Multi-objective manifold data analysis method, device, equipment and storage medium

Abstract

The application relates to a multi-target manifold data analysis method, device, equipment and storage medium. The architecture comprises the following steps: acquiring multi-target manifold data; inputting the multi-target manifold data into a pre-trained manifold data analysis model to obtain an analysis result of the multi-target manifold data; and training an initial analysis model in a manifold space by using sample manifold data and a sample analysis sequence corresponding to the sample manifold data. According to the embodiment of the application, the initial analysis model in the manifold space is trained by using the sample manifold data and the sample analysis sequence corresponding to the sample manifold data, so that the manifold data analysis model can be used to well analyze high-dimensional manifold data, the accuracy of the manifold data analysis is ensured, the manifold data analysis model can be applied to massive high-dimensional manifold data analysis, and the manifold data analysis model has good popularization and application value.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of big data analysis, and in particular to a multi-target manifold data analysis method, device, equipment and storage medium. BACKGROUND

[0002] In the era of big data, many types of data are affected by multi-dimensional characteristics, making the complexity of data higher and higher. The data with high complexity can be referred to as manifold data. For example, the dimensions of movie data include the dimensions of producer, shooting time, end time, sponsor company, actor and release time.

[0003] For a large amount of manifold data, users want to efficiently and accurately analyze the manifold data. However, the current data analysis means can effectively analyze low-dimensional data, and the analysis accuracy of high-dimensional manifold data is very low.

[0004] The manifold data analysis method is a generalized optimization solution algorithm, which finds the minimum value of the unit logarithmic likelihood function J(θ) under the constraint condition on the manifold, where θ is a model parameter set. J(θ) is defined as:

[0005]

[0006] Where y is the real data observation value, p is the model prediction value, ln() represents the natural logarithm function, <·> represents the mean function; λ is a constant, w is the weight of each group, and ρ is the signal manifold space, represents the group summation function.

[0007] The model prediction value p is a function of the real data x and the parameter set θ, and can be expressed according to the Fermi-Dirac distribution function as:

[0008]

[0009] Where e represents the exponential function, and f(x|θ) is as follows:

[0010] f(x|θ)=θ μ δ μ (x-α δ )+o(…)

[0011] Where x represents the real data, θ μ represents a variable parameter vector; δ μ (x-α δ ) is a row vector composed of Dirac delta functions (i.e. the group parameter vector corresponding to each variable); o(…) represents other terms, such as high-order infinitesimals, or neural network models, etc.

[0012] The signal manifold space p is a function on the packet set d defined as

[0013]

[0014] where d represents a group, <y> δ the mean of the real data observations over the groups, δ the mean of the model prediction values on each group, ln() represents a natural logarithm function.

[0015] The gradient vector and Hessian matrix of the unit log-likelihood function J(θ) are respectively:

[0016]

[0017]

[0018] wherein μ,v represents a variable, δ μ ,δ μ represents the group parameter corresponding to the variable, <·> represents a mean function; λ is a constant, w is the weight of each group, and ρ is the signal manifold space, represents a group summation function.

[0019] The iterative solution of the nonlinear optimization problem is:

[0020]

[0021]

[0022] wherein Δθ μ is the incremental vector of the variable parameter update in each iteration, represents the inverse matrix of , and Δα represents the row vector of . Δα δ is the incremental vector of the group parameter update in each iteration, β is a constant, ln() represents a natural logarithm function, and <·> represents a mean function. SUMMARY

[0023] In order to solve the above technical problems or at least partially solve the above technical problems, the present application provides a manifold data analysis method, device, equipment and storage medium of a multi-target manifold architecture.

[0024] In a first aspect, the present application provides a multi-target manifold data analysis method, which comprises:

[0025] acquiring multi-target manifold data;

[0026] inputting the multi-target manifold data into a pre-trained manifold data analysis model to obtain

[0027] an analysis result of the multi-target manifold data, wherein the manifold data analysis model is trained by using sample manifold data and a sample analysis sequence corresponding to the sample manifold data on an initial analysis model in a manifold space.

[0028] In a second aspect, the present application provides a multi-target manifold data analysis device, which comprises:

[0029] a multi-target manifold data acquisition module, configured to acquire multi-target manifold data;

[0030] an analysis result determination module, configured to input the multi-target manifold data into a pre-trained manifold data analysis model to obtain an analysis result of the multi-target manifold data, wherein the manifold data analysis model is trained based on sample manifold data and a sample analysis sequence corresponding to the sample manifold data.

[0031] In a third aspect, the present application further provides a multi-target manifold data analysis equipment, which comprises:

[0032] one or more processors;

[0033] a storage device, configured to store one or more programs,

[0034] when the one or more programs are executed by the one or more processors, the one or more processors implement the manifold data analysis method provided in the first aspect.

[0035] In a fourth aspect, the present application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the manifold data analysis method provided in the first aspect.

[0036] Compared with the prior art, the technical solution provided by the present application has the following advantages:

[0037] The multi-target manifold architecture manifold data analysis method, device, equipment and storage medium provided by the embodiment of the present application can input the multi-target manifold data into a pre-trained manifold data analysis model to obtain an analysis result of the multi-target manifold data after acquiring the multi-target manifold data, and since the manifold data analysis model is trained based on sample manifold data and a sample analysis sequence corresponding to the sample manifold data, the manifold data analysis model can be used to well analyze high-dimensional manifold data, ensure the accuracy of manifold data analysis, can be applied to massive high-dimensional manifold data for analysis, and has good popularization and application value. BRIEF DESCRIPTION OF DRAWINGS

[0038] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0039] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, for those skilled in the art, other drawings can also be obtained based on these drawings without any creative effort.

[0040] Figure 1 The flowchart of the manifold data analysis method provided by the embodiment of the present application is shown in the figure.

[0041] Figure 2 The detailed flowchart of another manifold data analysis method provided by the embodiment of the present application is shown in the figure.

[0042] Figure 3 The structure diagram of the manifold data analysis device provided by the embodiment of the present application is shown in the figure.

[0043] Figure 4 The structure diagram of the manifold data analysis device provided by the embodiment of the present application is shown in the figure.

[0044] Figure 5 The structure diagram of the field programmable logic gate array provided by the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, for those skilled in the art, other drawings can also be obtained based on these drawings without any creative effort.

[0046] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, but the present application can also be implemented in other different ways from those described herein; obviously, the embodiments in the specification are only some embodiments of the present application, not all embodiments.

