Disk health prediction method and construction method of disk health prediction model
By building a disk health prediction model and using neural networks and feature subsets to train disk data, the problem of difficult to guarantee the accuracy and efficiency of disk health prediction in the existing technology is solved, and more efficient and accurate prediction of disk health status is achieved.
Patent Information
- Application Number
- CN202411785592.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-06
- Publication Date
- 2025-03-28
AI Technical Summary
In the prior art, the accuracy and efficiency of disk health prediction are difficult to guarantee, and it mainly relies on manual intervention for analysis.
A disk health prediction method is adopted to build a disk health prediction model, and use preset neural network models and feature subsets to train and predict disk data. The feature subset is obtained through preset search algorithms and initialized populations, and optimized using the fitness function, and finally generates a model for predicting disk health status.
It improves the accuracy and efficiency of disk health prediction, reduces the dependence of manual intervention, reduces the false positive rate, and enhances the stability of the prediction results.
Smart Images
Figure CN119718823A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of computer storage technology, and in particular to a disk health prediction method and a method for constructing a disk health prediction model. Background Art
[0002] In today's digital age, the importance of data is increasingly recognized and valued. With the rapid development of the Internet of Things, cloud computing, and cloud storage, the total amount of data generated by human society has exploded. These data cover various industries and fields, including corporate business data, personal digital assets, and scientific research experimental data. However, data storage and management face a series of challenges.
[0003] Data centers play an important role as key infrastructure for storing and processing large amounts of data. They carry massive amounts of data and provide efficient access and reliable storage services. However, data centers also face some key issues, one of which is the risk of data loss due to hard drive failure. Therefore, it is very important to monitor the health status of hard drives.
[0004] In the prior art, the technical solution for hard disk health prediction is to monitor some disk data, issue an alarm if there is abnormal data, and conduct secondary analysis after manual intervention by operation and maintenance and R&D personnel to predict the health status of the disk. However, this solution is extremely dependent on manual labor, and the accuracy and efficiency of the prediction are difficult to guarantee. Summary of the invention
[0005] The technical solution of the present invention aims to provide a disk health prediction method and a method for constructing a disk health prediction model, so as to solve the problem that the accuracy and efficiency of disk health prediction are difficult to ensure in the prior art.
[0006] To achieve the above object, the present invention is achieved by:
[0007] In a first aspect, an embodiment of the present invention provides a disk health prediction method, comprising:
[0008] According to the disk health prediction model and the disk data to be predicted, a prediction result for indicating the health status of the disk is obtained; wherein, the disk health prediction model is obtained by training a preset neural network model according to a first training data set; the first training data set is obtained by selecting data from a first disk sample data set according to a first feature subset; each element in the first feature subset is a first value or a second value, the first value is used to indicate that data at a corresponding element position in the first disk sample data set is selected, and the second value is used to indicate that data at a corresponding element position in the first disk sample data set is not selected; the first feature subset is obtained by using a preset search algorithm and an initialized population, and each individual in the initialized population is obtained by probabilistically generating the first value and the second value according to a preset distribution function.
[0009] Optionally, the disk health prediction method further comprises:
[0010] Obtaining a second feature subset;
[0011] A second training data set is selected from a second disk sample data set according to the second feature subset; wherein the second disk sample data set includes the first disk sample data set and the disk data to be predicted; each element in the second feature subset is a first value or a second value, the first value is used to indicate that data at a corresponding element position in the second disk sample data set is selected, and the second value is used to indicate that data at a corresponding element position in the second disk sample data set is not selected;
[0012] The disk health prediction model is updated according to the second training data set to obtain an updated disk health prediction model.
[0013] Optionally, the disk health prediction method further comprises:
[0014] According to the following steps, a target feature subset is obtained; wherein the target feature subset is the first feature subset or the second feature subset;
[0015] Probabilistically generate the first value and the second value according to a preset distribution function to obtain each individual in the initialized population;
[0016] Iteratively updating each individual of the initialized population using a preset search algorithm until an iteration termination condition is met to obtain an updated population;
[0017] According to the fitness function, obtaining the best individual in the updated population;
[0018] Perform probabilistic binary mapping on the optimal individual to obtain the target feature subset.
[0019] Optionally, in the disk health prediction method, the fitness function is determined according to a first objective function and a second objective function, the first objective function is a function related to a classification error rate of a target training data set, and the second objective function is a function related to the number of the target training data set;
[0020] The target training data set is the first training data set or the second training data set.
[0021] Optionally, the disk health prediction method, wherein the optimal individual is subjected to probabilistic binary mapping to obtain the target feature subset, comprises:
[0022] The optimal individual is subjected to probabilistic binary mapping through a S-type function to obtain the target feature subset.
