient is obtained. The information in the information set can be the characteristics of each patient obtained after "looking, smelling, asking, and touching", as well as the cause of the disease and the prescription given by the doctor, the efficacy of the prescription, etc. The multiple information in the information set can be recorded in the form of text, or can be recorded in the form of audio description, image display, etc.; or, in the field of equipment maintenance, each sample can be each equipment The information in the information set may be the hardware and / or software data of each device when a failure occurs, such as CPU usage, memory usage, whether the hardware is operating normally, etc., as well as the conclusion and solution of the equipment failure given by the engineer, etc. The multiple information in the information set may be recorded in the form of text, video, audio description, image display, etc.; after obtaining the sample information set, the sample information is vectorized to obtain a sample vector set of each sample, and each sample vector set is stored in a data warehouse for subsequent data mining, business analysis, and decision making.
[0051] Step S206, obtaining prediction information of the target object based on the information of the N dimensions, performing vector conversion on the prediction information to obtain a prediction vector, and determining the vectors of the N dimensions and the prediction vector as a third vector set, wherein the prediction information is information of the first dimension, and the M dimensions include the first dimension;
[0052] The target object may be a new sample other than the samples stored in the target storage, such as a new patient or a new machine, etc.; the prediction information may be prediction information obtained by the target object based on information of N dimensions represented by the vector in the first vector set. For example, when the first vector set includes vectors of four dimensions of “observation, smell, questioning, and palpation”, the prediction information may be the cause of disease given by the doctor based on information of the four dimensions of “observation, smell, questioning, and palpation”; or, when the first vector set includes vectors of three dimensions of CPU usage, memory usage, and whether the hardware is operating normally when a device fails, the prediction information may be the cause of failure given by the engineer based on information of three dimensions of CPU usage, memory usage, and whether the hardware is operating normally; the prediction information obtained by the target object based on information of N dimensions is obtained, and the prediction information is vector-converted to obtain a prediction vector, and the vectors of N dimensions in the first vector set and the prediction vector are determined as a third vector set.
[0053] Step S208, determining in the target storage a fourth vector set that does not match the third vector set, and obtaining a second number of fourth vector sets, wherein the fourth vector set includes vectors of the M dimensions;
[0054] Step S210: determining the accuracy of the prediction information according to the first number and the second number.
[0055] Optionally, determining the accuracy of the prediction information based on the first number and the second number includes: determining a ratio of the second number to the first number as the accuracy of the prediction information.
[0056] Through the above steps, after the prediction information of the target object is determined based on the information of N dimensions represented by the vectors of N dimensions in the first vector set, the prediction information is vector-converted to obtain the prediction vector, and the third vector set is determined based on the prediction vector and the first vector set, and the accuracy of the prediction information is determined based on the first number of vector sets that do not match the first vector set and the second number that do not match the third vector set determined in the target storage, so that the accuracy of the prediction information can be determined by using only the relevant information through the vector set in the target storage while ensuring the security of the original data and preventing it from being leaked. The problem in the related art that data analysis cannot be realized without leaking the original data, thereby determining the accuracy of the prediction information, is solved, and data security is improved.
[0057] The execution subject of the above steps may be a server, a terminal, etc., but is not limited thereto.
[0058] The execution order of step S204 and step S206 is interchangeable, that is, step S206 may be executed first, and then step S204.
[0059] As an optional implementation, each information in a plurality of sample information sets is vector-converted according to the media type to obtain a plurality of sample vector sets, including: performing the following operations on each sample information in the sample information set, and taking the sample information subjected to the following operations as the current sample information: when the media type of the current sample information is a text type, performing a vector conversion on the current sample information in accordance with a text vector conversion method to obtain a current sample vector of the current sample information; when the type of the current sample information is an image type, performing a vector conversion on the current sample information in accordance with an image vector conversion method to obtain a current sample vector of the current sample information; when the media type of the current sample information is an audio type, performing a vector conversion on the current sample information in accordance with an audio vector conversion method to obtain a current sample vector of the current sample information; when the type of the current sample information is a video type, performing a vector conversion on the current sample information in accordance with a video vector conversion method to obtain a current sample vector of the current sample information.