[0047] Big data is ubiquitous in today's world, and has various forms, including familiar text, pictures, audio, video, etc. After digitization, the basic types of data are divided into integers, floating points and characters, and the forms of data storage can be scalars, vectors, tensors, etc.

[0048] For massive manifold data, the inventor found that the popular data generated by many events can be transformed into a nonlinear optimization problem. The classic algorithms for nonlinear optimization problems include gradient descent method and Newton method. The gradient descent method is a first-order convergence, that is, a plane topology is used to fit the local manifold of the current position. The Newton method is a second-order convergence, that is, a second-order topological surface is used to fit the current local manifold. Since the descent path involved in the Newton method is more consistent with the real optimal descent path. Therefore, for nonlinear optimization problems, the Newton method is faster.

[0049] In order to improve the analysis accuracy of high-dimensional manifold data, based on the above finding, the application adopts an optimization solving algorithm in a topological space to find the minimum value of a unit log-likelihood function through a first-order gradient vector and a second-order Hessian matrix on the manifold.

[0050] Based on the above analysis, the application provides a manifold data analysis method and device based on a multi-target manifold architecture, a device, and a storage medium.

[0051] Figure 1 A flowchart of the manifold data analysis method provided by the application is shown.

[0052] In some embodiments of the application, Figure 1 The manifold data analysis method shown can be executed by an electronic device or a server. The electronic device can include, but is not limited to, a mobile terminal such as a smartphone, a notebook computer, a personal digital assistant (PDA), a tablet computer (PAD), a portable multimedia player (PMP), a vehicle terminal (such as a vehicle navigation terminal), a wearable device, and the like, and a fixed terminal such as a digital TV, a desktop computer, a smart home device, and the like. The server can be a cloud server or a server cluster, or the like, which is a device with storage and computing functions.

[0053] As shown in Figure 1 The manifold data analysis method can include the following steps.

[0054] S110, obtaining multi-target manifold data.

[0055] In the embodiments of the application, when it is necessary to analyze high-dimensional manifold data, an electronic device can obtain high-dimensional manifold data as multi-target manifold data to further analyze the multi-target manifold data by using an optimization solving algorithm in a topological space.

[0056] In the embodiments of the application, the multi-target manifold data can be multi-dimensional feature data in a topological space.

[0057] Optionally, the multi-target manifold data can be data in the financial field, data in the telecommunications field, data in the Internet of Things field, data in the energy field, and the like. Optionally, the multi-target manifold data can be text data, picture data, audio data, and video data, and the like.

[0058] S120, inputting the multi-target manifold data into a pre-trained manifold data analysis model to obtain an analysis result of the multi-target manifold data, the manifold data analysis model being trained by using sample manifold data and a sample analysis sequence corresponding to the sample manifold data.

[0059] In the embodiment of the present application, after the electronic device obtains the multi-target manifold data, the multi-target manifold data can be input into the pre-trained manifold data analysis model to analyze the multi-target manifold data based on the multi-dimensional features of the multi-target manifold data by using the manifold data analysis model, and obtain the analysis result of the multi-target manifold data.

[0060] In the embodiment of the present application, since the manifold data analysis model can be trained by using the sample manifold data and the sample analysis sequence corresponding to the sample manifold data based on the initial analysis model in the manifold space, the trained manifold data analysis model can analyze the manifold data in the manifold space.

[0061] Specifically, when training the initial analysis model, the electronic device can determine the index of the sample manifold data, sort and group the sample manifold data by using the initial analysis model, and analyze the grouped sample manifold data to obtain the sample analysis sequence.

[0062] Specifically, the manifold data analysis model can combine seven criteria of resolution, accuracy, algorithm efficiency, storage efficiency, energy efficiency, data volume and cost performance. Among them, the resolution can be used to distinguish the degree of good and bad of the manifold data, which can be judged by the good and bad separation degree (KS) of the cross-time sample; the accuracy can be the statistical error of the data and the model, which can be judged by the cumulative square chi-square value on the model variable dimension; the algorithm efficiency can be the model training and prediction efficiency, which can be judged by the length of the required time; the storage efficiency can be the storage space required for model training and prediction, which can be judged by the size of the random access memory (RAM); the energy efficiency can be the power required for model prediction; the data volume can be the label and column of the optimization variable or model dimension; and the cost performance can be judged by the cost of the hardware architecture under the condition of certain accuracy and operation speed.

[0063] In the embodiment of the present application, after obtaining the multi-target manifold data, the multi-target manifold data can be input into the pre-trained manifold data analysis model to obtain the analysis result of the multi-target manifold data. Since the manifold data analysis model is trained by using the sample manifold data and the sample analysis sequence corresponding to the sample manifold data based on the initial analysis model in the manifold space, the manifold data analysis model can well analyze the high-dimensional manifold data, ensure the accuracy of the manifold data analysis, be applicable to massive high-dimensional manifold data analysis, and have good popularization and application value.

[0064] Further, after S120, the method can further include the following steps:

[0065] The analysis result of the multi-target manifold data is filtered to obtain the analysis result after noise filtering.

[0066] Specifically, after the electronic device determines the analysis result of the multi-target manifold data, the deconvolution algorithm can be used to filter noise from the analysis result of the multi-target manifold data to obtain the analysis result after noise filtering, so as to reduce the complexity of the analysis result and further analyze the analysis result with low complexity.

[0067] Therefore, in the embodiments of the present application, the manifold data analysis model and the deconvolution algorithm are combined to analyze the multi-target manifold data, which can improve the accuracy of manifold data analysis while reducing the complexity of the analysis result.

[0068] In the detailed embodiments of the present application, before S110, there is also a step of training the manifold data analysis model.

[0069] Figure 2 A detailed flowchart of another manifold data analysis method provided by the embodiments of the present application is shown.

[0070] S210, obtaining sample manifold data, a sample analysis sequence corresponding to the sample manifold data, and initial model parameters of an initial analysis model.

[0071] In the embodiments of the present application, the electronic device can obtain sample manifold data used to train the initial analysis model and a sample analysis sequence corresponding to the sample manifold data, and randomly determine model parameters of the initial analysis model to obtain the initial model parameters.

[0072] Taking the sample manifold data as movie data as an example, the movie data includes information such as producer information, shooting time, end time, sponsor company, actor information, and release time, and the sample analysis sequence can include information such as box office of the movie data.

[0073] S220, based on the sample manifold data and the sample analysis sequence, iteratively adjusting the initial model parameters of the initial analysis model in the manifold space until the current model parameters at the current adjustment times are stable, obtaining a manifold data analysis model used to determine the analysis result of the multi-target manifold data.

