A method and device for pre-storage of data

Abstract

The present invention provides a data pre-access method and device. The method includes the following steps: obtaining a data page to be predicted, extracting data features of the data page to be predicted; inputting the data features of the data page to be predicted into a pre-trained prediction model to obtain data-related values ​​and a target data page; confirming that the data-related value is greater than a preset threshold, then pre-accessing the target data page; wherein the pre-trained prediction model is trained based on historical call data page samples. The present invention uses a machine learning model to obtain the data page access pattern, thereby greatly improving the accuracy of data page pre-access, which is of great significance to the development of database systems, big data systems and other fields.

Description

Technical Field

[0001] The present invention relates to the field of machine learning technology, and in particular to a data pre-access method and device. Background Art

[0002] In recent years, although the capacity of memory on servers has greatly increased, the amount of data stored in databases and big data systems is often much larger than the available memory. Therefore, data in databases and big data systems is mainly stored on disk and read into memory for processing when needed. Since reading data from disk is much slower than accessing data in memory, it is necessary to put the data pages that may be used into the buffer in advance, which is called data page prefetching.

[0003] For example, the storage unit widely used in solid-state drives is generally NAND flash memory, which cannot directly perform overwrite operations on the location where the original data is stored. The specific reasons are that, first, the bytes in the NAND flash memory can only be changed from 1 to 0, and vice versa. In addition, for the write operation of the solid-state drive, the smallest operation unit is the page. Based on the above two characteristics, all write operations need to be performed on a new page with all bytes being 1, so that data storage is achieved by changing the bytes that need to be changed to 0 in the new page full of 1s to 0 and retaining the bytes that should be 1.

[0004] In order to improve the access performance of fixed hard disks, the prior art includes a stride-based data prefetching technology. The stride-based data prefetching technology in the prior art has at least the following disadvantages: when there are many types of applications and access modes involved in the data access process of the solid-state hard disk, especially when the stored data is unstructured data such as files, non-related database big data, etc., the prefetching performance of the stride-based data prefetching will be reduced or even unable to work.

[0005] Meanwhile, existing data page prefetching methods are mainly based on heuristic methods and cannot effectively cope with random access. Summary of the invention

[0006] The present invention provides a data pre-access method and device, which are used to solve the defect that the data pages pre-accessed by rules in the prior art cannot accurately meet the calling requirements, so as to enable database system and big data system developers to pre-access data pages more accurately in their systems and greatly improve the efficiency of data page pre-access.

[0007] The present invention provides a data pre-access method, the method comprising:

[0008] Acquire a data page to be predicted, and extract data features of the data page to be predicted;

[0009] Inputting the data features of the data page to be predicted into a pre-trained prediction model to obtain data-related values ​​and a target data page;

[0010] If it is confirmed that the data-related value is greater than a preset threshold, the target data page is pre-accessed;

[0011] The pre-trained prediction model is trained based on historical call data page samples.

[0012] According to a data pre-access method provided by the present invention, the training process of the prediction model includes:

[0013] Acquire an initial prediction model and the historical call data page sample, and extract an access sequence of the historical call data page sample;

[0014] converting the access sequence into a data page input vector;

[0015] Determining a correlation result of the data page input vector;

[0016] The data page input vector and the correlation result are input into the initial prediction model for training to obtain the prediction model.

[0017] According to a data pre-access method provided by the present invention, the step of converting the access sequence into a data page input vector specifically comprises the following steps:

[0018] Converting the access sequence into a data page access triplet sequence;

[0019] Confirming that the number of input windows of the initial prediction model is N;

[0020] Divide the data page access triplet sequence into a plurality of groups of input vectors according to N groups;

[0021] Wherein, N is a positive integer.

[0022] According to a data pre-access method provided by the present invention, the result of determining the correlation of the data page input vector is specifically:

[0023] The correlation of the N data page triplet sequences in each of the input vectors is determined to obtain a correlation result of the input vector.

[0024] According to a data pre-access method provided by the present invention, if the data-related value is less than or equal to the preset threshold, it is determined that the data page to be predicted does not meet the pre-access condition.

