Magnetic disk health state prediction method and device, equipment and medium
By excluding bad value and interpolation of the transfer probability matrix in disk health status prediction, a hidden Markov model was constructed, which solved the problem of low accuracy in disk health status prediction, and achieved higher prediction accuracy and data set reliability.
Patent Information
- Application Number
- CN202510124991.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-26
- Publication Date
- 2025-04-25
AI Technical Summary
When predicting the health status of a disk in the prior art, due to factors such as disk aging, wear, signal interference, etc., the observed or counted disk data has a large deviation, which in turn makes the model predicted disk health status less accurate.
By eliminating and interpolation of bad values in the implicit state transition probability matrix and the observed state transition probability matrix, a hidden Markov model is constructed for predicting disk health status.
Improves the prediction accuracy of disk health status, reduces the risk of incorrect predictions, and enhances the accuracy and usability of implicit and observable state datasets.
Smart Images
Figure CN119883850A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of data processing technology, and in particular to a method, device, equipment and medium for predicting the health status of a disk. Background Art
[0002] Disks are the core components for storing data in computers. The health status of disks directly affects the security of data and the normal operation of computers. If a disk fails, it may cause data loss and computer crashes, thus causing huge losses to users. Therefore, timely understanding of the health status of disks is crucial to ensure data security. Generally, training models are used to predict the health status of disks. However, considering factors such as disk aging and wear, signal interference, etc., there may be large deviations in the observed or statistical disk data, which may cause large deviations in the disk health status predicted by the model based on the disk data.
[0003] Therefore, how to improve the prediction accuracy of the disk health status is a problem that those skilled in the art need to solve. Summary of the invention
[0004] The purpose of the embodiments of the present invention is to provide a disk health status prediction method, device, equipment and medium, which can construct a hidden Markov model for predicting the disk health status by removing bad values in the implicit state transition probability matrix and the observed state transition probability matrix and interpolating and completing them, so as to improve the prediction accuracy of the disk health status. The specific scheme is as follows:
[0005] In a first aspect, the present application provides a method for predicting the health status of a disk, comprising:
[0006] Based on the disk implicit state and the disk health state, an initial implicit state transition probability matrix and an initial observed state transition probability matrix are respectively constructed, and based on each element in the initial implicit state transition probability matrix and the initial observed state transition probability matrix, an initial implicit state data set and an initial observed state data set are respectively constructed;
[0007] Eliminate bad values in the initial hidden state data set and the initial observed state data set, and perform genetic operations on the hidden state data set and the observed state data set after the bad values are eliminated, so that after the genetic operations are completed, interpolate the positions of the hidden state data set and the observed state data set obtained by the genetic operations at which the bad values are eliminated, and determine a target hidden state transition probability matrix and a target observed state transition probability matrix based on the interpolated hidden state data set and the observed state data set;
[0008] A hidden Markov model is constructed based on the target implicit state transition probability matrix and the target observed state transition probability matrix, and the hidden Markov model is used to predict the health status of the disk.
[0009] Optionally, constructing a hidden Markov model based on the target implicit state transition probability matrix and the target observed state transition probability matrix, and using the hidden Markov model to predict the health status of the disk includes:
[0010] Constructing a hidden Markov model based on the initial state probability matrix, the target hidden state transition probability matrix and the target observed state transition probability matrix;
[0011] The target observation sequence for the disk health status is input into the hidden Markov model so that the hidden Markov model uses the initial state probability matrix, the target hidden state transition probability matrix and the target observation state transition probability matrix to determine the occurrence probability and the optimal hidden state sequence corresponding to the target observation sequence, and uses the occurrence probability and the optimal hidden state sequence to predict the disk health status.