[0023] In a second aspect, an embodiment of the present invention further provides a method for constructing a disk health prediction model, comprising:
[0024] Probabilistically generate the first value and the second value according to a preset distribution function to obtain each individual in the initialized population;
[0025] Using a preset search algorithm and the initialization population, obtaining a first feature subset;
[0026] Selecting data from a first disk sample data set according to the first feature subset to obtain a first training data set; wherein each element in the first feature subset is the first value or the second value, the first value is used to indicate selecting data at a corresponding element position in the first disk sample data set, and the second value is used to indicate not selecting data at a corresponding element position in the first disk sample data set;
[0027] The preset neural network model is trained according to the first training data set to obtain a disk health prediction model.
[0028] Optionally, the method for constructing the disk health prediction model, wherein the using a preset search algorithm and the initialization population to obtain a first feature subset includes:
[0029] Iteratively updating each individual of the initialized population using a preset search algorithm until an iteration termination condition is met to obtain an updated population;
[0030] According to the fitness function, obtaining the best individual in the updated population;
[0031] Perform probabilistic binary mapping on the optimal individual to obtain the first feature subset.
[0032] Optionally, in the method for constructing the disk health prediction model, the fitness function is determined according to a first objective function and a second objective function, the first objective function is a function related to the classification error rate of a target training data set, and the second objective function is a function related to the number of the target training data set;
[0033] The target training data set is the first training data set or the second training data set.
[0034] Optionally, the method for constructing the disk health prediction model, wherein the optimal individual is subjected to probabilistic binary mapping to obtain the first feature subset, comprises:
[0035] The first feature subset is obtained by performing probability binary mapping on the optimal individual through a S-type function.
[0036] In a third aspect, an embodiment of the present invention further provides a disk health prediction device, comprising:
[0037] The first acquisition module is used to obtain a prediction result indicating the health status of the disk based on a disk health prediction model and disk data to be predicted; wherein the disk health prediction model is obtained by training a preset neural network model based on a first training data set; the first training data set is obtained by selecting data from a first disk sample data set according to a first feature subset; each element in the first feature subset is a first value or a second value, the first value is used to indicate the selection of data at a corresponding element position in the first disk sample data set, and the second value is used to indicate that data at a corresponding element position in the first disk sample data set is not selected; the first feature subset is obtained by using a preset search algorithm and an initialized population, and each individual in the initialized population is obtained by probabilistically generating the first value and the second value according to a preset distribution function.
[0038] In a fourth aspect, an embodiment of the present invention further provides a device for constructing a disk health prediction model, comprising:
[0039] A first obtaining module is used to generate the first value and the second value by probability according to a preset distribution function to obtain each individual in the initialized population;
[0040] A second acquisition module, used to acquire a first feature subset using a preset search algorithm and the initialization population;
[0041] A first selection module is used to select data from a first disk sample data set according to the first feature subset to obtain a first training data set; wherein each element in the first feature subset is the first value or the second value, the first value is used to indicate that data at a corresponding element position in the first disk sample data set is selected, and the second value is used to indicate that data at a corresponding element position in the first disk sample data set is not selected;
[0042] The second acquisition module is used to train the preset neural network model according to the first training data set to obtain a disk health prediction model.
[0043] In a fifth aspect, an embodiment of the present invention further provides a disk health prediction device, comprising: a processor, a memory, and a program stored in the memory and executable on the processor, wherein when the program is executed by the processor, the disk health prediction method as described in the first aspect is implemented.
[0044] In the sixth aspect, an embodiment of the present invention also provides a device for constructing a disk health prediction model, comprising: a processor, a memory, and a program stored in the memory and executable on the processor, wherein when the program is executed by the processor, the method for constructing a disk health prediction model as described in the second aspect is implemented.
[0045] In the seventh aspect, an embodiment of the present invention further provides a readable storage medium, on which a program is stored, and when the program is executed by a processor, the method for predicting disk health as described in the first aspect is implemented, or the method for constructing a disk health prediction model as described in the second aspect is implemented.
[0046] In an eighth aspect, an embodiment of the present invention further provides a computer program product, comprising computer instructions, which, when executed by a processor, implement the disk health prediction method as described in the first aspect, or implement the method for constructing a disk health prediction model as described in the second aspect.
[0047] The beneficial effects of the above technical solution of the present invention are as follows:
[0048] The disk health prediction method described in the embodiment of the present invention is used to obtain a prediction result for indicating the health status of the disk according to the disk health prediction model and the disk data to be predicted; wherein the disk health prediction model is obtained by training a preset neural network model according to a first training data set; the first training data set is obtained by selecting data from a first disk sample data set according to a first feature subset; each element in the first feature subset is a first value or a second value, the first value is used to indicate the selection of data at a corresponding element position in the first disk sample data set, and the second value is used to indicate that data at a corresponding element position in the first disk sample data set is not selected; the first feature subset is obtained by using a preset search algorithm and an initialization population, and each individual in the initialization population is obtained by probabilistically generating the first value and the second value according to a preset distribution function. In this way, the first training data set is selected according to the first feature subset, avoiding the loss of key features or hidden features that may be caused by subjective selection of features, and reducing redundant features. The disk health prediction model obtained by training based on the first training data set can effectively improve the accuracy and efficiency of disk health prediction. BRIEF DESCRIPTION OF THE DRAWINGS
[0049] Figure 1 A schematic diagram of a process of predicting disk health according to an embodiment of the present invention;
[0050] Figure 2 A flowchart of an implementation method of a disk health prediction method according to an embodiment of the present invention;
[0051] Figure 3 A schematic diagram of a process for constructing a disk health prediction model according to an embodiment of the present invention;
[0052] Figure 4 This is a schematic diagram of the architecture of the disk health prediction model described in an embodiment of the present invention;
[0053] Figure 5 It is a structural schematic diagram of the disk health prediction device according to an embodiment of the present invention;
[0054] Figure 6 It is a structural schematic diagram of a device for constructing a disk health prediction model according to an embodiment of the present invention;
[0055] Figure 7 A hardware block diagram of the disk health prediction device according to an embodiment of the present invention;
[0056] Figure 8 This is a hardware block diagram of a device for constructing a disk health prediction model according to an embodiment of the present invention. DETAILED DESCRIPTION
[0057] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.
