Intelligent identification method of energy sensitive data based on association mining

Abstract

The present invention relates to the field of sensitive data identification technology, and specifically to an intelligent identification method for energy sensitive data based on association mining, the method comprising: obtaining a strongly associated data sequence according to the strength of the association between energy data sequences; analyzing the data mutation in the first-order difference sequence, the slope of the energy data sequence fitting curve, and the seasonal change to obtain the influence sensitivity coefficient of each energy data sequence; obtaining the LOF abnormal score value of the energy data sequence key length, and combining the number of data transmissions of each energy data sequence to obtain the data sensitivity recognition of each energy data sequence; constructing a weighted undirected graph according to the data sensitivity recognition, and combining the Markov graph clustering algorithm to obtain the real energy sensitive data sequence. Thereby, intelligent identification of energy sensitive data based on association mining is realized, the problem that the association between energy data is relatively complex and the strongly associated data is not easy to obtain is solved, and the accuracy of identifying sensitive data in energy data is improved.

Description

Technical Field

[0001] The present application relates to the technical field of sensitive data identification, and specifically to a method for intelligently identifying energy sensitive data based on association mining. Background Art

[0002] With the advancement of digitalization, the digital content of various industries is gradually enriched. Among them, the energy industry involves a lot of important data, so the relevant processing of energy data has gradually become an important part of the research on data processing methods. Energy data usually contains a lot of sensitive data related to the enterprise, such as the enterprise's equipment energy consumption information, equipment operation data, user identity information and password information of the energy platform, etc. Once such data is leaked, it may cause huge losses, so sensitive data in energy data is usually protected.

[0003] Normally, sensitive data refers to some important information of individuals and enterprises, such as identity information, password information, address information, etc., which may leak important information. However, for energy data, in addition to the above-mentioned content, there may also be some relatively conventional data, but there may also be risks of leaking important information, such as the energy consumption of confidential equipment of enterprises. Traditional sensitive data identification methods can usually only identify some special fields, information categories, data categories, etc., which easily ignore sensitive data that may be leaked in the field of energy data, resulting in incomplete identification of sensitive data, and then ignoring energy sensitive data when protecting sensitive data. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides an energy sensitive data intelligent identification method based on association mining to solve the existing problems.

[0005] The energy sensitive data intelligent identification method based on association mining of the present invention adopts the following technical solutions:

[0006] An embodiment of the present invention provides a method for intelligently identifying energy sensitive data based on association mining, the method comprising the following steps:

[0007] Obtain each energy data sequence in each type of energy data; collect the encryption key and transmission times of each energy data;

[0008] In the same type of energy data, the strongly correlated data sequence of each energy data sequence is obtained according to the strength of the correlation between the energy data sequences; the influencing points of each energy data sequence are obtained according to the data mutation in the first-order difference sequence of each energy data sequence; the change impact intensity index of each influencing point is obtained according to the influencing point of the energy data sequence and the data change in the strongly correlated data sequence; the change correlation closeness of each energy data sequence is obtained according to the slope and seasonal intensity of the fitting curve of each energy data sequence; the impact sensitivity coefficient of each energy data sequence is obtained according to the change correlation closeness and the change impact intensity index; the LOF anomaly score value of the key length of each energy data sequence is obtained according to the length of the energy data sequence key and the LOF anomaly detection algorithm; for all categories of energy data sequences; the confidentiality strength index and transmission storage sensitivity of each energy data sequence are obtained according to the LOF anomaly score value of the key length of each energy data sequence and the number of data transmissions; the data sensitivity recognition degree of each energy data sequence is obtained according to the impact sensitivity coefficient, confidentiality strength index and transmission storage sensitivity;

[0009] Construct a true energy-sensitive data sequence based on data sensitivity identification.

[0010] Preferably, the method of obtaining the strongly correlated data sequence of each energy data sequence according to the strength of the correlation between the energy data sequences is specifically as follows:

[0011] The correlation between the b-th energy data sequence and each energy data sequence is obtained by a grey correlation analysis algorithm; the segmentation threshold of all the correlations is obtained by an Otsu threshold segmentation algorithm; and the energy data sequence whose correlation with the b-th energy data sequence is greater than or equal to the segmentation threshold is regarded as a strongly correlated data sequence of the b-th energy data sequence.