[0012] Optionally, the removing bad values from the initial hidden state data set and the initial observed state data set, and performing genetic operations on the hidden state data set and the observed state data set after the bad values are removed, respectively, so as to interpolate the positions of the hidden state data set and the observed state data set obtained by the genetic operation after the bad values are removed, respectively, including:
[0013] adding the initial implicit state data set to a first population, and adding the initial observed state data set to a second population;
[0014] Remove bad values from each data set in the current population, and calculate the fitness of each data set after removing the bad values, so as to select a target data set whose fitness meets a preset fitness condition from each data set after removing the bad values contained in the current population; wherein the current population is the first population or the second population;
[0015] Performing genetic operations on the target data set to obtain a new data set, adding the new data set to the current population, calculating the fitness of the new data set after removing bad values in the current population, and then determining whether a preset end condition is currently met;
[0016] If the preset end condition is not met currently, jump back to the step of selecting the target data set whose fitness meets the preset fitness condition from the data sets after removing bad values contained in the current population;
[0017] If the preset end condition is currently met, it indicates that the genetic operation has ended, and the optimal data set with the largest fitness is selected from the data sets after bad values are removed from the current population, and the positions where bad values are removed from the optimal data set are interpolated to obtain an interpolated data set;
[0018] If the current population is the first population, the interpolated data set is an interpolated implicit state data set; if the current population is the second population, the interpolated data set is an interpolated observed state data set.
[0019] Optionally, the process of removing bad values includes:
[0020] Determine benchmark data based on each data in any data set, and calculate the deviation between each data in any data set and the benchmark data;
[0021] The target data in any of the data sets whose deviation is greater than a preset deviation threshold is determined as a bad value, and the bad values in any of the data sets are eliminated.
[0022] Optionally, selecting the target data set whose fitness satisfies a preset fitness condition from the data sets contained in the current population after removing bad values includes:
[0023] Selecting a preset number of target data sets with the largest fitness from the data sets contained in the current population after removing bad values;
[0024] Or, a target data set whose fitness is greater than a preset fitness threshold is selected from each data set included in the current population after bad values are removed.
[0025] Optionally, the preset end condition includes that the cumulative number of genetic operations for the current population reaches a preset number of genetic operations, and / or the fitness change trend for the current population tends to be stable, and / or the bad value ratio determined based on the new data set is less than a preset ratio;
[0026] Among them, the fitness change trend is a change trend determined based on the fitness corresponding to each data set after bad values are removed from the current population; the bad value ratio is a ratio determined based on the number of bad values in the new data set and the total number of data.
[0027] Optionally, interpolating the positions where bad values are removed from the optimal data set to obtain an interpolated data set includes:
[0028] Based on the position where bad values are removed from the optimal data set, determining a target position from the optimal data set; wherein the target position is a position adjacent to the position where bad values are removed from the optimal data set;
[0029] Based on the data at the target position in the optimal data set, interpolation is performed on the position where bad values are removed in the optimal data set to obtain an interpolated data set.
[0030] In a second aspect, the present application provides a disk health status prediction device, comprising:
[0031] A data set construction module, used to construct an initial implicit state transition probability matrix and an initial observed state transition probability matrix based on the disk implicit state and the disk health state, and to construct an initial implicit state data set and an initial observed state data set based on each element in the initial implicit state transition probability matrix and the initial observed state transition probability matrix;
[0032] A matrix determination module, used to remove bad values from the initial hidden state data set and the initial observed state data set, and perform genetic operations on the hidden state data set and the observed state data set after the bad values are removed, so as to interpolate the positions of the hidden state data set and the observed state data set obtained by the genetic operation at which the bad values are removed, and determine a target hidden state transition probability matrix and a target observed state transition probability matrix based on the interpolated hidden state data set and the observed state data set;
[0033] The disk status prediction module is used to construct a hidden Markov model based on the target implicit state transition probability matrix and the target observed state transition probability matrix, and use the hidden Markov model to predict the disk health status.
[0034] In a third aspect, the present application provides an electronic device, including:
[0035] Memory for storing computer programs;
[0036] A processor is used to execute the computer program to implement the steps of the aforementioned disk health status prediction method.
[0037] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the aforementioned disk health status prediction method are implemented.
[0038] In the present invention, an initial implicit state transfer probability matrix and an initial observed state transfer probability matrix are respectively constructed based on the disk implicit state and the disk health state, and an initial implicit state data set and an initial observed state data set are respectively constructed based on each element in the initial implicit state transfer probability matrix and the initial observed state transfer probability matrix; bad values in the initial implicit state data set and the initial observed state data set are eliminated, and genetic operations are respectively performed on the implicit state data set and the observed state data set after the bad values are eliminated, so that after the genetic operation is completed, the positions of the bad values eliminated in the implicit state data set and the observed state data set obtained by the genetic operation are respectively interpolated, and a target implicit state transfer probability matrix and a target observed state transfer probability matrix are determined based on the interpolated implicit state data set and the observed state data set; a hidden Markov model is constructed based on the target implicit state transfer probability matrix and the target observed state transfer probability matrix, and the disk health state is predicted using the hidden Markov model.