[0058] The embodiments of the present invention, as Figure 1 As shown, a disk health prediction method is provided, comprising:
[0059] S101, obtaining a prediction result indicating the health status of the disk according to a disk health prediction model and disk data to be predicted; wherein the disk health prediction model is obtained by training a preset neural network model according to a first training data set; the first training data set is obtained by selecting data from a first disk sample data set according to a first feature subset; each element in the first feature subset is a first value or a second value, the first value is used to indicate the selection of data at a corresponding element position in the first disk sample data set, and the second value is used to indicate that data at a corresponding element position in the first disk sample data set is not selected; the first feature subset is obtained by using a preset search algorithm and an initialized population, and each individual in the initialized population is obtained by probabilistically generating the first value and the second value according to a preset distribution function.
[0060] In an embodiment of the present invention, the disk data to be predicted may be data generated in real time by the disk during operation of a distributed storage system. Based on the disk data to be predicted acquired in real time, the health status of the disk is monitored to obtain a prediction result indicating the health status of the disk, wherein the prediction result includes healthy or unhealthy.
[0061] The first training data set is obtained by selecting data from the first disk sample data set according to the first feature subset. Through feature selection, the first training data set is comprehensive and irreplaceable, retains the implicit relationship between features, and also avoids redundant relationships between features, thereby achieving data filtering and purification effects. Based on the first training data set, the false alarm rate of the prediction result is reduced and the stability is improved.
[0062] The first disk sample data set includes at least one of the following: SMART (Self-Monitoring Analysis and Reporting Technology) attributes; disk IO (input / output) attributes; system resource attributes; cluster performance indicators; and disk models.
[0063] The first value and the second value may, optionally, be of binary type, such as the first value is 1 and the second value is 0.
[0064] Therefore, the first feature subset is a binary array, for example, [01000111…0101].
[0065] In one embodiment, optionally, the method further comprises:
[0066] Obtaining a second feature subset;
[0067] A second training data set is selected from a second disk sample data set according to the second feature subset; wherein the second disk sample data set includes the first disk sample data set and the disk data to be predicted; each element in the second feature subset is a first value or a second value, the first value is used to indicate that data at a corresponding element position in the second disk sample data set is selected, and the second value is used to indicate that data at a corresponding element position in the second disk sample data set is not selected;
[0068] The disk health prediction model is updated according to the second training data set to obtain an updated disk health prediction model.
[0069] In an embodiment of the present invention, the second training data set is taken from the first disk sample data and the disk data to be predicted acquired in real time according to the second feature subset, and the disk health prediction model is retrained based on the second training data set to update the disk health prediction model.
[0070] Therefore, the disk health prediction method described in the embodiment of the present invention can adapt to the constantly updated actual operating environment and generate new training data sets, thereby updating the disk health prediction model.
[0071] In one embodiment, optionally, the method further comprises:
[0072] According to the following steps, a target feature subset is obtained; wherein the target feature subset is the first feature subset or the second feature subset;
[0073] Probabilistically generate the first value and the second value according to a preset distribution function to obtain each individual in the initialized population;
[0074] Iteratively updating each individual of the initialized population using a preset search algorithm until an iteration termination condition is met to obtain an updated population;
[0075] According to the fitness function, the optimal individual in the updated population is obtained; optionally, the fitness function is determined according to a first objective function and a second objective function, the first objective function is a function related to the classification error rate of the target training data set, and the second objective function is a function related to the number of the target training data set; the target training data set is the first training data set or the second training data set;
[0076] Perform probabilistic binary mapping on the optimal individual to obtain the target feature subset.
[0077] In the embodiment of the present invention, first, the fitness function is described:
[0078] The first objective function is expressed by the following formula (1):
[0079] f1=1-Acc (1)
[0080] Among them, f1 represents the classification error rate; Acc represents the classification accuracy.
[0081] The second objective function is expressed by the following formula (2):
[0082] f2=n / N (2)
[0083] Among them, n represents the number of the target training data sets. In principle, n needs to be as small as possible. The more n is, the greater the pressure on model training; N represents the number of the first training data sets or the second training data sets.