[0012] Preferably, the influencing cause points of each energy data sequence are obtained according to the data mutation in the first-order difference sequence of each energy data sequence, specifically:

[0013] The mutation points in the first-order difference sequence of each energy data sequence are obtained through the Pettitt mutation point detection algorithm; the corresponding moments of each mutation point in the first-order difference sequence of each energy data sequence are taken as the causal moments; the data points at each causal moment in each energy data sequence are taken as the influencing causal points of each energy data sequence.

[0014] Preferably, the change impact intensity index of each impact origin point is obtained according to the impact origin point of the energy data sequence and the data change in the strongly associated data sequence, specifically including:

[0015] For each strongly correlated data sequence of each energy data sequence, the data points at each originating time in the strongly correlated data sequence are taken as the starting point of the strongly correlated data sequence; the first mutation point after each of the said originating time in the first-order difference sequence of the strongly correlated data sequence is obtained, and the data point at the corresponding time of the said first mutation point in the strongly correlated data sequence is taken as the dependent change point; the specific expression for calculating the change influence intensity index of each influencing originating point of each energy data sequence is:

[0016]

[0017] In the formula, is the change impact intensity index of the ith influencing point of the bth energy data sequence, H is the total number of strongly associated data sequences of the bth energy data sequence, is the i-th starting point data value of the g-th strongly correlated data sequence of the b-th energy data sequence, is the i-th dependent point data value of the g-th strongly correlated data sequence, b i is the data value of the ith influencing point of the bth energy data sequence, is the i-th starting point of the g-th strongly correlated data sequence, f i g is the i-th dependent point of the g-th strongly correlated data sequence, T(*) is the acquisition time value corresponding to the data point, To obtain the acquisition time of the i-th starting point of the g-th strongly correlated data sequence, T(f i g ) is the acquisition time of the i-th dependent change point of the g-th strongly correlated data sequence.

[0018] Preferably, the step of obtaining the degree of correlation of changes of each energy data series according to the slope and seasonal intensity of the fitting curve of each energy data series specifically includes:

[0019] The polynomial curve fitting algorithm is used to fit the curves of each energy data series. The slope greater than or equal to 0 at the data point on each fitting curve is taken as the positive slope, and the slope less than 0 is taken as the reverse slope. The seasonal intensity of the seasonal term sequence of each energy data series is obtained through the time series decomposition algorithm. The specific expression for calculating the degree of correlation between the changes of each energy data series is as follows:

[0020]

[0021] C b is the degree of correlation of the changes in the b-th energy data sequence, exp() is an exponential function with e as the base, Z is the total number of positive slopes on the polynomial fitting curve of the b-th energy data sequence, and Z′ is the total number of reverse slopes on the polynomial fitting curve of the b-th energy data sequence. are the jth positive slopes on the polynomial fitting curves of the bth energy data sequence and the corresponding gth strongly correlated data sequence, are the j′th reverse slopes of the polynomial fitting curves of the bth energy data sequence and the gth strongly correlated data sequence, respectively. are the seasonal intensities of the bth energy data series and the corresponding gth strongly correlated data series, respectively, and α is a parameter to prevent the denominator from being zero.

[0022] Preferably, the influence sensitivity coefficient of each energy data series is obtained according to the change correlation closeness and the change influence intensity index, specifically including:

[0023] Calculate the mean of the change impact intensity index of all influencing origins of the b-th energy data sequence; take the product of the change correlation degree of the b-th energy data sequence and the mean as the impact sensitivity coefficient of the b-th energy data sequence.

[0024] Preferably, the LOF anomaly score value of each energy data sequence key length is obtained according to the length of the energy data sequence key and the LOF anomaly detection algorithm, specifically:

[0025] The length of each energy data sequence key is counted, and all energy data sequence key lengths are input into the LOF anomaly detection algorithm. The algorithm output is the LOF anomaly score value of each energy data sequence key length.