[0025] According to a data pre-access method provided by the present invention, the initial prediction model is one of the following:

[0026] SVM model, CNN model, RNN model or LSTM model.

[0027] The present invention also provides a data pre-access device, the device comprising:

[0028] An extraction module, used to obtain a data page to be predicted and extract data features of the data page to be predicted;

[0029] A prediction module, used for inputting the data features of the data page to be predicted into a pre-trained prediction model to obtain data-related values ​​and a target data page;

[0030] A judgment module, used for confirming that the data-related value is greater than a preset threshold, and then pre-accessing the target data page;

[0031] The pre-trained prediction model is trained based on historical call data page samples.

[0032] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, any of the above-mentioned data pre-access methods is implemented.

[0033] The present invention also provides a non-transitory computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the data pre-access method described in any one of the above methods is implemented.

[0034] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any of the above-mentioned data pre-access methods.

[0035] The present invention provides a data pre-access method and device, which use a machine learning model to obtain a data page access pattern, thereby greatly improving the accuracy of data page pre-access, and has important significance for the development of database systems, big data systems and other fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are 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.

[0037] Figure 1 This is one of the flow charts of a data pre-access method provided by the present invention;

[0038] Figure 2This is a second flow chart of a data pre-access method provided by the present invention;

[0039] Figure 3 This is a third flow chart of a data pre-access method provided by the present invention;

[0040] Figure 4 It is a structural schematic diagram of a data pre-access device provided by the present invention;

[0041] Figure 5 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are 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.

[0043] Data page refers to the unit of data exchange from external storage (such as disk or SSD) to internal storage in database system and big data system. Each data page has an integer number. The physical storage structure of data pages in the data file on the disk. All data in the database (including various tables and data in the tables) are ultimately stored in disk files, and the physical format stored in the file is the data page. A large number of data pages are stored one by one in sequence, and then two adjacent data pages are referenced by a bidirectional linked list. A data page is a piece of data in the disk file, which may be binary or other special format data. The data page contains two pointers: the physical address of the previous data page and the physical address of the next data page. The actual storage of MySQL is roughly similar to this. Each data page is a continuous piece of data in the disk file. Then in each data page, it can be considered as a continuous piece of data in the disk from the beginning of DataPage to the || symbol. You can think of each data page as such a continuous thing in the disk file. Then each data page: has a pointer pointing to the starting physical position of its previous data page in the disk file, such as linked_list_pre_pointer = 15367, which points to the starting physical position of the previous data page in the disk file. 15367 can be considered as the position or offset in the disk file. There is also a pointer pointing to the physical position of its next data page. This can help us understand how multiple data pages in a disk file form a bidirectional linked list through pointers.

[0044] A data page stores rows of data, that is, rows of data inserted into the table are stored in the data page. Each row of data in the data page is sorted and stored by the size of the primary key, and each row of data has a pointer pointing to the location of the next row of data, forming a one-way linked list.

[0045] In the prior art, the heuristic pre-access is generally based on certain rules. When the preset rules are met, the data page is pre-accessed, and when the preset rules are not met, the data page is not pre-accessed. The rules need to be constantly changed and adjusted to meet the user's call requirements.

[0046] At the same time, research has found that the stride-based data prefetching technology in the prior art has at least the following disadvantages: when there are many types of applications and access modes involved in the data access process of the solid-state drive, especially when the stored data is unstructured data such as files, non-related database big data, etc., the prefetching efficiency of the stride-based data prefetching will be reduced or even unable to work.

[0047] The present invention provides a novel data pre-access method, which will greatly improve the efficiency of data page pre-access. The technical solution of the present invention can realize efficient data page pre-access, which is of great significance to the development of database systems and big data systems.

[0048] Combine the following Figure 1-Figure 3 A data pre-access method of the present invention is described.

[0049] Figure 1 This is one of the flow charts of a data pre-access method provided by an embodiment of the present invention.