[0084] The fitness function is determined by processing the first objective function and the second objective function using a linear weighted method, and is specifically determined using the following formula (3):
[0085] f=αf1+(1-α)f2 (3)
[0086] Among them, α represents the weighting factor, which can be set according to empirical values.
[0087] Then, the initialization population is described:
[0088] The preset distribution function is a Gaussian distribution function.
[0089] Set the size of the initial population to T and the number of iterations to epoch.
[0090] Since the essence of feature selection is to generate a certain number of solutions to perform regular and efficient iterations. In order to avoid the randomness of random initialization, the embodiment of the present invention adopts directional initialization based on Gaussian distribution function, and the solution (i.e., target feature subset) is a binary array. Therefore, what affects the diversity of the population in a binary array is the distribution of 0 and 1. The probability of 0 and 1 is generated according to the following formulas (4) and (5), that is, each individual in the initialized population (i.e., target feature subset) is obtained.
[0091]
[0092] in, The meaning is to ensure that the probability of 0 appearing in the distribution of the initialized population decreases as the number of sequences increases. That is, the first individual contains the largest number of 0s, and the Nth individual contains the largest number of 1s.
[0093] Furthermore, the preset search algorithm is described:
[0094] The preset search algorithm is a CS (Cuckoo Search) algorithm.
[0095] Each individual in the initial population is iteratively updated through the Levy flight strategy in the cuckoo search algorithm. In each round of iteration, all individuals need to find new solutions and update them again. The Levy flight strategy is used in this process. The Levy flight strategy is the real flight strategy of birds in nature. If you move frequently in a small area but cannot find the real object, you will migrate a long distance to another area to search in a small area again. That is, frequent short-distance movement and low probability of long-distance migration.
[0096] Specifically, the following formulas (6), (7), and (8) are used to update each individual.
[0097]
[0098] α=α0(X i -X j ) (7)
[0099] Levy(λ)~μ=t -λ (1<λ<3) (8)
[0100] Where α is a step size coefficient, which is usually 1, but in order to make the algorithm more adaptable, the above formula (7) needs to be satisfied; X i , X j is any different solution in the search process. The step size S satisfies the Levy distribution, that is, formula (8).
[0101] Next, set the abandonment probability Pa, and generate a random number r∈(0, 1) that follows a uniform distribution to compare with the abandonment probability Pa. If r>Pa, the individual is abandoned. At the same time, through random walk, a new solution is regenerated to replace the abandoned solution. Specifically, the following formula (9) is used for random walk.
[0102]
[0103] Among them, ∈ is a random number that follows a uniform distribution, Heaviside is a jump function, and X i , X j is any different solution during the search process.
[0104] Furthermore, the values of new individuals are calculated according to the fitness function described above, and individuals with better values are retained.
[0105] Record the best individual in the current population. If the iteration termination condition is met, output the best individual. Otherwise, continue the iteration according to the Levy flight strategy.
[0106] In one implementation manner, optionally, performing probabilistic binary mapping on the optimal individual to obtain the target feature subset includes:
[0107] The optimal individual is subjected to probabilistic binary mapping through a S-type function to obtain the target feature subset.
[0108] It should be noted that since the initialized population is a discrete binary array, but after iteration, the population will become a continuous value, it is necessary to perform probabilistic binary mapping on the optimal individual through a sigmoid function to convert the optimal individual into a discrete value, that is, the target feature subset.
[0109] Here, the S-type function is expressed by the following formula:
[0110]
[0111] Among them, S i represents the position step of individual movement (i.e. update), x i is the target feature subset.
[0112] Therefore, the CS algorithm in the embodiment of the present invention is an improved CS algorithm, which can solve the problem of discretized feature selection and enable it to have stable performance in the optimization process.
[0113] Figure 2 The flowchart of the implementation method of the disk health prediction method according to the embodiment of the present invention is shown in FIG. In this implementation method, the disk health prediction method includes the following steps:
[0114] Selecting a training data set from the disk sample data set based on the feature subset;
[0115] Train a preset neural network model according to the training data set to obtain a disk health prediction model;
[0116] Using the disk health prediction model to monitor disk data in real time, and obtain prediction results indicating the health status of the disk;
[0117] Re-acquire feature subsets and disk sample data sets based on real-time monitored disk data;
[0118] Reselect the training data set from the newly acquired disk sample data set using the newly acquired feature subset;
[0119] The disk health prediction model is retrained based on the newly selected training data set.
[0120] In summary, using the disk health prediction method described in the embodiment of the present invention, the first disk sample data includes not only SMART attributes, but also disk IO attributes, system resource attributes and other data, which is comprehensive disk data, avoiding the problem of one-sided prediction, reducing the false alarm rate, and improving the accuracy of disk health prediction. Moreover, the first training data set selected based on the first feature subset retains the implicit relationship between features, avoids some redundant relationships between features, is comprehensive and irreplaceable, and further improves the accuracy and efficiency of disk health prediction.