[0026] Preferably, the confidentiality strength index and transmission and storage sensitivity of each energy data sequence are obtained according to the LOF abnormality score value of the key length of each energy data sequence and the number of data transmissions, and the expressions are:

[0027]

[0028]

[0029] Where M b is the confidentiality strength index of the b-th energy data sequence, V is the total number of influencing cause points in the b-th energy data sequence, P is the total number of data points in the b-th energy data sequence, is the xth influencing point in the bth energy data sequence, is the minimum value of all data in the bth energy data sequence, LOF b ,LOF r are the key length LOF values ​​of the bth and rth energy data sequences, respectively, R is the total number of energy data sequences of all categories, and Y b is the transmission storage sensitivity of the bth energy data sequence, f b ,f rare the data transmission times of the bth and rth energy data sequences, respectively, m b ,m r are the total number of energy data series within the data category to which the bth and rth energy data series belong respectively.

[0030] Preferably, the data sensitivity recognition of each energy data sequence is obtained according to the impact sensitivity coefficient, the confidentiality strength index and the transmission and storage sensitivity, specifically including:

[0031] The sum of the confidentiality strength index and the transmission and storage sensitivity of each energy data sequence is calculated; and the product of the impact sensitivity coefficient of each energy data sequence and the sum is used as the data sensitivity recognition degree of each energy data sequence.

[0032] Preferably, the construction of a true energy sensitive data sequence according to data sensitivity recognition is specifically as follows:

[0033] All energy data sequences are regarded as nodes of an undirected graph, and the data sensitivity recognition differences between energy data sequences are regarded as weights between nodes to construct a weighted undirected graph; the weighted undirected graph is used as the input of the Markov graph clustering algorithm, and the output is each cluster; each energy data sequence in the cluster with the largest mean data sensitivity recognition is regarded as a candidate energy sensitive data sequence, and the candidate energy sensitive data sequence with the largest data sensitivity recognition is regarded as a benchmark sequence, and the Pearson correlation coefficient between each candidate energy sensitive data sequence and the benchmark sequence is calculated respectively, and all Pearson correlation coefficients are used as the transaction set input of the Apriori algorithm, and the output is a frequent item set; the candidate energy sensitive data sequence in the frequent item set is regarded as the true energy sensitive data sequence.

[0034] The present invention has at least the following beneficial effects:

[0035] The present invention analyzes the sensitivity and particularity of energy data through the impact degree of energy data, the key length of energy data, and the data transmission situation, calculates the data sensitivity recognition degree of various energy data sequences, constructs a weighted undirected graph, and obtains the real energy sensitive data sequence in combination with the Markov graph clustering algorithm; avoids the problem that the traditional sensitive data recognition method is too simple and fails to consider the correlation between data, resulting in incomplete recognition of energy sensitive data, solves the problem that the correlation between energy data is relatively complex, resulting in the difficulty in obtaining strongly correlated data, and improves the accuracy of sensitive data recognition in energy data;

[0036] The present invention obtains a strongly correlated data sequence according to the strength of the correlation between energy data sequences; obtains the influence sensitivity coefficient of each energy data sequence by analyzing the data mutation in the first-order difference sequence, the slope of the energy data sequence fitting curve and the seasonal change; performs LOF anomaly detection on the length of the energy data sequence key to obtain its LOF anomaly score; obtains the data sensitivity recognition of each energy data sequence in combination with the number of data transmissions of each energy data sequence; constructs a weighted undirected graph according to the data sensitivity recognition, inputs it into a Markov graph clustering algorithm to obtain each clustering cluster, calculates the Pearson correlation coefficient between the energy data sequences in the clustering cluster with the largest sensitivity recognition mean, and inputs it into an Apriori algorithm to obtain a true energy sensitive data sequence, thereby realizing intelligent recognition of energy sensitive data based on association mining. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. 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.

[0038] Figure 1 A flow chart of the method for intelligently identifying energy sensitive data based on association mining provided by the present invention;

[0039] Figure 2 Schematic diagram of the steps of the intelligent identification method of energy sensitive data based on association mining. DETAILED DESCRIPTION

[0040] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation method, structure, characteristics and effects of the energy sensitive data intelligent identification method based on association mining proposed by the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0041] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0042] The specific scheme of the method for intelligently identifying energy sensitive data based on association mining provided by the present invention is described in detail below with reference to the accompanying drawings.