[0050] like Figure 1 As shown, a data pre-access method provided by an embodiment of the present invention includes the following steps:

[0051] Step 110, obtain the data page to be predicted, and extract the data features of the data page to be predicted. In this step, the data page to be predicted is first prepared, that is, the data page access sequence of the data page to be predicted is obtained. After the access sequence is obtained, the data page number in the data page access sequence to be predicted is converted into a data page access triple sequence, and the data page access triple sequence is used as the data feature of the data page to be predicted.

[0052] In some implementations, the triplet corresponding to the data page is <number, group number, sequence number within the group>, that is, each triplet includes the number, group number, and sequence number within the group of the corresponding data page.

[0053] A triple is a set of the form ((x, y), z) (that is, a triple is a pair whose first projection is also a pair), often abbreviated as (x, y, z). Triples are a concept in a common basic course for computer science majors - data structures. It is mainly used as a compression method to store sparse matrices, also called a triple table. Assuming that a triple table is represented by a sequential storage structure, a compressed storage method for sparse matrices is obtained, namely a triple sequence table, referred to as a triple table. For large sparse matrices that appear in practical problems, if they are stored in a computer using conventional allocation methods, a large amount of memory will be wasted, and a large amount of time will be wasted when accessing and operating. In order to solve this problem, a variety of solutions have been produced. Due to its own sparse characteristics, the memory cost of sparse matrices can be greatly saved by compression. The specific operation is: the row, column and value of the non-zero element are formed into a triple (i, j, v), and then these triples are stored according to a certain rule. This method can save storage space.

[0054] The so-called "triplet" refers to the geometric elements of a graph, the topological relationship between graph lines, and the size constraints. If the first two elements of a group of graphs are the same but only the size is different, then this group of graphs constitutes a family of serialized graphs with the same shape. The three numbers that make up an element are called a triple. A triple contains the following three parts: SDO_STARTING_OFFSET indicates the storage location of the first coordinate of each geometric element in the SDO_ORDINATES array. …Mt:N2) is called a triple...Mt is called a label, N1 and N2 are nodes, and R is a relationship. When n≠0, Li is called a modification of node N1. When t≠0, Mj is called a modification of node N2.

[0055] Step 120: Input the data features of the data page to be predicted into the pre-trained prediction model to obtain data-related values ​​and target data pages. The pre-trained prediction model is trained based on historical call data pages. The purpose of using the prediction model is to determine the data pages that the user may access, and provide a method for implementing prediction for the database and big data system.

[0056] Specifically, the data access triplet sequence is input into a pre-trained machine learning model M to obtain a data-related value p and A data page codes to be predicted.

[0057] Step 130, confirm that the data correlation value is greater than the preset threshold, and pre-access the target data page. Optionally, if the data correlation value is less than or equal to the preset threshold, it is determined that the data page to be predicted does not meet the pre-access condition. Specifically, the user determines the preset threshold T, and if p>T, then A data page codes to be predicted are returned. If p<=T, then empty is returned. Since in actual applications, the prediction model cannot make an absolutely accurate prediction of the data page to be predicted, the user can set the preset threshold T according to the actual situation, such as according to the performance of the prediction model, whether the detected related data pages meet the requirements of the database call, etc.

[0058] The present invention is a data pre-access method using machine learning, and this embodiment provides a method for implementing data preparation for a database and a big data system. A user uses a benchmark test such as TPC-H or a customized workload and data to run in a database or a big data system for a certain period of time to obtain a data page access sequence.

[0059] Figure 2 This is the second flow chart of a data pre-access method provided by an embodiment of the present invention.

[0060] like Figure 2 As shown, an embodiment of the present invention provides a data pre-access method, in which a pre-trained prediction model is used in step 120, and the training process of the prediction model includes the following steps:

[0061] Step 210: Obtain an initial prediction model and historical call data page samples, and extract access sequences of the historical call data page samples.