[0039] Beneficial effect: The present invention takes into account that there may be some bad values in the initial implicit state transfer probability matrix and the initial observed state transfer probability matrix. Therefore, the initial implicit state data set and the initial observed state data set are respectively constructed based on the initial implicit state transfer probability matrix and the initial observed state transfer probability matrix, and the bad values in the initial implicit state data set and the initial observed state data set are eliminated, so as to obtain a clean implicit state data set and an observed state data set, avoiding directly using the transfer probability matrix containing bad values to construct a hidden Markov model, thereby reducing the risk of incorrectly predicting the health status of the disk. In order to further improve the accuracy and usability of the implicit state data set and the observed state data set, the implicit state data set and the observed state data set after the bad values are eliminated are further genetically operated, and the positions of the bad values eliminated in the implicit state data set and the observed state data set obtained after the genetic operation are respectively interpolated to complete the implicit state data set and the observed state data set, and then the target implicit state transfer probability matrix and the target observed state transfer probability matrix can be determined based on the completed implicit state data set and the observed state data set to construct a hidden Markov model for predicting the health status of the disk, thereby improving the prediction accuracy of the health status of the disk. BRIEF DESCRIPTION OF THE DRAWINGS
[0040] In order to more clearly illustrate the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0041] Figure 1A flow chart of a disk health status prediction method provided by an embodiment of the present invention;
[0042] Figure 2 A schematic diagram of the structure of a disk health status prediction device provided by an embodiment of the present invention;
[0043] Figure 3 A structural diagram of an electronic device provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0045] The prediction of disk health status is generally carried out through training models. However, considering factors such as disk aging and wear, signal interference, etc., the observed or statistical disk data may have large deviations, which may cause the disk health status predicted by the model based on the disk data to have large deviations. To this end, the present invention provides a disk health status prediction method, which removes bad values in the implicit state transition probability matrix and the observed state transition probability matrix and interpolates and completes them, thereby constructing a hidden Markov model for predicting the disk health status, so as to improve the prediction accuracy of the disk health status.
[0046] See also Figure 1 As shown, an embodiment of the present invention provides a method for predicting the health status of a disk, comprising:
[0047] Step S11, constructing an initial implicit state transition probability matrix and an initial observed state transition probability matrix based on the disk implicit state and the disk health state respectively, and constructing an initial implicit state data set and an initial observed state data set based on each element in the initial implicit state transition probability matrix and the initial observed state transition probability matrix respectively.
[0048] In the embodiment of the present invention, an initial implicit state transition probability matrix is constructed based on the disk implicit state, so as to construct an initial implicit state data set based on each element in the initial implicit state transition probability matrix; an initial observed state transition probability matrix is constructed based on the disk health state, so as to construct an initial observed state data set based on each element in the initial observed state transition probability matrix. Each element in the initial implicit state transition probability matrix includes , the elements in the initial observation state transition probability matrix include For example, the initial hidden state data set is , the initial observation state data set is .
[0049] Among them, the disk implicit state refers to the disk attribute parameters that will affect the disk health status, such as the disk location, load, temperature, and can also be the remapped sector count, the total number of LBA (Logical Block Address) writes, the read error retry rate, etc., which are not specifically limited here. The disk health status includes the health status and the fault status. The fault status can be further expanded to the fault warning status, fault occurrence status, fault repair status, etc., and the health status can be further expanded to the sub-healthy status, general health status, normal health status, etc. In this way, with the refinement and expansion of the disk implicit state and the disk health status, the subsequent hidden Markov model constructed based on the implicit state transition probability matrix and the observed state transition probability matrix can be made more accurate, so that the model's disk health status prediction results are closer to the true value, thereby improving the prediction accuracy of the disk health status.
[0050] Step S12, remove bad values from the initial hidden state data set and the initial observed state data set, and perform genetic operations on the hidden state data set and the observed state data set after the bad values are removed, so that after the genetic operations are completed, the positions of the hidden state data set and the observed state data set obtained by the genetic operations where the bad values are removed are interpolated, and a target hidden state transition probability matrix and a target observed state transition probability matrix are determined based on the interpolated hidden state data set and the observed state data set.