[0074] In the embodiments of the present application, the electronic device can input the sample manifold data into the initial analysis model in the manifold space to obtain an output analysis sequence, adjust the initial model parameters based on the output analysis sequence and the sample analysis sequence, until the error value between the output analysis sequence and the sample analysis sequence is less than a preset error threshold, it is determined that the current model parameters at the current adjustment times are stable, and a manifold data analysis model used to determine the analysis result of the multi-target manifold data is obtained.

[0075] In the embodiment of the present application, the preset error threshold can be an error for judging whether the current model parameter under the current adjustment times is stable, which is determined in advance as required.

[0076] Specifically, the electronic device can sort the sample manifold data according to the characteristics of each dimension of the sample manifold data to obtain sorted sample manifold data; then, the sorted sample manifold data is grouped to obtain a grouping sequence of the sample manifold data, and a grouping sequence of the sample analysis sequence is determined according to the grouping sequence of the sample manifold data; then, the sum of the grouping parameters of the initial analysis model can be calculated according to each grouping sequence of the sample manifold data and the grouping sequence of the sample analysis sequence corresponding to each grouping sequence of the sample manifold data; further, the grouping parameter corresponding to each grouping can be determined according to the sum of the grouping parameters corresponding to all groupings until the initial analysis model is stable, and the model parameter of the manifold data analysis model is obtained.

[0077] Therefore, in the embodiment of the present application, the sample manifold data and the sample analysis sequence can be used to iteratively train the initial analysis model in the manifold space, and in the process of iterative training, the sample manifold data and the sample analysis sequence can be sorted, grouped and the sum of the grouping parameters corresponding to all groupings can be calculated to obtain the grouping parameter corresponding to each grouping until the initial analysis model is stable, and the model parameter of the manifold data analysis model is obtained. Therefore, the manifold data analysis model can be accurately trained, and the training speed of the manifold data analysis model can be improved by calculating the grouping parameter corresponding to each grouping.

[0078] S230, obtaining multi-target manifold data.

[0079] S230 is similar to S110, and will not be repeated here.

[0080] S240, inputting the multi-target manifold data into the pre-trained manifold data analysis model to obtain an analysis result of the multi-target manifold data, wherein the manifold data analysis model is trained by using sample manifold data and sample analysis sequences corresponding to the sample manifold data.

[0081] In the embodiment of the present application, the electronic device can sort and group the multi-target manifold data, determine the analysis result of the multi-target manifold data based on the model parameter of the manifold data analysis model and the multi-target manifold data after grouping, so that the analysis result of the multi-target manifold data can be accurately and efficiently determined based on the trained manifold data analysis model.

[0082] In yet another embodiment of the present application, the sample manifold data can be grouped in different ways, and a grouping parameter and corresponding to each first sample grouping sequence is determined, and then the initial model parameter is determined according to the grouping parameter and corresponding to each first sample grouping sequence.

[0083] In the embodiment of the present application, optionally, S220 can specifically include the following steps:

[0084] S2201, grouping the sample manifold data to obtain a plurality of first sample grouping sequences of the sample manifold data;

[0085] S2202, calculating a prediction analysis sequence of the initial analysis model based on the initial model parameters respectively corresponding to the sample manifold data in each first sample grouping sequence;

[0086] S2203, iteratively adjusting the initial model parameter based on the prediction analysis sequence and the sample analysis sequence until the current model parameter at the current adjustment time is stable, to obtain a manifold data analysis model used to determine the analysis result of the multi-target manifold data.

[0087] In the embodiment of the present application, S2201 can specifically include the following steps:

[0088] S22011, sorting the sample manifold data based on the characteristics of each dimension of the sample manifold data to obtain a first sorting sequence respectively corresponding to the sample manifold data;

[0089] S22012, grouping the sample manifold data corresponding to the first sorting sequence to obtain a plurality of first sample grouping sequences of the sample manifold data.

[0090] Specifically, the electronic device can sort the sample manifold data based on the characteristics of each dimension of the sample manifold data, determine a first sorting sequence respectively corresponding to the sample manifold data, and group the sample manifold data corresponding to the first sorting sequence, so as to divide the sample manifold data into a plurality of groups, to obtain a plurality of first sample grouping sequences of the sample manifold data.

[0091] In the embodiment of the present application, the sample manifold data can be grouped in different ways, and a grouping parameter and corresponding to each first sample grouping sequence is determined, and then the initial model parameter is determined according to the grouping parameter and corresponding to each first sample grouping sequence.

[0092] In some embodiments, S22012 can specifically include the following steps:

[0093] S1, determine a first grouping split point of the sample manifold data based on a data amount of the sample manifold data corresponding to the first ordering sequence;

[0094] S2, group the sample manifold data corresponding to the first ordering sequence according to the first grouping split point, to obtain at least two first sample grouping sequences of the sample manifold data.

[0095] Specifically, the electronic device can determine the data amount of the sample manifold data corresponding to the first ordering sequence, and determine the first grouping split point of the sample manifold data according to the data amount of the sample manifold data corresponding to the first ordering sequence, and then group the sample manifold data corresponding to the first ordering sequence according to the first grouping split point, so as to split the sample manifold data into multiple groups, and obtain at least two first sample grouping sequences of the sample manifold data.

[0096] The data amount can be a total number of data of the sample manifold data.

[0097] The first grouping split point can be determined according to a percentile of the sample manifold data corresponding to the first ordering sequence.

[0098] Taking 1000 data of the sample manifold data as an example, the grouping percentage can be 10%, each group can include 100 sample manifold data, and each 100 sample manifold data determines a grouping split point, 9 first grouping split points are obtained, and the sample manifold data is split into 10 groups based on the first grouping split point, to obtain 10 first sample grouping sequences of the sample manifold data.

[0099] In some other embodiments, S22012 can specifically include the following steps:

[0100] S3, input the sample manifold data corresponding to the first ordering sequence into an initial analysis model to obtain a prediction analysis sequence corresponding to the sample manifold data;

[0101] S4, determine a second grouping split point of the sample manifold data based on a probability value corresponding to each prediction analysis result in the prediction analysis sequence;

[0102] S5, group the sample manifold data corresponding to the first ordering sequence according to the second grouping split point, to obtain at least two first sample grouping sequences of the sample manifold data.