[0060] The sample information of the above text type can be vectorized by using the TF-IDF (Term Frequency-Inverse Document Frequency) formula; the sample information of the above image type can be vectorized by performing binarization, transposition and other operations on the pixel matrix of the image; the sample information of the above audio type can be vectorized by calculating the sampling rate, number of channels, quantization, time and other aspects of the audio; the sample information of the above frequency type can be vectorized by extracting features and encoding the features. Since different vector conversion methods are used for different types of media, the sample information becomes a vector with the same structure, so that the user cannot obtain the original information by performing inverse operations on the vector without knowing the media type of the original information, thereby ensuring the privacy and security of the original data.
[0061] Optionally, storing the multiple sample vector sets in the target storage includes: performing the following operations on each sample vector set in the multiple sample vector sets, and using the sample vector set subjected to the following operations as a first current sample vector set, wherein the first current sample vector set is a sample vector set obtained by performing the vector conversion on sample information of the current sample: determining the sample type of the current sample as the current sample type, wherein the multiple samples include multiple sample types; determining a current hypercube model that does not match the current sample type in the target storage, wherein the target storage contains multiple hypercube models, wherein the multiple hypercube models include the current hypercube model, each of the hypercube models is used to store a sample vector set of the same sample type, and each point on the hypercube model is used to store a sample vector set; and storing each sample vector in the first current sample vector set into the current hypercube model.
[0062] The sample type may be the subject of the sample or the category to which it belongs, for example, in the field of traditional Chinese medicine diagnosis, the sample type may be dizziness and headache, chest tightness and shortness of breath, limb weakness, etc. In the field of equipment maintenance, the sample type may be a black screen of the device, a system crash, or a device that cannot be started, etc.; the hypercube model may be a hypercube model, which is a multidimensional data model consisting of multiple dimensions and metrics. The dimensions describe the characteristics or attributes of the data, and the metrics are data indicators that need to be analyzed and calculated. In this embodiment, the hypercube model may be a structure used to organize and store data in a data warehouse (i.e., target storage). Specifically, the hypercube model may be mapped to the data warehouse, wherein the dimensions of the hypercube model may be mapped to the fields of the data warehouse, the vector groups may be mapped to the records of the data warehouse, and the vectors may be mapped to the field values corresponding to the records of the data warehouse, thereby realizing the logical loading and physical storage of the hypercube model; through the hypercube model, users can conveniently perform multidimensional data analysis, data slicing, dicing, drilling, and other operations, so as to better understand the data and make effective decisions. Determine the type of the current sample, determine the hypercube model corresponding to the type in the target storage, and store the sample vector set of the current sample in the corresponding hypercube model. The data warehouse and the hypercube model complement each other. The data warehouse provides the basis for data storage and management, while the hypercube model provides tools for data analysis and query, helping users better understand and use data.
[0063] Specifically, taking the three-dimensional hypercube model as an example, Figure 4 is a schematic diagram of the structure of a three-dimensional hypercube model according to an embodiment of the present application, such as Figure 4 As shown, the three dimensions of the hypercube model are Vx, Vy, and Vz, wherein each node of the hypercube model stores a sample vector set of a sample, and each dimension can be mapped to a field of the data warehouse, that is, the specific meaning of each dimension, for example, "look", "smell", "ask", "cut", cause of disease, etc., or CPU usage, memory usage, whether the hardware is running normally, cause of failure, etc., each vector in the sample vector set is mapped to a specific dimension, for example, Vx is mapped to the specific feature of "look", Vy is mapped to the specific feature of "smell", Vz is mapped to the specific feature of the cause of disease, etc., or Vx is mapped to CPU usage, Vy is mapped to whether the hardware is running normally, Vz is mapped to the cause of failure, etc.
[0064] As an optional implementation, the determining in the target storage a second vector set that does not match the first vector set includes: determining the type of the target object as the target type; determining a target hypercube model that does not match the target type from a plurality of the hypercube models in the target storage; traversing each node in the target hypercube model, and determining as the second vector set a sample vector set stored on each node whose similarity to the first vector set is greater than or equal to a preset similarity threshold.