[0051] In the embodiment of the present invention, considering that due to the aging wear of the disk or signal interference, the elements in the initial implicit state transfer probability matrix and the initial observed state transfer probability matrix may have bad values, and accordingly, the elements in the initial implicit state data set and the initial observed state data set constructed based on the initial implicit state transfer probability matrix and the initial observed state transfer probability matrix may also have bad values; therefore, it is necessary to first remove the bad values in the initial implicit state data set and the initial observed state data set, respectively, to obtain the implicit state data set and the observed state data set after removing the bad values. Further, in order to improve the accuracy and usability of the implicit state data set and the observed state data set, the implicit state data set and the observed state data set after removing the bad values can be genetically operated, and the positions of the bad values removed in the implicit state data set and the observed state data set obtained after the genetic operation are interpolated, respectively, to complete the implicit state data set and the observed state data set. Then, the target implicit state transfer probability matrix can be determined based on the interpolated implicit state data set, and the target observed state transfer probability matrix can be determined based on the interpolated observed state data set.
[0052] Specifically, an initial first population and a second population are randomly generated, wherein the initial first population contains a clean implicit state data set, wherein the clean implicit state data set is an implicit state data set that does not contain bad values; and the initial second population contains a clean observed state data set, wherein the clean observed state data set is an observed state data set that does not contain bad values. Further, the initial implicit state data set is added to the first population, and the initial observed state data set is added to the second population; the first population or the second population is used as the current population for illustration, bad values are removed from each data set in the current population, and the fitness of each data set after the bad values are removed is calculated respectively, so as to select a target data set whose fitness meets the preset fitness condition from each data set after the bad values are removed contained in the current population; then a genetic operation is performed on the target data set to obtain a new data set, and the new data set is added to the current population, and the fitness of the new data set after the bad values are removed is calculated in the current population, and then it is determined whether the preset end condition is currently met; if the preset end condition is not currently met, then jump to the above step of selecting a target data set whose fitness meets the preset fitness condition from each data set after the bad values are removed contained in the current population; if the preset end condition is currently met, it indicates that the genetic operation has ended, and the optimal data set with the largest fitness is selected from each data set after the bad values are removed contained in the current population, and the position where the bad values are removed in the optimal data set is interpolated to obtain the interpolated data set.
[0053] It should be noted that, if the current population is the first population, the interpolated data set is the interpolated implicit state data set; if the current population is the second population, the interpolated data set is the interpolated observed state data set.
[0054] For bad value removal, the benchmark data can be first determined based on each data in any data set, and the deviation between each data in any data set and the benchmark data can be calculated; then the target data in any data set whose deviation is greater than the preset deviation threshold is determined as a bad value, and the bad value in any data set is removed. Among them, any data set is any data set in the current population that has not removed the bad value; the benchmark data can be the mean square error determined based on each data in any data set; the preset deviation threshold can be set according to the user's accuracy requirements for disk health status prediction. If the accuracy requirement is high, the preset deviation threshold can be set to a smaller value.
[0055] In addition, for bad value removal, a data change trend graph can be drawn based on each data in any data set, and a corresponding data retention range can be determined based on the range where most of the data in the data change trend graph are located, and then the data in any data set that are not in the data retention range are determined as bad values, and the bad values in any data set are removed, that is, the data in any data set that are in the data retention range are retained. In this way, the embodiment of the present invention removes the bad values in the data set, and uses the data set after removing the bad values to restore the transition probability matrix, so as to use the clean transition probability matrix to construct a hidden Markov model for predicting the health status of the disk, so as to avoid directly using the transition probability matrix containing bad values to construct the hidden Markov model, and reduce the risk of incorrectly predicting the health status of the disk.
[0056] For fitness calculation, the fitness of the data set after removing bad values can be calculated by the mean square error (MSE) or likelihood function, that is, the fitness of each data in the data set after removing bad values can be calculated by the mean square error or likelihood function to obtain the fitness of the data set. Taking the mean square error MSE as an example, the fitness calculation formula is: ,in, represents fitness, Represents the first data, express is a non-bad value, correspondingly, express is a bad value, that is, for any data set The data after bad values are removed is The fitness of each data is calculated to obtain the fitness of the data set.