[0062] In a specific implementation process, the user selects the initial prediction model, namely the machine learning model M. Machine learning is a branch of artificial intelligence. The research history of artificial intelligence has a natural and clear thread from focusing on "reasoning", to focusing on "knowledge", and then to focusing on "learning". Obviously, machine learning is a way to achieve artificial intelligence, that is, to solve problems in artificial intelligence by machine learning. A machine learning model is an expression of an algorithm that combs through massive amounts of data to find patterns or make predictions. Machine learning (ML) models are powered by data and are the mathematical engines of artificial intelligence. For example, ML models for computer vision can recognize cars and pedestrians in real-time videos. ML models for natural language processing can translate words and sentences.

[0063] In the present invention, the machine learning model M can be a traditional model such as SVM (support vector machine), as well as a deep learning model CNN (convolutional neural network), RNN (recurrent neural network) and LSTM (long short-term memory model), etc. The user here refers to the developer of the database and big data system.

[0064] Furthermore, in the embodiment of the present invention, deep learning (DL) and reinforcement learning can be combined to construct the prediction model in the embodiment of the present invention, so that the prediction model can not only automatically and dynamically adjust to ensure that the output result does not degrade in performance (i.e., it has the effect of reinforcement learning), but also can use deep learning to simplify the data features used as input or generate new data features, thereby reducing the number of parameters for adjusting the input parameters of the prediction model. Based on the characteristics of deep learning, as time goes by, the prediction model in the embodiment of the present invention can continuously accumulate experience, thereby continuously improving the effect.

[0065] When extracting the access sequence of historical call data pages, you can use benchmarks such as TPC-H in the process of preparing the data page access sequence. Before training the prediction model, you must first obtain the user's usage habits. However, in actual application scenarios, you will not actually let a user collect usage data, but let TPC-H simulate the user, which is equivalent to TPC-H simulating a standard person or standard tester. Then obtain the data usage habits.

[0066] TPC-H is a performance evaluation benchmark for decision support systems developed by the TPC organization. It is used to evaluate the server's capabilities in data mining, analysis and processing. The TPC-H test model is a sales and management system for the retail industry. It defines 8 tables and 22 queries. The test has the characteristics of complexity, diverse access patterns, and randomness. It is very similar to actual applications and has a wide range of reference significance. The TPC organization only stipulates the test rules of TPC-H and provides tools for generating data used in the TPC-H test and 22 query statements. Each manufacturer can choose the hardware, database, operating system and other software and hardware environments, and choose the test data size from 1GB, 10GB, 30GB, ..., 30000GB, 100000GB test data scale. All data query statements must be generated by the official tool. The benchmark test includes 22 queries (Q1~Q22), and its main evaluation indicator is the response time of each query, that is, the time required from submitting a query to returning the result. The unit of measurement for the TPC-H benchmark is the number of queries executed per hour, QphH@size, where H represents the average number of complex queries executed by the system per hour, and size represents the size of the database, which reflects the system's ability to process queries.

[0067] In step 210, the purpose of using TPC-H is to simulate the operation of using data by users, thereby generating a historical call data page.

[0068] Step 220: Convert the access sequence of the historical call data page into a data page input vector.

[0069] Specifically, Figure 3 This is the third flow chart of a data pre-access method provided by an embodiment of the present invention.

[0070] like Figure 3 As shown, an embodiment of the present invention provides a data pre-access method, in step 220, converting the access sequence of the historical call data page into a data page input vector, specifically comprising the following steps:

[0071] Step 310: Convert the access sequence of historically called data pages into a data page access triplet sequence. In some implementations, the triplet corresponding to the data page is <number, group number, sequence number within the group>. That is, each triplet includes the number, group number, and sequence number within the group of the corresponding data page.

[0072] In actual operation, the user determines the number of input windows of M to be N, and divides the data page access triplet sequence into input vectors according to every N triplet.

[0073] Step 320: confirm that the number of input windows of the initial prediction model is N, where N is a positive integer. In the actual training process, the user or the standard tester will do many things, but during training, only the most recent things done by the user or the standard tester will be considered. The window refers to the most recent operation, such as the most recent few seconds, or the most recent few operations.