[0065] The type of the target object may be the subject of the target object or the category to which it belongs, for example, in the field of traditional Chinese medicine diagnosis, the type of the target object may be dizziness and headache, chest tightness and shortness of breath, limb weakness, etc. In the field of equipment maintenance, the type of the target object may be a black screen of the device, a system crash, or a device unable to start, etc.; the similarity calculation method may be a k-nearest neighbor algorithm (k-Nearest Neighbors, KNN) or a cosine similarity calculation method; a hypercube model matching the type of the target object is determined in the target storage, each node in the hypercube model is traversed, and a sample vector set stored on a node whose similarity to the first vector set is greater than or equal to a preset similarity threshold is obtained, wherein the target storage may map hypercube models of multiple subjects (such as Figure 3 By selecting the appropriate similarity calculation method and similarity threshold, adjustments and optimizations can be made according to specific application scenarios and requirements.
[0066] As an optional implementation, before determining the sample vector set stored on the nodes whose similarity with the first vector set is greater than or equal to a preset similarity threshold in each of the nodes as the second vector set, the method further includes: calculating the similarity between the first vector set and the sample vector set stored on each node in the target hypercube model in the following manner, including: traversing each node in the target hypercube model, taking the currently traversed node as the current node, and taking the sample vector set stored on the current node as the second current sample vector set; determining the vectors of N dimensions that match the vectors of N dimensions of the first vector set from the vectors of M dimensions in the second current sample vector set to obtain a second current sample vector subset; determining the vectors of N dimensions of the first vector set that match the vectors of N dimensions of the second current sample vector set to obtain a second current sample vector subset; The similarity between the vectors of N dimensions in the subset is determined to obtain N similarities; the sum of the N similarity values is determined as the similarity between the first vector set and the second current sample vector set stored on the current node; wherein, the similarity between the vectors of N dimensions in the first vector set and the vectors of N dimensions in the subset of the second current sample vector is determined to obtain N similarities, including: performing the following operations on each vector of the N dimensions in the first vector set, and the vectors performing the following operations are used as the first current vector, and the first current vector is the vector of the current dimension; determining the vector of the current dimension in the subset of the second current sample vector to obtain the second current vector, and calculating the similarity between the first current vector and the second current vector to obtain the current similarity, wherein the N similarities include the current similarity.
[0067] Optionally, the above calculation of the similarity between the sets can directly calculate the similarity between the first vector set and the vector set of the current sample vector subset; the above calculation of the similarity between the vector sets can also use parallel computing technology, that is, respectively calculating the similarity between the vectors of each dimension in the N dimensions and the vectors of the corresponding dimension in the current sample vector subset to obtain the similarity of N dimensions, and then performing sum operations on the similarities of N dimensions to obtain the similarity between the sample vector set in each node and the first vector set.
[0068] By computing the similarities between vector sets in parallel, we can utilize hardware resources such as multi-core processors and GPUs to further speed up the calculations. Especially when processing high-dimensional data, we can greatly shorten the calculation time and speed up the calculations, thereby improving processing efficiency.
[0069] As an optional implementation, determining in the target storage a fourth vector set that does not match the third vector set includes: determining the type of the target object as a target type; determining a target hypercube model that does not match the target type from a plurality of hypercube models in the target storage; traversing each node in the target hypercube model, and determining as the fourth vector set a sample vector set stored on each node whose similarity to the third vector set is greater than or equal to a preset similarity threshold.
[0070] As an optional implementation, after determining the accuracy of the prediction information, the method further includes: when the accuracy ratio of the prediction information is greater than or equal to a preset threshold, determining the prediction information as the diagnosis result of the target object, and outputting the prediction information, wherein the information of N dimensions corresponding to the vectors of the N dimensions in the first vector set is the detection information of the target object; when the accuracy ratio of the prediction information is less than the preset threshold, outputting a prompt message, wherein the prompt message is used to indicate that the prediction information is erroneous information.