[0103] Specifically, the electronic device can input the sample manifold data corresponding to the first sorting sequence into the initial analysis model to obtain a predicted analysis sequence corresponding to the sample manifold data, each predicted analysis result in the predicted analysis sequence corresponds to a probability value, based on the probability value, calculate the KS value and / or the information value (IV), take the maximum KS value and / or the maximum IV as the grouping cut point to obtain the second grouping cut point of the sample manifold data, and group the sample manifold data corresponding to the first sorting sequence according to the second grouping cut point to obtain at least two first sample grouping sequences of the sample manifold data.

[0104] wherein KS = max{<CSY = 1 | SP> - <CSY = 0 | SP>}.

[0105] wherein SP represents the sorted predicted value, CSY = 1 | SP is the sample manifold data corresponding to the actual value of 1 after sorting the predicted value, CSY = 0 | SP is the sample manifold data corresponding to the actual value of 0 after sorting the predicted value, <CSY = 1 | SP> is the cumulative probability of the actual value of 1, and <CSY = 0 | SP> is the cumulative probability of the actual value of 0.

[0106] wherein IV = (G-B) LN (G / B).

[0107] wherein G is the probability of the actual value of 1 in each group, B is the probability of the actual value of 0 in each group, and model IV is the maximum value of the IV of each group.

[0108] Therefore, in the embodiment of the present application, the cut point of the first sample grouping sequence can be determined according to the percentile of the sample manifold data to obtain the first grouping cut point, or the cut point of the first sample grouping sequence can be determined according to the KS value and / or the IV to obtain the second grouping cut point, and the sample manifold data corresponding to the first sorting sequence is grouped according to the first grouping cut point or the second grouping cut point to obtain at least two first sample grouping sequences of the sample manifold data.

[0109] wherein S2202 specifically can include the following steps:

[0110] S22021, determining the initial grouping parameters of the initial model parameters based on the first sample grouping sequence;

[0111] S22022, for each first sample grouping sequence, calculating the grouping parameter sum of the initial grouping parameters corresponding to the first sample grouping sequence to obtain the predicted analysis sequence of the initial analysis model, and the grouping parameter sum is the sum of all initial model parameters in the initial grouping parameters corresponding to the first sample grouping sequence.

[0112] Specifically, after determining the first sample grouping sequence, the electronic device can determine an initial grouping parameter of the initial model parameter, for each first sample grouping sequence, can calculate a grouping parameter sum of the initial grouping parameter corresponding to the first sample grouping sequence, calculate the grouping parameter sum of all rows of each column of the initial grouping parameter, and then according to the grouping parameter sum and the first sample grouping sequence, can determine a prediction analysis sequence of the initial analysis model, and obtain the sum of all initial model parameters in the initial grouping parameter corresponding to the first sample grouping sequence.

[0113] After S2202, the method can further include the following steps:

[0114] Based on the first sample grouping sequence, determine a plurality of second sample grouping sequences corresponding to the sample analysis sequence.

[0115] Correspondingly, S2203 can specifically include the following steps:

[0116] S22031, for each second sample grouping sequence, calculate the analysis result sum of the sample analysis result corresponding to the second sample grouping sequence;

[0117] S22032, based on the analysis result sum corresponding to each second sample grouping sequence and the prediction analysis sequence, iteratively adjust the initial model parameter until the current model parameter is stable under the current adjustment times, to obtain a manifold data analysis model for determining the analysis result of the multi-target manifold data.

[0118] Specifically, after determining the first sample grouping sequence, the electronic device can determine a plurality of second sample grouping sequences corresponding to the sample analysis sequence, for each second sample grouping sequence, calculate the analysis result sum of the sample analysis result corresponding to the second sample grouping sequence, calculate the sum of all columns of each row of the second sample grouping sequence, and then according to the analysis result sum corresponding to each second sample grouping sequence and the prediction analysis sequence, adjust the initial model parameter to obtain a manifold data analysis model for determining the analysis result of the multi-target manifold data.

[0119] S22031 can specifically include the following steps:

[0120] For each second sample grouping sequence, determine the sample analysis result corresponding to each global index between the initial index and the end index in the second sample grouping sequence;

[0121] The sum of the sample analysis results corresponding to each global index between the initial index and the end index is taken as the analysis result sum of the sample analysis result corresponding to the second sample grouping sequence.

[0122] Specifically, after determining the second sample grouping sequence, the electronic device can determine the sample analysis results corresponding to each global index between the initial index and the end index in the second sample grouping sequence, and then take the sum of the sample analysis results corresponding to each global index between the initial index and the end index as the analysis result sum of the sample analysis results corresponding to the second sample grouping sequence, to calculate the analysis result sum of the sample analysis results corresponding to each row and all columns of the second sample grouping sequence.

[0123] In some embodiments, S22032 can specifically include the following steps:

[0124] Based on the analysis result sum corresponding to the prediction analysis sequence and the second sample grouping sequence, a plurality of analysis difference values are calculated.

[0125] If the analysis difference value with the largest value is greater than a preset fault tolerance value, the initial model parameter is iteratively adjusted until the analysis difference value with the largest value is less than or equal to the preset fault tolerance value, to obtain a manifold data analysis model for determining the analysis result of the multi-target manifold data.

[0126] Specifically, the electronic device can subtract the analysis result sum corresponding to the prediction analysis sequence and the second sample grouping sequence to obtain a plurality of analysis difference values, and if the analysis difference value with the largest value is greater than a preset fault tolerance value, the initial model parameter is iteratively adjusted until the analysis difference value with the largest value is less than or equal to the preset fault tolerance value, to obtain a manifold data analysis model for determining the analysis result of the multi-target manifold data.

[0127] The preset fault tolerance value can be a fault tolerance value for determining whether to stop adjusting the initial model parameter, which can be determined in advance as needed.

[0128] In other embodiments, S22032 can specifically include the following steps:

[0129] Based on the analysis result sum corresponding to the prediction analysis sequence and the second sample grouping sequence, a plurality of analysis difference values are calculated.

[0130] If the mean of all analysis difference values is greater than a preset fault tolerance value, the initial model parameter is iteratively adjusted until the mean of all analysis difference values is less than or equal to the preset fault tolerance value, to obtain a manifold data analysis model for determining the analysis result of the multi-target manifold data.

[0131] Specifically, the electronic device can subtract the analysis result sum corresponding to the prediction analysis sequence and the second sample grouping sequence to obtain a plurality of analysis difference values, and if the mean of all analysis difference values is less than or equal to a preset fault tolerance value, the initial model parameter is iteratively adjusted until the mean of all analysis difference values is less than or equal to the preset fault tolerance value, to obtain a manifold data analysis model for determining the analysis result of the multi-target manifold data.