[0057] After calculating the fitness of each data set after removing bad values from the current population, it is necessary to select a target data set whose fitness meets the preset fitness condition from each data set after removing bad values from the current population. Specifically, one way is to select a preset number of target data sets with the largest fitness from each data set after removing bad values from the current population; another way is to select a target data set whose fitness is greater than a preset fitness threshold from each data set after removing bad values from the current population. The preset number and the preset fitness threshold can be set according to the actual needs of the user for genetic operations.
[0058] It should be noted that, in addition to the above two methods, the target data set can also be selected from the data sets after bad values are removed from the current population through the roulette algorithm or the tournament algorithm based on the fitness of each data set after bad values are removed. The probability of each data set being selected is the ratio between the fitness of the data set and the sum of the fitness of each data set after bad values are removed in the same population, that is, the data set with higher fitness has a higher probability of being selected as the target data set. In this way, a better data set can be newly generated by performing genetic operations on the data set with higher fitness.
[0059] Among them, genetic operations can include crossover operations and mutation operations. Specifically, after selecting a target data set whose fitness meets the preset fitness conditions from each data set after removing bad values from the current population, the target data set is crossovered at a certain crossover rate to generate an intermediate data set, and then some data of the intermediate data set is randomly changed at a certain mutation rate to introduce new genetic diversity, thereby obtaining a new data set. Among them, crossover operations include single-point crossover, multi-point crossover, and uniform crossover. In this way, through genetic operations, it is possible to help the newly generated data set introduce new genetic diversity and jump out of the local optimal solution.
[0060] It should be noted that the preset end condition can be that the cumulative number of genetic operations for the current population reaches the preset number of genetic operations, or that the fitness change trend for the current population tends to be stable, or that the bad value ratio determined based on the new data set is less than the preset ratio. Among them, the fitness change trend is the fitness change trend determined based on the fitness corresponding to each data set after the bad values are removed from the current population; the bad value ratio is the ratio determined based on the number of bad values in the new data set and the total number of data.
[0061] When it is determined that the preset end condition is currently met, the optimal data set with the largest fitness is selected from the data sets after bad values are removed from the current population, and based on the position where the bad values are removed in the optimal data set, a target position adjacent to the position where the bad values are removed is determined from the optimal data set, that is, the target position is a position adjacent to the position where the bad values are removed in the optimal data set, for example, the target position may be a position adjacent to the position before and after the position where the bad values are removed in the optimal data set; then, based on the data at the target position in the optimal data set, the position where the bad values are removed in the optimal data set is interpolated to obtain an interpolated data set.
[0062] The position where bad values are removed from the optimal data set is For example, the target position can be two positions adjacent to the position where bad values are removed in the optimal data set. and , then based on the data at the target position in the optimal data set, linear interpolation is performed on the position where bad values are removed in the optimal data set to complete The data at the position is then interpolated to obtain the interpolated data set. The linear interpolation calculation formula is as follows:
[0063] ;
[0064] in, express The data to be interpolated at the position, express Known data on the location, express Known data at the location.
[0065] Step S13: construct a hidden Markov model based on the target implicit state transition probability matrix and the target observed state transition probability matrix, and use the hidden Markov model to predict the health status of the disk.
[0066] In the embodiment of the present invention, a hidden Markov model is constructed based on an initial state probability matrix, a target hidden state transition probability matrix and a target observed state transition probability matrix, so as to predict the health status of a disk using the hidden Markov model.
[0067] Specifically, the target observation sequence for the disk health status is input into the hidden Markov model, so that the hidden Markov model uses the initial state probability matrix, the target hidden state transition probability matrix and the target observation state transition probability matrix to determine the occurrence probability and the optimal hidden state sequence corresponding to the target observation sequence, and then uses the occurrence probability and the optimal hidden state sequence corresponding to the target observation sequence to predict the disk health status. Among them, the occurrence probability can reflect the occurrence probability of the target observation sequence, and the optimal hidden state sequence can reflect which disk hidden state is most likely to cause the disk health status at each moment in the target observation sequence, thereby realizing the backtracking of the disk hidden state and finding the cause of the disk failure.
[0068] The disk implicit state includes position 1, load 2 and temperature 3, that is, S = {1, 2, 3}, and the disk health state includes healthy state H and fault state F, that is, O = {H, F}. The initial state probability matrix is , the target hidden state transition probability matrix is , the target observation state transition probability matrix is As an example, a hidden Markov model is constructed, and the target observation sequence "HFHHF" for the health status of the disk is input into the hidden Markov model to determine the occurrence probability corresponding to the target observation sequence and the optimal hidden state sequence, thereby realizing the prediction of the health status of the disk.