[0071] As an optional implementation, Figure 5 It is a schematic diagram of the overall process according to the embodiment of the present application, such as Figure 5 The specific mileage includes:
[0072] Step S1, design a hypercube model of a multidimensional vector group. Hypercube means using three or more dimensions to describe an object, with each dimension being perpendicular to each other. Taking a three-dimensional cube as an example, X, Y and Z represent three different dimensions, and Vx, Vy and Vz represent the vector values in each dimension. Then the vector set {Vx, Vy, Vz} represents a point in the three-dimensional cube model. Each point in the cube is a multidimensional vector group, and the cube dimension corresponds to the vector group dimension one-to-one. By analogy, the three-dimensional cube is expanded to the multidimensional cube, and a hypercube model for a multidimensional vector group is designed and constructed.
[0073] Step S2, using the data warehouse, completes the logical loading and physical storage of the hypercube model. A data warehouse is a warehouse for storing data, which stores various kinds of data, and these data need to be organized and stored according to some structures and rules. Based on the characteristics and features of the data warehouse, the hypercube model can be mapped to the data warehouse, the dimensions of the hypercube model can be mapped to the fields of the data warehouse, the vector group can be mapped to the records of the data warehouse, and the vector can be mapped to the field values corresponding to the records of the data warehouse, so that the logical loading and physical storage of the hypercube model are realized.
[0074] Step S3, design VSS-SQL language and algorithm to implement statistical analysis operations on hypercube model data. VSS-SQL is defined as a structured query language for vector similarity search, which consists of a set of SQL-like languages, language parsers and VSS algorithm sets.
[0075] SQL-like languages allow users to work on high-level data structures. They do not require users to specify data access methods, nor do they require users to understand specific data access methods. They are mainly responsible for defining input, executing operations, and outputting results in a structured manner. The following is an example:
[0076] (1) Storage operation definition:
[0077] Surprise:
[0078] INSERT INTO <t>VALUES(<V1,V2,……,Vn> );
[0079] illustrate:
[0080] INSERT INTO: instruction, indicating the operation of storing vector data;
[0081] <t>: variable, the value is the subject to be stored;
[0082] VALUES(): instruction, indicating the stored value;
[0083] <V1,V2,……,Vn> : variable, the value is the vector group with storage;
[0084] (2) Take the operational definition:
[0085] Surprise:
[0086] SELECT COUNT() FROM <t>WHERE <f1> ∽( <v1> , <s1> )&& <f1> ∽( <v1> , <s1>);
[0087] illustrate:
[0088] SELECT COUNT()FROM: instruction, indicating the operation of retrieving statistical results;
[0089] <t>: variable, the value is the subject to be taken out;
[0090]
[0091] The language parser is a low-level middleware program that converts the SQL-like language edited by the user into an operable data warehouse and executable VSS algorithm set. During the storage process, the language parser maps the SQL-like language to the data warehouse language and inserts the data records into the hypercube model by subject; during the retrieval process, the language parser maps the SQL-like language to the data warehouse language, firstly retrieves the data records from the hypercube model by subject and temporarily saves them to the memory; secondly, executes the VSS algorithm set from the memory according to the conditions; finally, returns the execution result after the operation to the user. The language parser pseudo code is as follows:
[0092]
[0093] Int getSQL(String sql){
[0094] / / Based on the incoming SQL, extract the corresponding instructions
[0095] String cmd = cmd(sql);
[0096] / / According to the instructions, judge and execute the corresponding operations
[0097] switch(cmd){
[0098] / / Store instructions and execute operations
[0099] case INSERT:return insert(sql);
[0100] / / Store instructions and execute operations
[0101] case SELECT:return select(sql);}
[0102] }
[0103] / / Judgement instruction
[0104] String cmd(sql){
[0105] / / Separate by spaces and select the first string to return return sql.split("")[0];
[0106] }
[0107] / / Deposit method
[0108] Int insert(String sql){
[0109] / / Map to data warehouse language
[0110] String d_sql=mapping(sql);
[0111] / / Call the data warehouse to save to the hypercube model
[0112] Int result = save(d_sql);
[0113] return result;
[0114] }
[0115]
[0116] The VSS algorithm set consists of a group of algorithms, and its purpose is to complete the search and statistical calculation work based on given conditions. It mainly involves three links: hypercube model data traversal, conditional vector similarity comparison, and statistical result counting. Considering the support for multi-dimensional, big data, and distributed, for hypercube model data traversal, the block forward traversal algorithm is selected; for conditional vector similarity comparison, the Yuxuan similarity algorithm is selected; for statistical result counting, the MapReduce programming model is selected. The best implementation is to traverse the records (multidimensional vector group) in the hypercube model in blocks and in parallel one by one, and judge each record according to the dimension where the given condition is located. If the judgment result meets the expectation (similarity is greater than or equal to the threshold), the statistical counter is increased by 1, and all traversals are completed to obtain the best statistical results.