[0043] An embodiment of the present invention provides an intelligent identification method for energy sensitive data based on association mining.

[0044] Specifically, the following energy sensitive data intelligent identification method based on association mining is provided, please refer to Figure 1 , the method comprises the following steps:

[0045] Step S001, collecting energy data from the energy data platform and performing preprocessing.

[0046] In the database of the energy data platform, the energy data stored in the energy data platform in the past month is obtained and classified according to the existing categories in the energy data platform. All energy data in each data of each category of energy data are arranged in ascending order according to their time sequence, and each energy data sequence of each category of energy data is constructed, where the bth energy data sequence of the ath category of energy data is used express.

[0047] In the process of energy data transmission, data is usually encrypted using encryption keys of different lengths according to the sensitivity and importance of the energy data sequence. The more important and sensitive the data is, the longer the key length is used, and vice versa. When the energy data platform receives data, the encryption key of the corresponding length is used to decrypt the ciphertext to obtain the plaintext data of the energy data sequence. Usually these key data are stored in the key database of the energy data platform.

[0048] For each of the above energy data, the encryption key data and transmission times data of each of the above energy data are read in the key database of the energy data platform using the pandas library function in python. Since data may be lost during the transmission process, this embodiment uses the existing data and fills the data using the mean filling method. The mean filling algorithm and the pandas library are well-known technologies, and the process is not repeated in this embodiment.

[0049] Step S002, obtain a strongly correlated data sequence according to the strength of the correlation between energy data; obtain an influencing sensitivity coefficient according to the changes in the energy data sequence and its first-order difference sequence; and construct the data sensitivity recognition of each energy data sequence in combination with the LOF anomaly score value of the key length and the number of data transmissions.

[0050] Generally speaking, sensitive data refers to data that is at risk of being leaked to individuals or companies, such as personal identity data, corporate revenue data, etc. For sensitive energy data, the definition of sensitive data is relatively broad. For example, when a certain energy data undergoes a slight change, the greater the impact it has on other energy data, the more likely it is that it is the energy consumption data of a key equipment used by the company in the production process. Then the data may be more sensitive to the company, and if it is leaked, it may cause losses to the company. Therefore, before identifying energy sensitive data, it is also necessary to analyze the impact of energy data.

[0051] Specifically, take the a-th energy data sequence as an example. All energy data sequences in the a-th energy data are taken as the input of the grey relational analysis algorithm, and the b-th energy data sequence is set is the reference sequence, and the output is the sum of each energy data sequence and the bth energy data sequence Then the correlation of all energy data sequences is used as the input of the Otsu threshold method, and the output segmentation threshold is recorded as the correlation segmentation threshold. The energy data sequence whose correlation with the bth energy data sequence is greater than or equal to the correlation segmentation threshold is regarded as the strongly correlated data sequence of the bth energy data sequence, and recorded as in The meaning is the gth strongly correlated data sequence of the bth energy data sequence; the energy data sequence whose correlation with the bth energy data sequence is lower than the correlation segmentation threshold is taken as the low-correlation data sequence of the bth energy data sequence.

[0052] The bth energy data sequence and its strongly associated data sequence Perform first-order difference processing to obtain the bth energy data sequence and its strong correlation data sequence The first-order difference sequence of where τ b The meaning of is the first-order difference sequence of the b-th energy data sequence, The meaning is the first-order difference sequence of the g-th strongly correlated data sequence of the b-th energy data sequence. Then all the first-order difference sequences are used as the input of the Pettitt mutation point detection algorithm, and the output is all the mutation points in each first-order difference sequence. The calculation of the first-order difference sequence, the grey correlation analysis and the Pettitt mutation point detection algorithm are all well-known technologies, and the process will not be repeated in this embodiment.

[0053] Considering that the mutual influence between energy data may have a certain lag, the first-order difference sequence τ bThe corresponding moment of each mutation point in the bth energy data sequence is taken as the cause moment, and the corresponding data point of each cause moment in the bth energy data sequence is taken as the impact cause point, and the strongly associated data sequence is taken as All energy data points corresponding to the cause time in are taken as the starting point, and the first-order difference sequence of the strongly correlated data sequence after the cause time is The first mutation point in the strongly correlated data sequence The data points at the corresponding time in are taken as dependent points.