[0069] Among them, for the occurrence probability corresponding to the target observation sequence "HFHHF", the following steps are included:
[0070] (1) The probability of occurrence when the disk implicit states are “position”, “load”, and “temperature” at time t=1 and the disk health state is H:
[0071] ;
[0072] (2) The probability of occurrence when the disk implicit states are “position”, “load”, and “temperature” and the disk health state is F at time t=2:
[0073] ;
[0074] (3) The probability of occurrence when the disk implicit states are “position”, “load”, and “temperature” and the disk health state is H at time t=3:
[0075] ;
[0076] (4) The probability of occurrence when the disk implicit states are “position”, “load”, and “temperature” and the disk health state is H at time t=4:
[0077] ;
[0078] (5) The probability of occurrence when the disk implicit states are “position”, “load”, and “temperature” and the disk health state is F at time t=5:
[0079] ;
[0080] According to the above derivation, the final calculated probability of occurrence of the target observation sequence "HFHHF" is: .
[0081] For the optimal hidden state sequence corresponding to the target observation sequence "HFHHF", the following steps are included:
[0082] (1) At time t=1, the probability of each implicit state when the disk implicit states are "position", "load", and "temperature" and the disk health state is H:
[0083] ;
[0084] (2) The probability of each implicit state when the disk implicit states are "position", "load", and "temperature" at time t=2 and the disk health state is F:
[0085] ;
[0086] (3) At time t=3, the probability of each implicit state when the disk implicit states are "position", "load", and "temperature" and the disk health state is H:
[0087] ;
[0088] (4) At time t=4, the probability of each implicit state when the disk implicit states are "position", "load", and "temperature" and the disk health state is H:
[0089] ;
[0090] (5) At time t=5, the probability of each implicit state when the disk implicit states are "position", "load", and "temperature" and the disk health state is F:
[0091] ;
[0092] According to the above derivation, the optimal implicit state sequence corresponding to the target observation sequence "HFHHF" is finally determined as follows: , that is, temperature → load → load → temperature → load, thereby tracing back the disk failure.
[0093] Beneficial effect: The present invention takes into account that there may be some bad values in the initial implicit state transfer probability matrix and the initial observed state transfer probability matrix. Therefore, the initial implicit state data set and the initial observed state data set are respectively constructed based on the initial implicit state transfer probability matrix and the initial observed state transfer probability matrix, and the bad values in the initial implicit state data set and the initial observed state data set are eliminated, so as to obtain a clean implicit state data set and an observed state data set, avoiding directly using the transfer probability matrix containing bad values to construct a hidden Markov model, thereby reducing the risk of incorrectly predicting the health status of the disk. In order to further improve the accuracy and usability of the implicit state data set and the observed state data set, the implicit state data set and the observed state data set after the bad values are eliminated are further genetically operated, and the positions of the bad values eliminated in the implicit state data set and the observed state data set obtained after the genetic operation are respectively interpolated to complete the implicit state data set and the observed state data set, and then the target implicit state transfer probability matrix and the target observed state transfer probability matrix can be determined based on the completed implicit state data set and the observed state data set to construct a hidden Markov model for predicting the health status of the disk, thereby improving the prediction accuracy of the health status of the disk.
[0094] See also Figure 2 As shown, an embodiment of the present invention provides a disk health status prediction device, comprising:
[0095] The data set construction module 11 is used to respectively construct an initial implicit state transition probability matrix and an initial observed state transition probability matrix based on the disk implicit state and the disk health state, and respectively construct an initial implicit state data set and an initial observed state data set based on each element in the initial implicit state transition probability matrix and the initial observed state transition probability matrix;
[0096] A matrix determination module 12 is used to remove bad values from the initial hidden state data set and the initial observed state data set, and perform genetic operations on the hidden state data set and the observed state data set after the bad values are removed, so that after the genetic operations are completed, the positions of the hidden state data set and the observed state data set obtained by the genetic operations where the bad values are removed are interpolated, and a target hidden state transition probability matrix and a target observed state transition probability matrix are determined based on the interpolated hidden state data set and the observed state data set;
[0097] The disk status prediction module 13 is used to construct a hidden Markov model based on the target implicit state transition probability matrix and the target observed state transition probability matrix, and use the hidden Markov model to predict the disk health status.