[0117] Among them, the block forward order traversal algorithm traverses the hypercube model from top to bottom one by one until each record (each sample) is traversed once; the cosine similarity algorithm uses the cosine value of the angle between two vectors in the vector space as a measure of the size of the difference between two individuals. -1 means that the directions of the two vectors are exactly opposite, 1 means that their directions are exactly the same, and 0 usually means that they are independent, and the value in between indicates the similarity or difference between them; the MapReduce programming model is used for parallel computing of large-scale data sets. Based on the block forward order traversal algorithm and the cosine similarity algorithm, the input meets the vectorized search conditions defined by the dimension, and the output meets the statistical analysis results that meet the similarity matching requirements.
[0118] Step S4, define the type (topic) and statistical analysis dimension according to the business application needs and the private original data (sample information set) held. Take TCM consultation as an example to explain the process of defining topics and dimensions. First, the patient is asked about the symptoms, and then the TCM doctor determines the cause of the patient's disease through "looking, smelling, asking, and palpating", and finally gives a prescription and tracks the efficacy. Among them, symptoms (dizziness and headache, chest tightness and shortness of breath, limb weakness, etc.) are the themes, and "looking, smelling, asking, palpating, cause of disease, prescription, efficacy" are dimensions, and each dimension can be data of different modalities but with the same meaning.
[0119] Step S5, using a multimodal vectorization tool, normalizes and vectorizes the private multimodal data to obtain irreversible secure vector data after conversion.
[0120] Step S6, using VSS-SQL, performs storage operations on the hypercube model data, and stores the processed vector set into the hypercube model item by item according to the subject and dimension, forming a multi-subject, multi-dimensional, highly secure, and shareable multi-dimensional vector group data warehouse.
[0121] Step S7, using VSS-SQL, perform operations on the hypercube model data, input vectorized search conditions that meet the dimension definition, and output statistical analysis results that meet the similarity matching requirements.
[0122] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, disk, CD), including a number of instructions to enable a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in each embodiment of the present application.
[0123] In this embodiment, a device for determining the accuracy of prediction information is provided, and the device is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made are omitted. As used below, the term "module" can implement a combination of software and / or hardware for a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.
[0124] Figure 6 is a structural block diagram of a device for determining the accuracy of prediction information according to an embodiment of the present application, such as Figure 6 As shown, the device includes: a first acquisition module 602, used to acquire a first vector set, wherein the first vector set includes vectors of N dimensions, and the vectors of the N dimensions are vectors obtained by performing vector transformation on information of the N dimensions; a first determination module 604, used to determine a second vector set that does not match the first vector set in the target storage, and obtain a first number of second vector sets, wherein the second vector set includes vectors of M dimensions, and the vectors of the M dimensions are vectors obtained by performing vector transformation on information of the M dimensions, and M is greater than N; a second acquisition module 606, used to acquire the target object based on the N dimensions. The method comprises the following steps: obtaining prediction information from information of the first dimension and performing vector conversion on the prediction information to obtain a prediction vector, determining the vectors of the N dimensions and the prediction vector as a third vector set, wherein the prediction information is information of the first dimension, and the M dimensions include the first dimension; a second determination module 608 is used to determine a fourth vector set that does not match the third vector set in the target storage, and obtain a second number of fourth vector sets, wherein the fourth vector set includes the vectors of the M dimensions; a third determination module 610 is used to determine the accuracy of the prediction information according to the first number and the second number.