[0054] For example, the first-order difference sequence τ b If the corresponding time of a certain influencing point in is 1, then the strongly correlated data sequence The data point at time 1 is the starting point, and the first-order difference sequence of the strongly correlated data sequence after time 1 is The first mutation point in the sequence corresponds to time 3. The data points at time 3 are taken as dependent points.

[0055] By strongly correlating data sequences The method for obtaining the starting point and dependent point of the b-th energy data sequence is used to obtain the starting point and dependent point of each strongly correlated data sequence.

[0056] Then the b-th energy data sequence and each of its strongly associated data sequences are respectively used as the input of the polynomial curve fitting algorithm, and the output is the polynomial fitting curve of the b-th energy data sequence and each of its strongly associated data sequences. The polynomial curve fitting algorithm is a well-known technology, and the present embodiment will not repeat its process. Obtain the slope of each data point on the polynomial fitting curve in the b-th energy data sequence, and record the slope greater than or equal to 0 as the forward slope, and record the slope less than 0 as the reverse slope. By obtaining the forward slope and reverse slope on the polynomial fitting curve of the b-th energy data sequence, the forward slope and reverse slope on the polynomial fitting curve of each strongly associated data sequence are obtained. The b-th energy data sequence and each of its strongly associated data sequences are respectively used as the input of the STL decomposition algorithm, and the output is the residual term sequence, seasonal term sequence, and trend term sequence of the b-th energy data sequence and each of its strongly associated data sequences, and the seasonal intensity of the b-th energy data sequence and each of its strongly associated data sequences is calculated using the seasonal intensity calculation formula. The STL decomposition algorithm and the slope calculation formula are both well-known technologies, and the present embodiment will not repeat the process. Construct the impact sensitivity coefficient of each energy data series:

[0057]

[0058]

[0059]

[0060] In the formula, is the change impact intensity index of the ith influencing point of the bth energy data sequence, H is the total number of strongly associated data sequences of the bth energy data sequence, A strongly correlated data sequence The i-th starting point data value, A strongly correlated data sequence The i-th dependent change point data value, b i is the data value of the ith influencing point of the bth energy data sequence, A strongly correlated data sequence The i-th starting point of i g A strongly correlated data sequence The i-th dependent variable point, T(*) is the value of the acquisition time corresponding to the data point, To obtain strongly correlated data sequences The collection time of the i-th starting point, T(f i g ) is to obtain a strongly associated data sequence The collection time of the i-th dependent variable point;

[0061] C b is the degree of correlation of the changes in the b-th energy data sequence, exp() is an exponential function with e as the base, Z is the total number of positive slopes on the polynomial fitting curve of the b-th energy data sequence, and Z′ is the total number of reverse slopes on the polynomial fitting curve of the b-th energy data sequence. are the bth energy data sequence and its strongly associated data sequence respectively The jth positive slope on the polynomial fitting curve is are the bth energy data sequence and its strongly associated data sequence respectively The j′th reverse slope of the polynomial fitting curve, are the bth energy data sequence and its strongly associated data sequence respectively The seasonal intensity, α is a parameter to prevent the denominator from being 0. It should be noted that the value of α can be set by the implementer. In this embodiment, the value of α is set to 0.001;

[0062] K b is the impact sensitivity coefficient of the bth energy data series, and E is the total number of influencing cause points of the bth energy data series.

[0063] The more similar the slope changes of the bth energy data sequence are to the other strongly correlated data sequences, and the smaller the difference in the periodicity, The larger the value, the closer the change characteristics of the energy data sequence are to the change characteristics of other strongly correlated data sequences. When the value of the influencing point in the a-th energy data sequence is smaller, but the data difference between the starting point and the dependent point of the other energy data is larger, and the time difference is smaller, The larger the value, the stronger the influence of the influencing point. When the change intensity index of each influencing variable point in the b-th energy data sequence is larger, and the change correlation of the b-th energy data sequence is more closely related, that is, The larger it is, the greater the influence of the energy data sequence on its strongly correlated data sequence, and the higher its sensitivity.