[0132] The sample manifold data is taken as The sample analysis result is taken as For example.

[0133] Wherein, m is the column number, representing m dimensional variables, and n is the row number.

[0134] Specifically, the sample manifold data X[N] and the sample analysis result Y[N] can be stored in a random access memory (RAM), and then

[0135] Wherein, based on the characteristics of each dimension of the sample manifold data, the sample manifold data is sorted to obtain a first sorting sequence corresponding to the sample manifold data, and thus the first sorting sequence And the first sorting sequence SX is continuously stored in the RAM, and then

[0136] Thus, the sample manifold data is the mapping relationship between X[N] and the first sorting sequence SX,

[0137] Wherein, the sample manifold data corresponding to the first sorting sequence is grouped to obtain at least two first sample grouping sequences of the sample manifold data, and thus the first sample grouping sequence GX[N] = SX[BX:EX], Wherein, 11 is the first grouping of the first dimensional variable, g11 is the g1th grouping of the first dimensional variable, and so on, 1m is the first grouping of the mth dimensional variable, and g m M is the gmth grouping of the mth dimensional variable. BX represents the starting index of each grouping of each dimensional variable, and EX represents the ending index of each grouping of each dimensional variable. Thus, the above process is X[N]->SX[N]->GX[G] = SX[BX:EX] in turn.

[0138] Further, based on the first sample sorting sequence, a plurality of second sample sorting sequences corresponding to the sample analysis sequence are determined, and thus The corresponding relationship between the sample manifold data X[N], Y[N], SX[BX:EX] and SY is:

[0139] Further, according to the second sample sorting sequence SY, the cumulative row sum of SY can be obtained Wherein, The analysis result of each grouping of each dimensional variable can also be represented as Y g = CSY(EX)-CSY(BX).

[0140] Thus, the above process is in turn:

[0141] Y[N]-><SY|SX>-><CSY|SX>-><GY|GX>=<CXY|[BX:EX]|SX>。

[0142] Assuming that each second sample group sequence corresponding to the sample analysis sequence: The initial group parameter corresponding to the first sample group sequence: The initial model parameter Further, for each first sample group sequence, calculate the group parameter sum of the initial group parameter corresponding to the first sample group sequence, that is, for each row of initial group data, calculate the group parameter sum of all columns of initial group parameters, and the group parameter sum of the initial group parameter: Thus, the above process is in turn:<GP|GX>-><P|X>->P[N]=Sum<P|X>。

[0143] Further, based on the prediction analysis sequence and the sample analysis sequence, the initial model parameter is iteratively adjusted until the current model parameter under the current adjustment times is stable, and a manifold data analysis model for determining the analysis result of the multi-target manifold data is obtained.

[0144] Wherein, the sample manifold data, the sample analysis sequence and the first sorting sequence can be determined in turn X->SX->GX, wherein GX[G]=SX[BX:EX], and then Y-><Y|X>-><SY|SX>-><CSY|SX> is determined in turn, wherein <GY|GX>=<CSY(BX:EX)|SX>.

[0145] Wherein, <GP|GX>=<GY|GX>→P→<SP|SX> is determined in turn, and the group parameter sum P=∑P m Thus, the above process is in turn:

[0146] P[N]-><SP|SX>-><CSP|SX>-><GP|GX>=<CSP[BX:EX]|SX>。

[0147] Further, based on the prediction analysis sequence and the sample analysis sequence, the initial model parameter is iteratively adjusted until the current model parameter under the current adjustment times is stable, and a manifold data analysis model for determining the analysis result of the multi-target manifold data is obtained. The mean value of the analysis difference value is:

[0148] Optionally, if the analysis difference with the largest value is greater than the preset fault tolerance value, or the average of the analysis differences is greater than the preset fault tolerance value, the initial model parameters are iteratively adjusted until the analysis difference with the largest value is less than or equal to the preset fault tolerance value, or the average of the analysis differences is less than or equal to the preset fault tolerance value, and the manifold data analysis model for determining the analysis result of the multi-target manifold data can be obtained.

[0149] Therefore, in the embodiment of the present application, the sample manifold data can be grouped in different ways, and the grouping parameter sum corresponding to each first sample grouping sequence is determined, and then the initial model parameters are accurately determined according to the grouping parameter sum corresponding to each first sample grouping sequence.

[0150] Further, the sample manifold data can be the sample analysis sequence The above process is further analyzed by taking x1, …, x5 belonging to the first group G1, x6, …, x 10 belonging to the second group G2, BX=[1, 6] is the initial index of G1 and G2, EX=[5, 10] is the end index of G1 and G2, GY=[gy1, gy2] is set, wherein gy1=y1+…+y5, gy2=y6+…+y 10 , and the grouping parameter sum P0 of the initial model parameters is set as P0=[p 01 , p 02 ].

[0151] Further, the analysis difference with the largest value is calculated If the analysis difference with the largest value is greater than the preset fault tolerance value, p 11 =gy1 / p 01 , p 12 =gy2 / p 02 , that is, P new =P1=[p 11， , p 12 ], P old =P0, according to P new =P1=[p 11 , p 12 ], P old =P0, the analysis difference with the largest value is continued to be calculated If the analysis difference with the largest value is greater than the preset fault tolerance value, p 21 =p 11 , p 22 =gy2 / p 12 , that is, P new =P2=[p 21 p 22 ], P old =P1, according to P new =P2=[p 21 p 22 ], P old =P1, continue calculating the analytical difference with the largest value. If the largest analytical difference is less than or equal to the preset tolerance value, then P k =[p k1 p k2 The parameters are the final model parameters. Therefore, the above process is as follows:<GP|GX> new =<GY|GX> / <GP|GX> old max|<GP|GX> -<GY|GX> | <Err。

[0152] This invention also provides a manifold data analysis apparatus for implementing the above-described manifold data analysis method, which will be described below in conjunction with... Figure 3 The following explanation is provided. In this embodiment of the invention, the manifold data analysis device can be an electronic device. This electronic device may include mobile terminals, tablet computers, in-vehicle terminals, wearable electronic devices, virtual reality (VR) all-in-one machines, smart home devices, and other devices with communication functions.

[0153] Figure 3 A schematic diagram of the structure of the manifold data analysis device provided in an embodiment of the present invention is shown.

[0154] like Figure 3 As shown, the manifold data analysis device 300 may include: a multi-target manifold data acquisition module 310 and an analysis result determination module 320.