[0125] In an exemplary embodiment, the device is used to obtain sample information of multiple samples to obtain multiple sample information sets, wherein each of the sample information sets includes information of a sample in M dimensions, and the information in the M dimensions includes information of multiple media types; performing vector conversion on each information in the multiple sample information sets according to the media type to obtain multiple sample vector sets, wherein the sample vector sets include vectors of the M dimensions, and the vectors of the M dimensions are vectors obtained by performing vector conversion on each information in the corresponding sample information sets, and the vector conversion method for sample information of different media types is different; and storing the multiple sample vector sets in the target storage.
[0126] In an exemplary embodiment, the device is further configured to perform the following operations on each sample information in the sample information set, and the sample information subjected to the following operations is used as the current sample information: when the media type of the current sample information is a text type, performing vector conversion on the current sample information in accordance with a text vector conversion method to obtain a current sample vector of the current sample information; when the type of the current sample information is an image type, performing vector conversion on the current sample information in accordance with an image vector conversion method to obtain a current sample vector of the current sample information; when the media type of the current sample information is an audio type, performing vector conversion on the current sample information in accordance with an audio vector conversion method to obtain a current sample vector of the current sample information; when the type of the current sample information is a video type, performing vector conversion on the current sample information in accordance with a video vector conversion method to obtain a current sample vector of the current sample information.
[0127] In an exemplary embodiment, the device is further configured to perform the following operations on each sample vector set in the multiple sample vector sets, and the sample vector set performing the following operations is used as a first current sample vector set, wherein the first current sample vector set is a sample vector set obtained by performing the vector conversion on sample information of the current sample: determining the sample type of the current sample as the current sample type, wherein the multiple samples include multiple sample types; determining a current hypercube model that does not match the current sample type in the target storage, wherein the target storage contains multiple hypercube models, the multiple hypercube models include the current hypercube model, each of the hypercube models is used to store a sample vector set of the same sample type, and each point on the hypercube model is used to store a sample vector set; and storing each sample vector in the first current sample vector set into the current hypercube model.
[0128] In an exemplary embodiment, the device is also used to determine the type of the target object as the target type; determine a target hypercube model that does not match the target type among the multiple hypercube models stored in the target; traverse each node in the target hypercube model, and determine the sample vector set stored on the node whose similarity with the first vector set is greater than or equal to a preset similarity threshold in each node as the second vector set.
[0129] In an exemplary embodiment, the device is also used to calculate the similarity between the first vector set and the sample vector set stored on each node in the target hypercube model in the following manner, including: traversing each node in the target hypercube model, taking the currently traversed node as the current node, and taking the sample vector set stored on the current node as the second current sample vector set; determining the vectors of N dimensions that match the vectors of N dimensions of the first vector set from the vectors of M dimensions in the second current sample vector set to obtain a second current sample vector subset; determining the vectors of N dimensions of the first vector set that match the vectors of the second current sample vector subset; The similarities between vectors of N dimensions in the set are calculated to obtain N similarities; the sum of the N similarity values is determined as the similarity between the first vector set and the second current sample vector set stored on the current node; wherein, the following operations are performed on each vector of the N dimensions in the first vector set, and the vector subjected to the following operations is used as the first current vector, and the first current vector is the vector of the current dimension; the vector of the current dimension is determined in the second current sample vector subset to obtain the second current vector, and the similarity between the first current vector and the second current vector is calculated to obtain the current similarity, wherein the N similarities include the current similarity.
[0130] In an exemplary embodiment, the device is further used to determine the ratio of the second number to the first number as the accuracy of the prediction information; after determining the accuracy of the prediction information, when the accuracy ratio of the prediction information is greater than or equal to a preset threshold, determine the prediction information as the diagnosis result of the target object, and output the prediction information, wherein the information of N dimensions corresponding to the vectors of the N dimensions in the first vector set is the detection information of the target object; when the accuracy ratio of the prediction information is less than the preset threshold, output a prompt message, wherein the prompt message is used to indicate that the prediction information is erroneous information.