[0064] The impact sensitivity coefficient of each energy data series is obtained by calculating the impact sensitivity coefficient of the bth energy data series.

[0065] For different companies, energy sensitive data may be different. For example, for a certain company, the energy consumption data of production equipment, the corresponding equipment type, energy consumption type and other data are public to the company, or they are not sensitive data for the company (such as the monthly power consumption of textile equipment of a textile company, or the type of textile equipment corresponding to energy consumption, etc.); but it is also possible that conventional data is sensitive data for a certain company (for example, a textile company uses a new type of textile equipment, then the energy data related to the equipment may leak the company's confidential information). Therefore, companies usually perform different degrees of data encryption according to the different sensitivity of the data. In traditional encryption algorithms, the most common way to perform different degrees of encryption is to modify the key length during encryption, add data noise, and then achieve different degrees of encryption for different data. Obtain the data value of the influencing cause point in each energy data sequence and record it as in The meaning is the data value of the xth influencing cause point of the bth energy data sequence. The encryption key data of each energy data sequence is calculated using the len() function in the python library to obtain the encryption key length data of each energy data sequence. The encryption key length data of all energy data sequences is used as the input of the LOF anomaly detection algorithm, the proximity distance is set to the mean of all energy data key length data, and the output is the LOF anomaly score value of the key length data of each energy data sequence. The LOF anomaly detection algorithm and the len() function are well-known technologies, and the process will not be repeated in this embodiment. And because sensitive data has certain particularities, the number of users who can access and transmit it is relatively small, so the number of transmission accesses is relatively small, and the data in the corresponding data category is less than that of ordinary data. Based on the above characteristics, the data sensitivity recognition degree of each energy data sequence is constructed:

[0066]

[0067]

[0068] Xb=Kb×(Mb+Y,)

[0069] Where M b is the confidentiality strength index of the b-th energy data sequence, V is the total number of influencing cause points in the b-th energy data sequence, P is the total number of data points in the b-th energy data sequence, is the xth influencing point in the bth energy data sequence, is the minimum value of all data in the b-th energy data sequence, LOFb and LOFr are the key length LOF values ​​of the b-th and r-th energy data sequences respectively, R is the total number of energy data sequences of all categories, and e* is an exponential function with e as the base;

[0070] Y b is the transmission storage sensitivity of the bth energy data sequence, f b ,f r are the data transmission times of the bth and rth energy data sequences, respectively, m b ,m r are the total number of energy data series in the data category to which the bth and rth energy data series belong respectively;

[0071] X b is the data sensitivity recognition of the b-th energy data sequence, K b is the impact sensitivity coefficient of the bth energy data series.

[0072] When the difference between all the influencing cause points in the bth energy data sequence and the minimum value in the energy data sequence is larger, and the difference between the encryption key length LOF value of the bth energy data and the other energy data sequences is larger, The larger the value is, the more noise data may be added to the energy data sequence, the larger the key length is, and the higher the confidentiality strength is. When the transmission frequency of the bth energy data sequence is larger than the transmission frequency of the other energy data sequences and smaller than the other energy data sequences, the number of the bth energy data sequence is larger than the total number of data sequences stored in the same type of data compared with the total number of data sequences of the other energy data, that is, The larger the K is, the rarer the b-th energy data sequence is compared to the other energy data sequences, and the higher the sensitivity of transmission and storage. When the sensitivity coefficient of the b-th energy data sequence is larger, and the confidentiality strength index and transmission and storage sensitivity are both larger, K b × The larger the value of is, the higher the possibility that the energy data sequence is sensitive data. b The bigger.

[0073] The data sensitivity recognition degree of each energy data sequence is obtained by calculating the data sensitivity recognition degree of the bth energy data sequence.

[0074] Step S003: using the data sensitivity recognition degree of each energy data sequence, intelligently identify sensitive data of the energy data.