[0098] Since the embodiments of the device part correspond to the above embodiments, please refer to the description of the embodiments of the method part for the embodiments of the device part, and will not be repeated here.
[0099] Beneficial effect: The present invention takes into account that there may be some bad values in the initial implicit state transfer probability matrix and the initial observed state transfer probability matrix. Therefore, the initial implicit state data set and the initial observed state data set are respectively constructed based on the initial implicit state transfer probability matrix and the initial observed state transfer probability matrix, and the bad values in the initial implicit state data set and the initial observed state data set are eliminated, so as to obtain a clean implicit state data set and an observed state data set, avoiding directly using the transfer probability matrix containing bad values to construct a hidden Markov model, thereby reducing the risk of incorrectly predicting the health status of the disk. In order to further improve the accuracy and usability of the implicit state data set and the observed state data set, the implicit state data set and the observed state data set after the bad values are eliminated are further genetically operated, and the positions of the bad values eliminated in the implicit state data set and the observed state data set obtained after the genetic operation are respectively interpolated to complete the implicit state data set and the observed state data set, and then the target implicit state transfer probability matrix and the target observed state transfer probability matrix can be determined based on the completed implicit state data set and the observed state data set to construct a hidden Markov model for predicting the health status of the disk, thereby improving the prediction accuracy of the health status of the disk.
[0100] Furthermore, the present application also discloses an electronic device. Figure 3 This is a structural diagram of an electronic device according to an exemplary embodiment. The content in the diagram cannot be regarded as any limitation on the scope of use of this application. The electronic device may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 is used to store a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the disk health status prediction method disclosed in any of the aforementioned embodiments. In addition, the electronic device in this embodiment may specifically be an electronic computer.
[0101] In this embodiment, the power supply 23 is used to provide working voltage for each hardware device on the electronic device; the communication interface 24 can create a data transmission channel between the electronic device and the external device, and the communication protocol it follows is any communication protocol that can be applied to the technical solution of the present application, and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs and is not specifically limited here.
[0102] In addition, the memory 22, as a carrier for storing resources, can be a read-only memory, a random access memory, a disk or an optical disk, etc. The resources stored thereon can include an operating system 221, a computer program 222, etc., and the storage method can be temporary storage or permanent storage.
[0103] The operating system 221 is used to manage and control various hardware devices on the electronic device and the computer program 222, which can be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program that can be used to complete the disk health status prediction method performed by the electronic device disclosed in any of the aforementioned embodiments, the computer program 222 can further include a computer program that can be used to complete other specific tasks.
[0104] Furthermore, the present application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, the disk health status prediction method disclosed above is implemented. For the specific steps of the method, reference may be made to the corresponding contents disclosed in the above embodiments, and no further description will be given here.
[0105] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.
[0106] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.
[0107] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.
[0108] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.
[0109] The technical solution provided by the present application is introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for general technicians in this field, according to the idea of the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.
Claims
1. A disk health status prediction method, characterized in that: include: Based on the disk implicit state and the disk health state, an initial implicit state transition probability matrix and an initial observed state transition probability matrix are respectively constructed, and based on each element in the initial implicit state transition probability matrix and the initial observed state transition probability matrix, an initial implicit state data set and an initial observed state data set are respectively constructed; Eliminate bad values in the initial hidden state data set and the initial observed state data set, and perform genetic operations on the hidden state data set and the observed state data set after the bad values are eliminated, so that after the genetic operations are completed, interpolate the positions of the hidden state data set and the observed state data set obtained by the genetic operations at which the bad values are eliminated, and determine a target hidden state transition probability matrix and a target observed state transition probability matrix based on the interpolated hidden state data set and the observed state data set; A hidden Markov model is constructed based on the target implicit state transition probability matrix and the target observed state transition probability matrix, and the hidden Markov model is used to predict the health status of the disk.
2. The disk health status prediction method according to claim 1, characterized in that: The step of constructing a hidden Markov model based on the target hidden state transition probability matrix and the target observed state transition probability matrix, and using the hidden Markov model to predict the health status of the disk includes: Constructing a hidden Markov model based on the initial state probability matrix, the target hidden state transition probability matrix and the target observed state transition probability matrix; The target observation sequence for the disk health status is input into the hidden Markov model so that the hidden Markov model uses the initial state probability matrix, the target hidden state transition probability matrix and the target observation state transition probability matrix to determine the occurrence probability and the optimal hidden state sequence corresponding to the target observation sequence, and uses the occurrence probability and the optimal hidden state sequence to predict the disk health status.