[0121] The embodiments of the present invention, as Figure 3 As shown, a method for constructing a disk health prediction model is provided, including:
[0122] S301, generating a probability of a first value and a second value according to a preset distribution function to obtain each individual in the initialized population;
[0123] S302, using a preset search algorithm and the initialized population to obtain a first feature subset;
[0124] S303, selecting data from the first disk sample data set according to the first feature subset to obtain a first training data set; wherein each element in the first feature subset is the first value or the second value, the first value is used to indicate that data at a corresponding element position in the first disk sample data set is selected, and the second value is used to indicate that data at a corresponding element position in the first disk sample data set is not selected;
[0125] S304: Train a preset neural network model according to the first training data set to obtain a disk health prediction model.
[0126] In an embodiment of the present invention, the preset neural network model is a PNN (Product-based Neural Network, probabilistic neural network) model. Figure 4 Schematic diagram of the network structure of the preset neural network model according to the embodiment of the present invention. Figure 4 As shown, from the perspective of network structure, the entire preset neural network model consists of four layers. The first layer is the feature embedding layer, the second layer is the Product layer, and the third and fourth layers are fully connected network layers.
[0127] Specifically, the number of nodes in the input layer of the preset neural network model corresponds to the dimension of the first training data set, and each node corresponds to one data.
[0128] The feature embedding layer is used to convert discrete features into continuous vector representations.
[0129] The Product layer is used to calculate cross information between data, and the output dimension of the Product layer is the same as that of the feature embedding layer.
[0130] The fully connected network layer is used to learn higher-level feature representations, and the network depth, learning rate, loss function are set to a multi-class cross entropy loss function, activation function is Tanh, and model optimizer is Adam.
[0131] In one implementation manner, optionally, the using a preset search algorithm and the initialization population to obtain a first feature subset includes:
[0132] Iteratively updating each individual of the initialized population using a preset search algorithm until an iteration termination condition is met to obtain an updated population;
[0133] According to the fitness function, obtaining the best individual in the updated population;
[0134] Perform probabilistic binary mapping on the optimal individual to obtain the first feature subset.
[0135] Optionally, the fitness function is determined according to a first objective function and a second objective function, wherein the first objective function is a function related to a classification error rate of a target training data set, and the second objective function is a function related to the number of the target training data set;
[0136] The target training data set is the first training data set or the second training data set.
[0137] In one implementation manner, optionally, performing probabilistic binary mapping on the optimal individual to obtain the first feature subset includes:
[0138] The first feature subset is obtained by performing probability binary mapping on the optimal individual through a S-type function.
[0139] It should be noted that the specific steps of using the preset search algorithm and the initialized population to obtain the first feature subset are the same as the above disk health prediction method, which will not be repeated here.
[0140] like Figure 5 As shown, an embodiment of the present invention further provides a disk health prediction device, comprising:
[0141] The first acquisition module 501 is used to obtain a prediction result indicating the health status of the disk based on a disk health prediction model and disk data to be predicted; wherein the disk health prediction model is obtained by training a preset neural network model based on a first training data set; the first training data set is obtained by selecting data from a first disk sample data set according to a first feature subset; each element in the first feature subset is a first value or a second value, the first value is used to indicate that data at a corresponding element position in the first disk sample data set is selected, and the second value is used to indicate that data at a corresponding element position in the first disk sample data set is not selected; the first feature subset is obtained by using a preset search algorithm and an initialized population, and each individual in the initialized population is obtained by probabilistically generating the first value and the second value according to a preset distribution function.
[0142] Optionally, the disk health prediction device further comprises:
[0143] A third acquisition module, used to acquire a second feature subset;
[0144] A second selection module is used to select a second training data set from a second disk sample data set according to the second feature subset; wherein the second disk sample data set includes the first disk sample data set and the disk data to be predicted; each element in the second feature subset is a first value or a second value, the first value is used to indicate the selection of data at a corresponding element position in the second disk sample data set, and the second value is used to indicate the non-selection of data at a corresponding element position in the second disk sample data set;
[0145] The third acquisition module is used to update the disk health prediction model according to the second training data set to obtain an updated disk health prediction model.
[0146] Optionally, the disk health prediction device further comprises:
[0147] A fourth acquisition module is used to acquire a target feature subset according to the following steps; wherein the target feature subset is the first feature subset or the second feature subset;
[0148] Probabilistically generate the first value and the second value according to a preset distribution function to obtain each individual in the initialized population;
[0149] Iteratively updating each individual of the initialized population using a preset search algorithm until an iteration termination condition is met to obtain an updated population;
[0150] According to the fitness function, obtaining the best individual in the updated population;
[0151] Perform probabilistic binary mapping on the optimal individual to obtain the target feature subset.
[0152] Optionally, in the disk health prediction device, the fitness function is determined according to a first objective function and a second objective function, the first objective function is a function related to a classification error rate of a target training data set, and the second objective function is a function related to the number of the target training data set;
[0153] The target training data set is the first training data set or the second training data set.
[0154] Optionally, in the disk health prediction device, the fourth acquisition module is specifically used to:
[0155] The optimal individual is subjected to probabilistic binary mapping through a S-type function to obtain the target feature subset.