[0131] It should be noted that the above modules can be implemented by software or hardware. For the latter, it can be implemented in the following ways, but not limited to: the above modules are all located in the same processor; or the above modules are located in different processors in any combination.
[0132] An embodiment of the present application also provides a computer-readable storage medium, in which a computer program is stored, wherein the computer program is configured to execute the steps of any of the above method embodiments when running.
[0133] In an exemplary embodiment, the computer-readable storage medium may include, but is not limited to, various media that can store computer programs, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.
[0134] An embodiment of the present application also provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any of the above method embodiments.
[0135] In an exemplary embodiment, the electronic device may include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.
[0136] An embodiment of the present application also provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the steps in any of the above method embodiments are implemented.
[0137] For specific examples in this embodiment, reference may be made to the examples described in the above embodiments and exemplary implementation modes, and this embodiment will not be described in detail herein.
[0138] Obviously, those skilled in the art should understand that the above modules or steps of the present application can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. In this way, the present application is not limited to any specific combination of hardware and software.
[0139] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the principles of the present application shall be included in the protection scope of the present application.< / t> < / v1> < / f1> < / s1> < / v1> < / f1> < / t> < / t> < / t>
Claims
1. A method for determining the accuracy of forecast information, It is characterized in that include: Acquire a first vector set, wherein the first vector set includes vectors of N dimensions, and the vectors of N dimensions are vectors obtained by performing vector conversion on information of N dimensions; Determine in the target storage a second vector set that matches the first vector set, and obtain a first number of second vector sets, wherein the second vector set includes vectors of M dimensions, and the vectors of M dimensions are vectors obtained by performing vector conversion on information of M dimensions, where M is greater than N; Acquire prediction information of the target object based on the information of the N dimensions, perform vector conversion on the prediction information to obtain a prediction vector, and determine the vectors of the N dimensions and the prediction vector as a third vector set, wherein the prediction information is information of the first dimension, and the M dimensions include the first dimension; Determining in the target storage a fourth vector set that matches the third vector set, to obtain a second number of fourth vector sets, wherein the fourth vector set includes vectors of the M dimensions; The accuracy of the prediction information is determined according to the first number and the second number.
2. The method according to claim 1, It is characterized in that Before determining in the target storage a second vector set matching the first vector set, the method includes: Acquire sample information of multiple samples to obtain multiple sample information sets, wherein each of the sample information sets includes information of one sample in M dimensions, and the information in the M dimensions includes information of multiple media types; Performing vector conversion on each information in the plurality of sample information sets according to the media type to obtain a plurality of sample vector sets, wherein the sample vector sets include the vectors of the M dimensions, and the vectors of the M dimensions are vectors obtained by performing vector conversion on each information in the corresponding sample information sets, and the vector conversion methods of sample information of different media types are different; The plurality of sample vector sets are stored in the target storage.
3. The method according to claim 2, It is characterized in that Performing vector conversion on each information in the plurality of sample information sets according to the media type to obtain a plurality of sample vector sets, including: Perform the following operations on each sample information in the sample information set, and use the sample information subjected to the following operations as the current sample information: When the media type of the current sample information is a text type, performing vector conversion on the current sample information according to a text vector conversion method to obtain a current sample vector of the current sample information; When the type of the current sample information is an image type, performing vector conversion on the current sample information according to an image vector conversion method to obtain a current sample vector of the current sample information; When the media type of the current sample information is an audio type, performing vector conversion on the current sample information according to an audio vector conversion method to obtain a current sample vector of the current sample information; In the case that the type of the current sample information is a video type, vector conversion is performed on the current sample information according to a video vector conversion method to obtain a current sample vector of the current sample information.