[0155] Multi-objective manifold data acquisition module 310 is used to acquire multi-objective manifold data;

[0156] The analysis result determination module 320 is used to input multi-objective manifold data into a pre-trained manifold data analysis model to obtain the analysis results of the multi-objective manifold data. The manifold data analysis model is trained on an initial analysis model in the manifold space using sample manifold data and the sample analysis sequence corresponding to the sample manifold data.

[0157] In the embodiment of the present application, after the multi-target manifold data is acquired, the multi-target manifold data can be input into the pre-trained manifold data analysis model to obtain the analysis result of the multi-target manifold data. Since the manifold data analysis model is trained by using the sample manifold data and the sample analysis sequence corresponding to the sample manifold data, the initial analysis model in the manifold space, the manifold data analysis model can be used to well analyze the high-dimensional manifold data, ensure the accuracy of the manifold data analysis, be applicable to massive high-dimensional manifold data for analysis, and have good popularization and application value.

[0158] Optionally, the apparatus further includes a sample acquisition module and a training module.

[0159] The sample acquisition module can be used to acquire the sample manifold data, the sample analysis sequence corresponding to the sample manifold data, and the initial model parameter of the initial analysis model.

[0160] The training module can be used to iteratively adjust the initial model parameter of the initial analysis model in the manifold space based on the sample manifold data and the sample analysis sequence until the current model parameter at the current adjustment time is stable, and obtain the manifold data analysis model used to determine the analysis result of the multi-target manifold data.

[0161] Optionally, the training module can include a grouping unit, a predicted analysis sequence calculation unit, and a training unit.

[0162] The grouping unit can be used to group the sample manifold data to obtain a plurality of first sample grouping sequences of the sample manifold data.

[0163] The predicted analysis sequence calculation unit can be used to calculate the predicted analysis sequence of the initial analysis model based on the initial model parameter respectively corresponding to the sample manifold data in each first sample grouping sequence.

[0164] The training unit can be used to iteratively adjust the initial model parameter based on the predicted analysis sequence and the sample analysis sequence until the current model parameter at the current adjustment time is stable, and obtain the manifold data analysis model used to determine the analysis result of the multi-target manifold data.

[0165] Optionally, the grouping unit can be further used to sort the sample manifold data based on the feature of each dimension of the sample manifold data to obtain a first sorting sequence respectively corresponding to the sample manifold data.

[0166] The sample manifold data corresponding to the first sorting sequence is grouped to obtain a plurality of first sample grouping sequences of the sample manifold data.

[0167] Optionally, the grouping unit can be further configured to determine the first grouping split point of the sample manifold data based on a data amount of the sample manifold data corresponding to the first ordering sequence.

[0168] According to the first grouping split point, the sample manifold data corresponding to the first ordering sequence is grouped to obtain at least two first sample grouping sequences of the sample manifold data.

[0169] Optionally, the grouping unit can be further configured to input the sample manifold data corresponding to the first ordering sequence into the initial analysis model to obtain a prediction analysis sequence corresponding to the sample manifold data.

[0170] Based on a probability value corresponding to each prediction analysis result in the prediction analysis sequence, the second grouping split point of the sample manifold data is determined.

[0171] According to the second grouping split point, the sample manifold data corresponding to the first ordering sequence is grouped to obtain at least two first sample grouping sequences of the sample manifold data.

[0172] Optionally, the prediction analysis sequence calculation unit can be further configured to determine an initial grouping parameter of the initial model parameter based on the first sample grouping sequence.

[0173] For each first sample grouping sequence, a grouping parameter sum of the initial grouping parameter corresponding to the first sample grouping sequence is calculated to obtain the prediction analysis sequence of the initial analysis model, and the grouping parameter sum is a sum of all initial model parameters in the initial grouping parameter corresponding to the first sample grouping sequence.

[0174] Optionally, the apparatus further comprises a second sample grouping sequence determination module.

[0175] The second sample grouping sequence determination module can be configured to determine a plurality of second sample grouping sequences corresponding to the sample analysis sequence based on the first sample grouping sequence.

[0176] Optionally, the training unit can be further configured to calculate, for each second sample grouping sequence, an analysis result sum of the sample analysis result corresponding to the second sample grouping sequence.

[0177] Based on the analysis result sum corresponding to each second sample grouping sequence and the prediction analysis sequence, the initial model parameter is iteratively adjusted until a current model parameter at a current adjustment time is stable, to obtain a manifold data analysis model for determining an analysis result of multi-target manifold data.

[0178] Optionally, the training unit can be further configured to determine, for each second sample grouping sequence, a sample analysis result corresponding to each global index between an initial index and an end index in the second sample grouping sequence.

[0179] The sum of the sample analysis results corresponding to each global index between the initial index and the end index is taken as the analysis result sum of the sample analysis results corresponding to the second sample grouping sequence.

[0180] Optionally, the training unit can be further configured to calculate a plurality of analysis differences based on the analysis result sum corresponding to the second sample grouping sequence and the prediction analysis sequence.

[0181] If the analysis difference with the largest value is greater than the preset fault tolerance value, the initial model parameter is iteratively adjusted until the analysis difference with the largest value is less than or equal to the preset fault tolerance value, and a manifold data analysis model for determining the analysis result of the multi-target manifold data is obtained.

[0182] Optionally, the training unit can be further configured to calculate a plurality of analysis differences based on the analysis result sum corresponding to the second sample grouping sequence and the prediction analysis sequence.

[0183] If the mean of all analysis differences is greater than the preset fault tolerance value, the initial model parameter is iteratively adjusted until the mean of all analysis differences is less than or equal to the preset fault tolerance value, and a manifold data analysis model for determining the analysis result of the multi-target manifold data is obtained.

[0184] Optionally, the apparatus can further include a noise filtering module.

[0185] The noise filtering module can be configured to perform noise filtering on the analysis result of the multi-target manifold data to obtain a noise-filtered analysis result.

[0186] It should be noted that, Figure 3 The video display apparatus 1300 shown can perform each step in the method embodiment shown, and achieve each process and effect in the method embodiment shown, and thus will not be described here. Figures 1 to 2 The video display apparatus 1300 shown can perform each step in the method embodiment shown, and achieve each process and effect in the method embodiment shown, and thus will not be described here. Figures 1 to 2 The video display apparatus 1300 shown can perform each step in the method embodiment shown, and achieve each process and effect in the method embodiment shown, and thus will not be described here.

[0187] Figure 4 A structure diagram of a manifold data analysis device provided by an embodiment of the application is shown.