[0156] The disk health prediction device provided in the embodiment of the present invention can execute the above-mentioned disk health prediction method embodiment, and its implementation principle and technical effect are similar, which will not be repeated in this embodiment.
[0157] like Figure 6 As shown, an embodiment of the present invention further provides a device for constructing a disk health prediction model, comprising:
[0158] A first obtaining module 601 is used to generate a first value and a second value by probability according to a preset distribution function to obtain each individual in the initialized population;
[0159] A second acquisition module 602 is used to acquire a first feature subset using a preset search algorithm and the initialization population;
[0160] A first selection module 603 is used to select data from a first disk sample data set according to the first feature subset to obtain a first training data set; wherein each element in the first feature subset is the first value or the second value, the first value is used to indicate that data at a corresponding element position in the first disk sample data set is selected, and the second value is used to indicate that data at a corresponding element position in the first disk sample data set is not selected;
[0161] The second obtaining module 604 is used to train the preset neural network model according to the first training data set to obtain a disk health prediction model.
[0162] Optionally, in the device for constructing a disk health prediction model, the second acquisition module 602 includes:
[0163] An updating unit, used for iteratively updating each individual of the initialized population using a preset search algorithm until an iteration termination condition is met to obtain an updated population;
[0164] A first acquisition unit, used for acquiring the optimal individual in the updated population according to a fitness function;
[0165] The second acquisition unit is used to perform probabilistic binary mapping on the optimal individual to obtain the first feature subset.
[0166] Optionally, in the device for constructing the disk health prediction model, the fitness function is determined according to a first objective function and a second objective function, the first objective function is a function related to the classification error rate of a target training data set, and the second objective function is a function related to the number of the target training data set;
[0167] The target training data set is the first training data set or the second training data set.
[0168] Optionally, in the device for constructing a disk health prediction model, the second acquisition unit is specifically used to:
[0169] The first feature subset is obtained by performing probability binary mapping on the optimal individual through a S-type function.
[0170] The disk health prediction model construction device provided in the embodiment of the present invention can execute the above-mentioned disk health prediction model construction method embodiment, and its implementation principle and technical effect are similar, which will not be repeated in this embodiment.
[0171] like Figure 7As shown, an embodiment of the present invention also provides a disk health prediction device, including: a processor 701; and a memory 702 connected to the processor 701 through a bus interface, the memory 702 is used to store programs and data used by the processor 701 when performing operations, and the processor 701 calls and executes the programs and data stored in the memory 702.
[0172] The disk health prediction device further includes a transceiver 703 , which is connected to the bus interface and is used to receive and send data under the control of the processor 701 .
[0173] Specifically, the processor 701 is used to read the program and execute the following process:
[0174] According to the disk health prediction model and the disk data to be predicted, a prediction result for indicating the health status of the disk is obtained; wherein, the disk health prediction model is obtained by training a preset neural network model according to a first training data set; the first training data set is obtained by selecting data from a first disk sample data set according to a first feature subset; each element in the first feature subset is a first value or a second value, the first value is used to indicate that data at a corresponding element position in the first disk sample data set is selected, and the second value is used to indicate that data at a corresponding element position in the first disk sample data set is not selected; the first feature subset is obtained by using a preset search algorithm and an initialized population, and each individual in the initialized population is obtained by probabilistically generating the first value and the second value according to a preset distribution function.
[0175] Among them, Figure 7 In the embodiment, the bus architecture may include any number of interconnected buses and bridges, specifically linking together various circuits of one or more processors represented by processor 701 and memory represented by memory 902. The bus architecture may also link together various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides a user interface 704. The transceiver 703 may be a plurality of components, i.e., including a transmitter and a receiver, providing a unit for communicating with various other devices on a transmission medium. The processor 701 is responsible for managing the bus architecture and general processing, and the memory 702 may store data used by the processor 701 when performing operations.
[0176] The processor 701 is responsible for managing the bus architecture and general processing, and the memory 702 can store data used by the processor 701 when performing operations.
[0177] Optionally, in the disk health prediction device, the processor 701 is further configured to read the program and execute the following process:
[0178] Obtaining a second feature subset;
[0179] A second training data set is selected from a second disk sample data set according to the second feature subset; wherein the second disk sample data set includes the first disk sample data set and the disk data to be predicted; each element in the second feature subset is a first value or a second value, the first value is used to indicate that data at a corresponding element position in the second disk sample data set is selected, and the second value is used to indicate that data at a corresponding element position in the second disk sample data set is not selected;
[0180] The disk health prediction model is updated according to the second training data set to obtain an updated disk health prediction model.
[0181] Optionally, in the disk health prediction device, the processor 701 is further configured to read the program and execute the following process:
[0182] According to the following steps, a target feature subset is obtained; wherein the target feature subset is the first feature subset or the second feature subset;
[0183] Probabilistically generate the first value and the second value according to a preset distribution function to obtain each individual in the initialized population;
[0184] Iteratively updating each individual of the initialized population using a preset search algorithm until an iteration termination condition is met to obtain an updated population;
[0185] According to the fitness function, obtaining the best individual in the updated population;
[0186] Perform probabilistic binary mapping on the optimal individual to obtain the target feature subset.