3. The disk health status prediction method according to claim 1, characterized in that: The step of removing bad values from the initial hidden state data set and the initial observed state data set, and performing genetic operations on the hidden state data set and the observed state data set after the bad values are removed, so as to interpolate the positions of the hidden state data set and the observed state data set obtained by the genetic operation after the bad values are removed, respectively, includes: adding the initial implicit state data set to a first population, and adding the initial observed state data set to a second population; Remove bad values from each data set in the current population, and calculate the fitness of each data set after removing the bad values, so as to select a target data set whose fitness meets a preset fitness condition from each data set after removing the bad values contained in the current population; wherein the current population is the first population or the second population; Performing genetic operations on the target data set to obtain a new data set, adding the new data set to the current population, calculating the fitness of the new data set after removing bad values in the current population, and then determining whether a preset end condition is currently met; If the preset end condition is not met currently, jump back to the step of selecting the target data set whose fitness meets the preset fitness condition from the data sets after removing bad values contained in the current population; If the preset end condition is currently met, it indicates that the genetic operation has ended, and the optimal data set with the largest fitness is selected from the data sets after bad values are removed from the current population, and the positions where bad values are removed from the optimal data set are interpolated to obtain an interpolated data set; If the current population is the first population, the interpolated data set is an interpolated implicit state data set; if the current population is the second population, the interpolated data set is an interpolated observed state data set.
4. The disk health status prediction method according to claim 3, characterized in that: The process of removing bad values includes: Determine benchmark data based on each data in any data set, and calculate the deviation between each data in any data set and the benchmark data; The target data in any of the data sets whose deviation is greater than a preset deviation threshold is determined as a bad value, and the bad values in any of the data sets are eliminated.
5. The disk health status prediction method according to claim 3, characterized in that: The step of selecting the target data set whose fitness satisfies the preset fitness condition from the data sets contained in the current population after removing bad values includes: Selecting a preset number of target data sets with the largest fitness from the data sets contained in the current population after removing bad values; Or, a target data set whose fitness is greater than a preset fitness threshold is selected from each data set included in the current population after bad values are removed.
6. The disk health status prediction method according to claim 3, characterized in that: The preset termination conditions include that the cumulative number of genetic operations for the current population reaches a preset number of genetic operations, and / or the fitness change trend for the current population tends to be stable, and / or the bad value ratio determined based on the new data set is less than a preset ratio; Among them, the fitness change trend is a change trend determined based on the fitness corresponding to each data set after bad values are removed from the current population; the bad value ratio is a ratio determined based on the number of bad values in the new data set and the total number of data.
7. The disk health status prediction method according to any one of claims 3 to 6, characterized in that: The interpolating the positions where bad values are removed from the optimal data set to obtain an interpolated data set includes: Based on the position where bad values are removed from the optimal data set, determining a target position from the optimal data set; wherein the target position is a position adjacent to the position where bad values are removed from the optimal data set; Based on the data at the target position in the optimal data set, interpolation is performed on the position where bad values are removed in the optimal data set to obtain an interpolated data set.
8. A disk health status prediction device, characterized in that: include: A data set construction module, used to construct an initial implicit state transition probability matrix and an initial observed state transition probability matrix based on the disk implicit state and the disk health state, and to construct an initial implicit state data set and an initial observed state data set based on each element in the initial implicit state transition probability matrix and the initial observed state transition probability matrix; A matrix determination module, used to remove bad values from the initial hidden state data set and the initial observed state data set, and perform genetic operations on the hidden state data set and the observed state data set after the bad values are removed, so as to interpolate the positions of the hidden state data set and the observed state data set obtained by the genetic operation at which the bad values are removed, and determine a target hidden state transition probability matrix and a target observed state transition probability matrix based on the interpolated hidden state data set and the observed state data set; The disk status prediction module is used to construct a hidden Markov model based on the target implicit state transition probability matrix and the target observed state transition probability matrix, and use the hidden Markov model to predict the disk health status.
9. An electronic device, characterized in that: include: Memory for storing computer programs; A processor is used to execute the computer program to implement the steps of the disk health status prediction method as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the disk health status prediction method according to any one of claims 1 to 7 are implemented.