[0188] As Figure 4 shown, the data acquisition device can include a processor 401 and a memory 402 storing computer program instructions.

[0189] Specifically, the processor 401 can include a central processing unit (CPU), or a specific integrated circuit (ASIC), or can be configured as one or more integrated circuits that implement one or more embodiments of the application.

[0190] The memory 402 can include mass storage for information or instructions. For example, and without limitation, the memory 402 can include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a Universal Serial Bus (USB) drive or a combination of two or more of these. Storage memory 402, where appropriate, can be removable or non-removable (or fixed) media. Storage memory 402, where appropriate, can be internal or external to the integrated gateway device. In certain embodiments, the storage memory 402 is non-volatile solid-state memory. In certain embodiments, the storage memory 402 includes read-only memory (ROM). Where appropriate, this ROM can be mask- programmed ROM, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), electrically alterable ROM (EAROM), or flash memory, or a combination of two or more of these. The storage memory 402 can include one or more components or elements selected from a group consisting of a solid-state memory, a magnetic storage media, an optical storage media, or a combination of two or more of these. Where appropriate, this storage memory 402 can be removable or non-removable (or fixed) media. Where appropriate, this storage memory 402 can be internal or external to the integrated gateway device. In certain embodiments, the storage memory 402 is non-volatile solid-state memory. In certain embodiments, the storage memory 402 includes read-only memory (ROM). Where appropriate, this ROM can be mask-programmed ROM, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), electrically alterable ROM (EAROM), or flash memory, or a combination of two or more of these.

[0191] The processor 401 performs the steps of the manifold data analysis method provided by embodiments of the present application by reading and executing computer program instructions stored in the memory 402.

[0192] In one example, the manifold data analysis device can further include a transceiver 403 and a bus 404. As shown, the processor 401, the memory 402, and the transceiver 403 are connected by the bus 404 and complete communication among each other. Figure 4

[0193] ​Bus 404 includes hardware, software, or both. For example, and not limitingly, a bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industrial Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a MicroChannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, bus 404 may include one or more buses. Although specific buses are described and illustrated in the embodiments of this application, this application considers any suitable bus or interconnection.

[0194] In this embodiment of the invention, optionally, the processor may be a field-programmable gate array.

[0195] Among them, field-programmable gate arrays (FPGAs) can be used to process multi-objective manifold data with n rows and m columns. FPGAs can process multi-objective manifold data with n rows and m columns through pipelines.

[0196] Figure 5 A schematic diagram of the structure of a field-programmable gate array provided in an embodiment of the present invention is shown.

[0197] As shown in Figure 5, the field-programmable gate array includes n pipes, CSY is the analysis result of the sample analysis result corresponding to each second sample grouping sequence, and Y is the result of the analysis of the sample analysis result. g For each second sample grouping sequence corresponding to the sample analysis sequence, BX is the initial index of GX[N], and EX is the ending index of GX[N].

[0198] Specifically, for each column m, that is, the mth dimension variable, starting from the first row of the first group of data, that is, starting from n=1, g=1, the starting position BXgm of the gth group of the mth dimension variable is obtained, and the ending position EXgm of the gth group of the mth dimension variable is obtained, and the analysis result and CSY of the sample analysis result corresponding to each second sample group sequence are determined through EXgm BXgm,m , the analysis result and CSY of the sample analysis result corresponding to each second sample group sequence are determined through EXgm EXgm,m , then the y value ygm of the gth group of the mth variable is ygm=CSY EXgm,m —CSY BXgm,m Based on the analysis result and CSY corresponding to each second sample group sequence and the predicted analysis sequence, n=n+1 is iteratively calculated, and if BXgm=EXgm, g=g+1, that is, the initial model parameters are iteratively adjusted until the current model parameters under the current adjustment times are stable, and a manifold data analysis model for determining the analysis result of the multi-target manifold data is obtained.

[0199] The following is an embodiment of a computer-readable storage medium provided by the embodiment of the application, which belongs to the same inventive concept as the manifold data analysis method of each embodiment described above. Details not described in the embodiment of the computer-readable storage medium can be referred to the embodiments of the manifold data analysis method described above.

[0200] The embodiment provides a storage medium of computer executable instructions, which is used to execute a manifold data analysis method when executed by a computer processor, and the method comprises:

[0201] Obtaining multi-target manifold data;

[0202] Inputting the multi-target manifold data into a pre-trained manifold data analysis model to obtain an analysis result of the multi-target manifold data, wherein the manifold data analysis model is trained by using sample manifold data and a sample analysis sequence corresponding to the sample manifold data.

[0203] Of course, the storage medium containing computer executable instructions provided by the embodiment of the application is not limited to the method operations described above, and can also perform related operations in the manifold data analysis method provided by any embodiment of the application.

[0204] Through the above description of the embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software and necessary universal hardware, and of course can also be implemented by hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, and the computer software product can be stored in a computer readable storage medium, such as a floppy disk, a read-only memory (ROM), a random access memory (RAM), a FLASH, a hard disk, or an optical disc, etc., including a plurality of instructions for causing a computer cloud platform (which can be a personal computer, a server, or a network cloud platform, etc.) to execute the manifold data analysis method provided by each embodiment of the present application.

[0205] Note that the above are only the preferred embodiments of the present application and the technical principles applied. Those skilled in the art will understand that the present application is not limited to the specific embodiments described herein, and that various obvious changes, re-adjustments and substitutions can be made by those skilled in the art without departing from the scope of the present application. Therefore, although the present application has been described in more detail through the above embodiments, the present application is not limited to the above embodiments, and can include more other equivalent embodiments without departing from the concept of the present application, and the scope of the present application is determined by the scope of the appended claims. < / y>