[0074] Step 330: Divide the data page access triple sequence into multiple groups of input vectors in groups of N. The user's actions include multiple operations, and n operations are selected as training. For example, if the user performs 100,000 operations, 50,000 operations can be selected as training, and N is 50,000.

[0075] During the specific operation, adjacent data page access triplet sequences are grouped together, and the feature similarity of adjacent data page access triplet sequences is relatively high.

[0076] Step 230: Determine the relevance result of the data page input vector. In this step, the user determines the relevance result value of the input vector, similar to labeling training data in machine learning, to determine whether the data meets the actual call requirements.

[0077] Furthermore, in a data pre-access method of the present invention, the correlation result of the data page input vector is determined in step 230, specifically:

[0078] Determine the correlation of the N data page triple sequences in each input vector, and thus determine the correlation result of the input vector. In actual applications, for each of the N triples in the input vector after segmentation, the user determines K results as positive and the others as negative. The K value is specified by the user.

[0079] Step 240: Input the data page input vector and the correlation result into the initial prediction model for training to obtain a prediction model. Specifically, the input vector and the corresponding result are used to train the machine learning model M, thereby obtaining the trained prediction model M. The trained prediction model can more accurately understand the user's usage habits.

[0080] In order to effectively obtain the data page access pattern, the technical solution of the present invention utilizes the idea of ​​discretization to formalize data page prefetching as a classification problem, and uses a machine learning model to obtain the data page access pattern, thereby greatly improving the accuracy of data page prediction, which is of great significance to the development of fields such as database systems and big data systems.

[0081] A data pre-access device provided by the present invention is described below. The data pre-access device described below and the data pre-access method described above can be referred to each other.

[0082] Figure 4 It is a structural schematic diagram of a data pre-access device provided by an embodiment of the present invention.

[0083] like Figure 4 As shown, an embodiment of the present invention provides a data pre-access device, which includes the following modules: an extraction module 410 , a prediction module 420 and a judgment module 430 .

[0084] Specifically, the extraction module 410 is used to obtain the data page to be predicted and extract the data features of the data page to be predicted. The prediction module 420 is used to input the data features of the data page to be predicted into the pre-trained prediction model to obtain the data correlation value and the target data page. The judgment module 430 is used to confirm the size of the data correlation value relative to the preset threshold value. When the data correlation value is greater than the preset threshold value, the target data page is pre-accessed. Among them, the pre-trained prediction model is trained based on the historical call data page sample.

[0085] In order to effectively obtain the data page access pattern, a data pre-storage device of the present invention utilizes the idea of ​​discretization to formalize data page pre-fetching as a classification problem, and uses a machine learning model to obtain the data page access pattern, thereby greatly improving the accuracy of data page prediction, which is of great significance to the development of fields such as database systems and big data systems.

[0086] Figure 5 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 5 As shown, the electronic device may include: a processor 510, a communication interface 520, a memory 530 and a communication bus 540, wherein the processor 510, the communication interface 520 and the memory 530 communicate with each other through the communication bus 540. The processor 510 may call the logic instructions in the memory 530 to execute a data pre-access method, which includes the following steps: obtaining a data page to be predicted, extracting data features of the data page to be predicted; inputting the data features of the data page to be predicted into a pre-trained prediction model to obtain a data correlation value and a target data page; confirming that the data correlation value is greater than a preset threshold, then pre-accessing the target data page; wherein the pre-trained prediction model is obtained by training based on historical call data page samples.

[0087] Electronic devices that apply this method use the idea of ​​discretization to formalize data page prefetching as a classification problem, and use machine learning models to obtain the data page access pattern, thereby greatly improving the accuracy of data page prediction, which is of great significance to the development of fields such as database systems and big data systems.

[0088] In addition, the logic instructions in the above-mentioned memory 530 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product 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 all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0089] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute a data pre-access method provided by the above methods, which method includes the following steps: obtaining a data page to be predicted and extracting data features of the data page to be predicted; inputting the data features of the data page to be predicted into a pre-trained prediction model to obtain data-related values ​​and a target data page; confirming that the data-related value is greater than a preset threshold, then pre-accessing the target data page; wherein the pre-trained prediction model is trained based on historical call data page samples.