[0187] Optionally, in the disk health prediction device, the fitness function is determined according to a first objective function and a second objective function, the first objective function is a function related to a classification error rate of a target training data set, and the second objective function is a function related to the number of the target training data set;
[0188] The target training data set is the first training data set or the second training data set.
[0189] Optionally, in the disk health prediction device, the processor 701 is specifically configured to read the program and execute the following process:
[0190] The optimal individual is subjected to probabilistic binary mapping through a S-type function to obtain the target feature subset.
[0191] like Figure 8 As shown, an embodiment of the present invention also provides a device for constructing a disk health prediction model, including: a processor 801; and a memory 802 connected to the processor 801 via a bus interface, the memory 802 being used to store programs and data used by the processor 801 when performing operations, and the processor 801 calls and executes the programs and data stored in the memory 802.
[0192] The disk health prediction model construction device further includes a transceiver 803 , which is connected to the bus interface and is used to receive and send data under the control of the processor 801 .
[0193] Specifically, the processor 801 is used to read the program and execute the following process:
[0194] Probabilistically generate the first value and the second value according to a preset distribution function to obtain each individual in the initialized population;
[0195] Using a preset search algorithm and the initialization population, obtaining a first feature subset;
[0196] Selecting data from a first disk sample data set according to the first feature subset to obtain a first training data set; wherein each element in the first feature subset is the first value or the second value, the first value is used to indicate selecting data at a corresponding element position in the first disk sample data set, and the second value is used to indicate not selecting data at a corresponding element position in the first disk sample data set;
[0197] The preset neural network model is trained according to the first training data set to obtain a disk health prediction model.
[0198] Among them, Figure 8 In the embodiment, the bus architecture may include any number of interconnected buses and bridges, specifically linking together various circuits of one or more processors represented by processor 801 and memory represented by memory 802. The bus architecture may also link together various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides a user interface 804. The transceiver 803 may be a plurality of components, i.e., including a transmitter and a receiver, providing a unit for communicating with various other devices on a transmission medium. The processor 801 is responsible for managing the bus architecture and general processing, and the memory 802 may store data used by the processor 801 when performing operations.
[0199] The processor 801 is responsible for managing the bus architecture and general processing, and the memory 802 can store data used by the processor 801 when performing operations.
[0200] Optionally, in the device for constructing the disk health prediction model, the processor 801 is specifically configured to read the program and execute the following process:
[0201] Iteratively updating each individual of the initialized population using a preset search algorithm until an iteration termination condition is met to obtain an updated population;
[0202] According to the fitness function, obtaining the best individual in the updated population;
[0203] Perform probabilistic binary mapping on the optimal individual to obtain the first feature subset.
[0204] Optionally, in the device for constructing the disk health prediction model, the fitness function is determined according to a first objective function and a second objective function, the first objective function is a function related to a classification error rate of a target training data set, and the second objective function is a function related to the number of the target training data set;
[0205] The target training data set is the first training data set or the second training data set.
[0206] Optionally, in the device for constructing the disk health prediction model, the processor 801 is specifically configured to read the program and execute the following process:
[0207] The first feature subset is obtained by performing probability binary mapping on the optimal individual through a S-type function.
[0208] A specific embodiment of the present invention also provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, it implements the steps in the above-mentioned disk health prediction method or the method for constructing a disk health prediction model, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.
[0209] In addition, an embodiment of the present invention further provides a computer program product, including computer instructions, which, when executed by a processor, implement the above Figure 1 or Figure 3 The various processes of the method embodiment shown can achieve the same technical effect, and will not be described again here to avoid repetition.
[0210] In the several embodiments provided by the present invention, it should be understood that the disclosed methods and devices can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.
[0211] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may be physically included separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional units.
[0212] The above-mentioned integrated unit implemented in the form of a software functional unit can be stored in a computer-readable storage medium. The above-mentioned software functional unit is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform some steps of the sending and receiving methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store program codes.
[0213] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.
Claims
1. A disk health prediction method, characterized in that: include: According to the disk health prediction model and the disk data to be predicted, a prediction result for indicating the health status of the disk is obtained; wherein, the disk health prediction model is obtained by training a preset neural network model according to a first training data set; the first training data set is obtained by selecting data from a first disk sample data set according to a first feature subset; each element in the first feature subset is a first value or a second value, the first value is used to indicate that data at a corresponding element position in the first disk sample data set is selected, and the second value is used to indicate that data at a corresponding element position in the first disk sample data set is not selected; the first feature subset is obtained by using a preset search algorithm and an initialized population, and each individual in the initialized population is obtained by probabilistically generating the first value and the second value according to a preset distribution function.