[0075] All energy data sequences are taken as nodes of an undirected graph, and the data sensitivity recognition differences between energy data sequences are taken as weights between nodes to construct a weighted undirected graph of energy data sequences. The weighted undirected graph of energy data sequences is taken as input, and the Markov graph clustering algorithm is used to output multiple clusters. Markov graph clustering is a well-known technology, and the process is not repeated in this embodiment. The mean of data sensitivity recognition in each cluster is calculated, and the cluster with the largest mean of data sensitivity recognition is taken as a candidate sensitive data cluster, all energy data sequences in the cluster are taken as candidate energy sensitive data sequences, and the candidate energy sensitive data sequence with the largest data sensitivity recognition is selected as a benchmark sequence, and the Pearson correlation coefficient between each candidate energy sensitive data sequence and the benchmark sequence is calculated respectively, and the Pearson correlation coefficient of each candidate energy sensitive data sequence except the benchmark sequence is taken as a set, and all sets are taken as an item set, and the item set is taken as the transaction set input of the Apriori algorithm, the minimum support is set to 2, and the output is a frequent item set. The candidate energy-sensitive data sequences included in the frequent item set are the real energy-sensitive data sequences and are marked in the database of the energy data platform. This completes the intelligent identification method of energy-sensitive data based on association mining. The steps of the above method are shown in the figure below: Figure 2 shown.

[0076] In summary, the embodiment of the present invention analyzes the sensitivity and particularity of energy data through the influence degree of energy data, the key length of energy data, and the data transmission situation, calculates the data sensitivity recognition degree of various energy data sequences, constructs a weighted undirected graph, and obtains the real energy sensitive data sequence in combination with the Markov graph clustering algorithm; avoids the problem that the traditional sensitive data recognition method is too simple and fails to consider the correlation between data, resulting in incomplete recognition of energy sensitive data, solves the problem that the correlation between energy data is relatively complex, resulting in the difficulty in obtaining strongly correlated data, and improves the accuracy of sensitive data recognition in energy data;

[0077] This embodiment obtains a strongly correlated data sequence according to the strength of the correlation between energy data sequences; obtains the influence sensitivity coefficient of each energy data sequence by analyzing the data mutation in the first-order difference sequence, the slope of the energy data sequence fitting curve and the seasonal change; performs LOF anomaly detection on the length of the energy data sequence key to obtain its LOF anomaly score; obtains the data sensitivity recognition of each energy data sequence in combination with the number of data transmissions of each energy data sequence; constructs a weighted undirected graph according to the data sensitivity recognition, inputs it into the Markov graph clustering algorithm to obtain each clustering cluster, calculates the Pearson correlation coefficient between the energy data sequences in the cluster with the largest mean sensitivity recognition, and inputs it into the Apriori algorithm to obtain the true energy sensitive data sequence, thereby realizing intelligent recognition of energy sensitive data based on association mining.