[0090] The computer program product that applies this method uses the idea of ​​discretization to formalize data page prefetching as a classification problem, and uses a machine learning model to obtain the data page access pattern, thereby greatly improving the accuracy of data page prediction, which is of great significance to the development of fields such as database systems and big data systems.

[0091] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, it is implemented to execute a data pre-access method provided by the above-mentioned methods, and the method includes the following steps: obtaining a data page to be predicted, and extracting data features of the data page to be predicted; inputting the data features of the data page to be predicted into a pre-trained prediction model to obtain data-related values ​​and a target data page; confirming that the data-related value is greater than a preset threshold, then pre-accessing the target data page; wherein the pre-trained prediction model is trained based on historical call data page samples.

[0092] The non-transitory computer-readable storage medium to which this method is applied utilizes the idea of ​​discretization to formalize data page prefetching as a classification problem, and uses a machine learning model to obtain the data page access pattern, thereby greatly improving the accuracy of data page prediction, which is of great significance to the development of fields such as database systems and big data systems.

[0093] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0094] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0095] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A data pre-access method, It is characterized in that The method comprises: Acquire a data page to be predicted, and extract data features of the data page to be predicted; Inputting the data features of the data page to be predicted into a pre-trained prediction model to obtain data-related values ​​and a target data page; If it is confirmed that the data-related value is greater than a preset threshold, the target data page is pre-accessed; Wherein, the pre-trained prediction model is trained based on historical call data page samples; The data feature extraction process of the data page to be predicted is as follows: Prepare the data page to be predicted, that is, obtain the data page access sequence of the data page to be predicted. After obtaining the access sequence, convert the data page number in the data page access sequence to be predicted into a data page access triplet sequence, and use the data page access triplet sequence as the data feature of the data page to be predicted.

2. The data pre-access method according to claim 1, It is characterized in that The training process of the prediction model includes: Acquire an initial prediction model and the historical call data page sample, and extract an access sequence of the historical call data page sample; converting the access sequence into a data page input vector; Determining a correlation result of the data page input vector; The data page input vector and the correlation result are input into the initial prediction model for training to obtain the prediction model.

3. The data pre-access method according to claim 2, It is characterized in that Converting the access sequence into a data page input vector specifically comprises the following steps: Converting the access sequence into a data page access triplet sequence; Confirming that the number of input windows of the initial prediction model is N; Divide the data page access triplet sequence into a plurality of groups of input vectors according to N groups; Wherein, N is a positive integer.

4. The data pre-access method according to claim 2, It is characterized in that Determine the correlation result of the data page input vector, specifically: The correlation of the N data page triplet sequences in each of the input vectors is determined to obtain a correlation result of the input vector.

5. The data pre-access method according to claim 1, It is characterized in that If the data related value is less than or equal to the preset threshold, it is determined that the data page to be predicted does not meet the pre-access condition.

6. The data pre-access method according to claim 2, It is characterized in that The initial prediction model is one of the following: SVM model, CNN model, RNN model or LSTM model.

7. A data pre-access device, It is characterized in that The device comprises: An extraction module, used to obtain a data page to be predicted and extract data features of the data page to be predicted; A prediction module, used for inputting the data features of the data page to be predicted into a pre-trained prediction model to obtain data-related values ​​and a target data page; A judgment module, used for confirming that the data-related value is greater than a preset threshold, and then pre-accessing the target data page; Wherein, the pre-trained prediction model is trained based on historical call data page samples; The data feature extraction process of the data page to be predicted is as follows: Prepare the data page to be predicted, that is, obtain the data page access sequence of the data page to be predicted. After obtaining the access sequence, convert the data page number in the data page access sequence to be predicted into a data page access triplet sequence, and use the data page access triplet sequence as the data feature of the data page to be predicted.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the program, a data pre-access method as claimed in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, a data pre-access method as claimed in any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program, It is characterized in that When the computer program is executed by a processor, a data pre-access method as claimed in any one of claims 1 to 6 is implemented.

Images