2. The disk health prediction method according to claim 1, characterized in that: The method further comprises: Obtaining a second feature subset; A second training data set is selected from a second disk sample data set according to the second feature subset; wherein the second disk sample data set includes the first disk sample data set and the disk data to be predicted; each element in the second feature subset is a first value or a second value, the first value is used to indicate that data at a corresponding element position in the second disk sample data set is selected, and the second value is used to indicate that data at a corresponding element position in the second disk sample data set is not selected; The disk health prediction model is updated according to the second training data set to obtain an updated disk health prediction model.
3. The disk health prediction method according to claim 1 or 2, characterized in that: The method further comprises: According to the following steps, a target feature subset is obtained; wherein the target feature subset is the first feature subset or the second feature subset; Probabilistically generate the first value and the second value according to a preset distribution function to obtain each individual in the initialized population; Iteratively updating each individual of the initialized population using a preset search algorithm until an iteration termination condition is met to obtain an updated population; According to the fitness function, obtaining the best individual in the updated population; Perform probabilistic binary mapping on the optimal individual to obtain the target feature subset.
4. The disk health prediction method according to claim 3, characterized in that: The fitness function is determined according to a first objective function and a second objective function, wherein the first objective function is a function related to a classification error rate of a target training data set, and the second objective function is a function related to the number of the target training data set; The target training data set is the first training data set or the second training data set.
5. The disk health prediction method according to claim 3, characterized in that: The optimal individual is subjected to probabilistic binary mapping to obtain the target feature subset, including: The optimal individual is subjected to probabilistic binary mapping through a S-type function to obtain the target feature subset.
6. A method for constructing a disk health prediction model, characterized in that: include: Probabilistically generate the first value and the second value according to a preset distribution function to obtain each individual in the initialized population; Using a preset search algorithm and the initialization population, obtaining a first feature subset; Selecting data from a first disk sample data set according to the first feature subset to obtain a first training data set; wherein each element in the first feature subset is the first value or the second value, the first value is used to indicate selecting data at a corresponding element position in the first disk sample data set, and the second value is used to indicate not selecting data at a corresponding element position in the first disk sample data set; The preset neural network model is trained according to the first training data set to obtain a disk health prediction model.
7. The method for constructing a disk health prediction model according to claim 6, characterized in that: The using of a preset search algorithm and the initialization population to obtain a first feature subset includes: Iteratively updating each individual of the initialized population using a preset search algorithm until an iteration termination condition is met to obtain an updated population; According to the fitness function, obtaining the best individual in the updated population; Perform probabilistic binary mapping on the optimal individual to obtain the first feature subset.
8. The method for constructing a disk health prediction model according to claim 7, characterized in that: The fitness function is determined according to a first objective function and a second objective function, wherein the first objective function is a function related to a classification error rate of a target training data set, and the second objective function is a function related to the number of the target training data set; The target training data set is the first training data set or the second training data set.
9. The method for constructing disk health prediction according to claim 7, characterized in that: Performing probabilistic binary mapping on the optimal individual to obtain the first feature subset includes: The first feature subset is obtained by performing probability binary mapping on the optimal individual through a S-type function.
10. A disk health prediction device, characterized in that: include: The first acquisition module is used to obtain a prediction result indicating the health status of the disk based on a disk health prediction model and disk data to be predicted; wherein the disk health prediction model is obtained by training a preset neural network model based on a first training data set; the first training data set is obtained by selecting data from a first disk sample data set according to a first feature subset; each element in the first feature subset is a first value or a second value, the first value is used to indicate the selection of data at a corresponding element position in the first disk sample data set, and the second value is used to indicate that data at a corresponding element position in the first disk sample data set is not selected; the first feature subset is obtained by using a preset search algorithm and an initialized population, and each individual in the initialized population is obtained by probabilistically generating the first value and the second value according to a preset distribution function.
11. A device for constructing a disk health prediction model, characterized in that: include: A first obtaining module is used to generate the first value and the second value by probability according to a preset distribution function to obtain each individual in the initialized population; A second acquisition module, used to acquire a first feature subset using a preset search algorithm and the initialization population; A first selection module is used to select data from a first disk sample data set according to the first feature subset to obtain a first training data set; wherein each element in the first feature subset is the first value or the second value, the first value is used to indicate that data at a corresponding element position in the first disk sample data set is selected, and the second value is used to indicate that data at a corresponding element position in the first disk sample data set is not selected; The second acquisition module is used to train the preset neural network model according to the first training data set to obtain a disk health prediction model.
12. A disk health prediction device, characterized in that: include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the disk health prediction method as claimed in any one of claims 1 to 5.
13. A device for constructing a disk health prediction model, characterized in that: include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein when the program is executed by the processor, the method for constructing a disk health prediction model as described in any one of claims 6 to 9 is implemented.
14. A readable storage medium, characterized in that: The readable storage medium stores a program, and when the program is executed by the processor, it implements the disk health prediction method as described in any one of claims 1 to 5, or implements the method for constructing a disk health prediction model as described in any one of claims 6 to 9.
15. A computer program product, characterized in that It comprises computer instructions, which, when executed by a processor, implement the disk health prediction method as described in any one of claims 1 to 5, or implement the method for constructing a disk health prediction model as described in any one of claims 6 to 9.