Claims

1. A multi-objective manifold data analysis method, characterized in that, The method comprises the following steps: acquiring multi-target manifold data; inputting the multi-target manifold data into a pre-trained manifold data analysis model to obtain an analysis result of the multi-target manifold data, wherein the manifold data analysis model is trained by using sample manifold data and a sample analysis sequence corresponding to the sample manifold data, and an initial analysis model in a manifold space; combining the manifold data analysis model and a deconvolution algorithm to analyze the multi-target manifold data; after the electronic device determines the analysis result of the multi-target manifold data, filtering noise from the analysis result of the multi-target manifold data by using the deconvolution algorithm; the manifold data analysis model is trained by the following steps: acquiring sample manifold data, a sample analysis sequence corresponding to the sample manifold data, and initial model parameters of an initial analysis model; iteratively adjusting the initial model parameters of the initial analysis model in the manifold space based on the sample manifold data and the sample analysis sequence until the current model parameters at the current adjustment time are stable, to obtain the manifold data analysis model used to determine the analysis result of the multi-target manifold data; the step of iteratively adjusting the initial model parameters of the initial analysis model in the manifold space based on the sample manifold data and the sample analysis sequence until the current model parameters at the current adjustment time are stable, to obtain the manifold data analysis model used to determine the analysis result of the multi-target manifold data, comprises the following steps: grouping the sample manifold data to obtain a plurality of first sample grouping sequences of the sample manifold data; calculating a predicted analysis sequence of the initial analysis model based on the initial model parameters corresponding to the sample manifold data in each first sample grouping sequence; iteratively adjusting the initial model parameters based on the predicted analysis sequence and the sample analysis sequence until the current model parameters at the current adjustment time are stable, to obtain the manifold data analysis model used to determine the analysis result of the multi-target manifold data; the step of grouping the sample manifold data to obtain a plurality of first sample grouping sequences of the sample manifold data comprises the following steps: sorting the sample manifold data based on the characteristics of each dimension of the sample manifold data to obtain a first sorting sequence corresponding to the sample manifold data; grouping the sample manifold data corresponding to the first sorting sequence to obtain a plurality of first sample grouping sequences of the sample manifold data.

2. The method of claim 1, wherein, the step of grouping the sample manifold data corresponding to the first sorting sequence to obtain at least two first sample grouping sequences of the sample manifold data comprises the following steps: determining a first grouping division point of the sample manifold data based on the data amount of the sample manifold data corresponding to the first sorting sequence; grouping the sample manifold data corresponding to the first sorting sequence according to the first grouping division point to obtain at least two first sample grouping sequences of the sample manifold data.

3. The method of claim 1, wherein, The grouping of the sample manifold data corresponding to the first sorting sequence obtains at least two first sample grouping sequences of the sample manifold data, including: The sample manifold data corresponding to the first sorting sequence is input into the initial analysis model to obtain a prediction analysis sequence corresponding to the sample manifold data; Based on a probability value corresponding to each prediction analysis result in the prediction analysis sequence, a second grouping split point of the sample manifold data is determined; According to the second grouping split point, the sample manifold data corresponding to the first sorting sequence is grouped to obtain at least two first sample grouping sequences of the sample manifold data.

4. The method of claim 1, wherein, The calculation of the prediction analysis sequence of the initial analysis model based on the initial model parameters respectively corresponding to the sample manifold data in each first sample grouping sequence, includes: Based on the first sample grouping sequence, initial grouping parameters of the initial model parameters are determined; For each first sample grouping sequence, a grouping parameter sum of the initial grouping parameters corresponding to the first sample grouping sequence is calculated to obtain the prediction analysis sequence of the initial analysis model, and the grouping parameter sum is a sum of all initial model parameters in the initial grouping parameters corresponding to the first sample grouping sequence.

5. The method of claim 1, wherein, After the grouping of the sample manifold data to obtain a plurality of first sample grouping sequences of the sample manifold data, the method further includes: Based on the first sample grouping sequence, a plurality of second sample grouping sequences corresponding to the sample analysis sequence are determined; Wherein, the iterative adjustment of the initial model parameters based on the prediction analysis sequence and the sample analysis sequence until the current model parameters under the current adjustment times are stable to obtain a manifold data analysis model for determining the analysis result of the multi-target manifold data, includes: For each second sample grouping sequence, an analysis result sum of the sample analysis result corresponding to the second sample grouping sequence is calculated; Based on the analysis result sum corresponding to each second sample grouping sequence and the prediction analysis sequence, the initial model parameters are iteratively adjusted until the current model parameters under the current adjustment times are stable to obtain a manifold data analysis model for determining the analysis result of the multi-target manifold data.

6. The method of claim 5, wherein, The calculation of the analysis result sum of the sample analysis result corresponding to each second sample grouping sequence, includes: For each second sample grouping sequence, a sample analysis result respectively corresponding to each global index between an initial index and an end index in the second sample grouping sequence is determined; The sum of the sample analysis results respectively corresponding to each global index between the initial index and the end index is taken as the analysis result sum of the sample analysis result corresponding to the second sample grouping sequence.

7. The method of claim 5, wherein, The iterative adjustment of the initial model parameters based on the analysis result sum corresponding to each second sample grouping sequence and the prediction analysis sequence until the current model parameters under the current adjustment times are stable to obtain a manifold data analysis model for determining the analysis result of the multi-target manifold data, includes: calculating a plurality of analysis difference values based on the analysis result sum corresponding to the second sample grouping sequence and the predicted analysis sequence; if the analysis difference value with the largest value is greater than a preset fault tolerance value, iteratively adjusting the initial model parameter until the analysis difference value with the largest value is less than or equal to the preset fault tolerance value, to obtain a manifold data analysis model for determining the analysis result of the multi-target manifold data.

8. The method of claim 5, wherein, The method for iteratively adjusting the initial model parameter based on the analysis result sum corresponding to each second sample grouping sequence and the predicted analysis sequence until the current model parameter at the current adjustment times is stable, to obtain a manifold data analysis model for determining the analysis result of the multi-target manifold data, comprises: calculating a plurality of analysis difference values based on the analysis result sum corresponding to the second sample grouping sequence and the predicted analysis sequence; if the mean of all the analysis difference values is greater than a preset fault tolerance value, iteratively adjusting the initial model parameter until the mean of all the analysis difference values is less than or equal to the preset fault tolerance value, to obtain a manifold data analysis model for determining the analysis result of the multi-target manifold data.

9. The method of claim 1, wherein, After obtaining the analysis result of the multi-target manifold data, the method further comprises: performing noise filtering on the analysis result of the multi-target manifold data to obtain an analysis result after noise filtering.

10. A multi-objective manifold data analysis device, characterized by, comprises: a processor; a memory for storing executable instructions; wherein the processor is configured to read the executable instructions from the memory and execute the executable instructions to implement the manifold data analysis method of any one of claims 1-9.

11. The apparatus of claim 10, wherein, The processor is a field programmable logic gate array.

12. A computer readable storage medium having stored thereon a computer program, characterized in that, The storage medium stores a computer program, and when the computer program is executed by the processor, the processor implements the manifold data analysis method of any one of claims 1-9. The storage medium stores a computer program, and when the computer program is executed by the processor, the processor implements the manifold data analysis method of any one of claims 1-9.

Images