4. The method according to claim 3, It is characterized in that Storing the plurality of sample vector sets in the target storage comprises: Performing the following operations on each sample vector set in the multiple sample vector sets, and using the sample vector set that performs the following operations as a first current sample vector set, wherein the first current sample vector set is a sample vector set obtained by performing the vector conversion on sample information of a current sample: Determining a sample type of the current sample as a current sample type, wherein the plurality of samples include a plurality of sample types; Determining a current hypercube model matching the current sample type in the target storage, wherein the target storage contains a plurality of hypercube models, the plurality of hypercube models include the current hypercube model, each of the hypercube models is used to store a sample vector set of the same sample type, and each point on the hypercube model is used to store a sample vector set; Each sample vector in the first current sample vector set is stored in the current hypercube model.
5. The method according to claim 4, It is characterized in that The determining, in the target storage, a second vector set matching the first vector set comprises: Determine the type of the target object as a target type; Determining a target hypercube model matching the target type among the plurality of hypercube models stored in the target; Each node in the target hypercube model is traversed, and a sample vector set stored in a node whose similarity with the first vector set is greater than or equal to a preset similarity threshold among the nodes is determined as the second vector set.
6. The method according to claim 5, It is characterized in that Before determining the sample vector sets stored on the nodes whose similarity with the first vector set is greater than or equal to a preset similarity threshold among the nodes as the second vector set, the method further includes: The similarity between the first vector set and the sample vector set stored on each node in the target hypercube model is calculated in the following manner, including: Traverse each node in the target hypercube model, take the currently traversed node as the current node, and use the sample vector set stored in the current node as the second current sample vector set: Determine, from the vectors of M dimensions in the second current sample vector set, vectors of N dimensions that match the vectors of N dimensions in the first vector set, to obtain a second current sample vector subset; Determine similarities between vectors of N dimensions in the first vector set and vectors of N dimensions in the second current sample vector subset to obtain N similarities; Determine the sum of the N similarity values as the similarity between the first vector set and the second current sample vector set stored on the current node; Wherein, determining the similarity between the vectors of N dimensions in the first vector set and the vectors of N dimensions in the second current sample vector subset to obtain N similarities includes: Perform the following operations on each vector of the N-dimensional vectors in the first vector set, and use the vectors subjected to the following operations as first current vectors, where the first current vector is a vector of the current dimension; Determine the vector of the current dimension in the second current sample vector subset to obtain a second current vector, calculate the similarity between the first current vector and the second current vector to obtain a current similarity, wherein the N similarities include the current similarity.
7. The method according to claim 1, It is characterized in that The determining the accuracy of the prediction information according to the first number and the second number includes: determining a ratio of the second number to the first number as the accuracy of the prediction information; After determining the accuracy of the prediction information, the method further includes: In a case where the accuracy ratio of the prediction information is greater than or equal to a preset threshold, determining the prediction information as a diagnosis result of the target object, and outputting the prediction information, wherein the information of N dimensions corresponding to the vectors of the N dimensions in the first vector set is the detection information of the target object; When the accuracy ratio of the prediction information is less than the preset threshold, a prompt message is output, wherein the prompt message is used to indicate that the prediction information is erroneous information.
8. A device for determining the accuracy of prediction information, It is characterized in that include: A first acquisition module, configured to acquire a first vector set, wherein the first vector set includes vectors of N dimensions, and the vectors of N dimensions are vectors obtained by performing vector conversion on information of N dimensions; A first determination module is configured to determine, in a target storage, a second vector set matching the first vector set, to obtain a first number of second vector sets, wherein the second vector set includes vectors of M dimensions, and the vectors of M dimensions are vectors obtained by performing vector conversion on information of M dimensions, where M is greater than N; A second acquisition module is used to obtain prediction information of the target object based on the information of the N dimensions, and perform vector conversion on the prediction information to obtain a prediction vector, and determine the vectors of the N dimensions and the prediction vector as a third vector set, wherein the prediction information is information of the first dimension, and the M dimensions include the first dimension; a second determination module is used to determine a fourth vector set matching the third vector set in the target storage to obtain a second number of fourth vector sets, wherein the fourth vector set includes the vectors of the M dimensions; a third determination module is used to determine the accuracy of the prediction information based on the first number and the second number.
9. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores a computer program, wherein the computer program implements the steps of the method described in any one of claims 1 to 7 when executed by a processor.
10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the computer program, the steps of the method described in any one of claims 1 to 7 are implemented.