[0078] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The above is a description of a specific embodiment of this specification. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0079] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0080] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Modifications to the technical solutions recorded in the aforementioned embodiments, or equivalent replacement of some of the technical features therein, do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. An intelligent identification method for energy sensitive data based on association mining, characterized in that: The method comprises the following steps: Obtain each energy data sequence in each type of energy data; collect the encryption key and transmission times of each energy data; In the same type of energy data, the strongly correlated data sequence of each energy data sequence is obtained according to the strength of the correlation between the energy data sequences; the influencing points of each energy data sequence are obtained according to the data mutation in the first-order difference sequence of each energy data sequence; the change impact intensity index of each influencing point is obtained according to the influencing point of the energy data sequence and the data change in the strongly correlated data sequence; the change correlation closeness of each energy data sequence is obtained according to the slope and seasonal intensity of the fitting curve of each energy data sequence; the impact sensitivity coefficient of each energy data sequence is obtained according to the change correlation closeness and the change impact intensity index; the LOF anomaly score value of the key length of each energy data sequence is obtained according to the length of the energy data sequence key and the LOF anomaly detection algorithm; for all categories of energy data sequences; the confidentiality strength index and transmission storage sensitivity of each energy data sequence are obtained according to the LOF anomaly score value of the key length of each energy data sequence and the number of data transmissions; the data sensitivity recognition degree of each energy data sequence is obtained according to the impact sensitivity coefficient, confidentiality strength index and transmission storage sensitivity; Construct a true energy-sensitive data sequence based on data sensitivity identification; The influencing cause points of each energy data sequence are obtained according to the data mutation in the first-order difference sequence of each energy data sequence, specifically: The mutation points in the first-order difference sequence of each energy data sequence are obtained by the Pettitt mutation point detection algorithm; the corresponding moments of each mutation point in the first-order difference sequence of each energy data sequence are taken as the causal moments; the data points at each causal moment in each energy data sequence are taken as the influencing causal points of each energy data sequence; The change impact intensity index of each impact origin point is obtained according to the impact origin point of the energy data sequence and the data change in the strongly associated data sequence, specifically including: For each strongly correlated data sequence of each energy data sequence, the data points at each originating time in the strongly correlated data sequence are taken as the starting point of the strongly correlated data sequence; the first mutation point after each of the said originating time in the first-order difference sequence of the strongly correlated data sequence is obtained, and the data point at the corresponding time of the said first mutation point in the strongly correlated data sequence is taken as the dependent change point; the specific expression for calculating the change influence intensity index of each influencing originating point of each energy data sequence is: ； In the formula, is the change impact intensity index of the ith influencing point of the bth energy data sequence, H is the total number of strongly associated data sequences of the bth energy data sequence, is the i-th starting point data value of the g-th strongly correlated data sequence of the b-th energy data sequence, is the i-th dependent change point data value of the g-th strongly correlated data sequence, is the data value of the ith influencing point of the bth energy data sequence, is the i-th starting point of the g-th strongly correlated data sequence, is the i-th dependent change point of the g-th strongly correlated data sequence, To obtain the data point corresponding to the collection time value, To obtain the collection time of the i-th starting point of the g-th strongly correlated data sequence, To obtain the collection time of the i-th dependent change point of the g-th strongly correlated data sequence; The degree of correlation between the changes of each energy data series is obtained according to the slope and seasonal intensity of the fitting curve of each energy data series, specifically including: The polynomial curve fitting algorithm is used to fit the curves of each energy data series. The slope greater than or equal to 0 at the data point on each fitting curve is taken as the positive slope, and the slope less than 0 is taken as the reverse slope. The seasonal intensity of the seasonal term sequence of each energy data series is obtained through the time series decomposition algorithm. The specific expression for calculating the degree of correlation between the changes of each energy data series is as follows: ； is the degree of correlation between the changes in the b-th energy data series, exp() is an exponential function with e as the base, Z is the total number of positive slopes on the polynomial fitting curve of the b-th energy data series, is the total number of reverse slopes on the polynomial fitting curve of the bth energy data series, are the jth positive slopes on the polynomial fitting curves of the bth energy data sequence and the corresponding gth strongly correlated data sequence, The polynomial fitting curves of the bth energy data sequence and the gth strongly correlated data sequence are The reverse slope, are the seasonal intensities of the b-th energy data series and the corresponding g-th strongly correlated data series, To prevent the parameter with denominator equal to 0; The influence sensitivity coefficient of each energy data series is obtained according to the change correlation degree and the change impact intensity index, specifically including: Calculate the mean value of the change impact intensity index of all the influencing origin points of the b-th energy data sequence; take the product of the change correlation degree of the b-th energy data sequence and the mean value as the impact sensitivity coefficient of the b-th energy data sequence; The confidentiality strength index and transmission and storage sensitivity of each energy data sequence are obtained according to the LOF abnormality score value of the key length of each energy data sequence and the number of data transmissions. The expressions are: ； ； In the formula, is the confidentiality strength index of the b-th energy data sequence, V is the total number of influencing cause points in the b-th energy data sequence, P is the total number of data points in the b-th energy data sequence, is the xth influencing point in the bth energy data sequence, is the minimum value of all data in the b-th energy data sequence, are the key length LOF values ​​of the bth and rth energy data sequences respectively, R is the total number of energy data sequences of all categories, is the transmission storage sensitivity of the bth energy data sequence, are the data transmission times of the bth and rth energy data sequences, respectively. are the total number of energy data series in the data category to which the bth and rth energy data series belong respectively; The data sensitivity recognition of each energy data sequence is obtained according to the impact sensitivity coefficient, confidentiality strength index and transmission and storage sensitivity, specifically including: Calculate the sum of the confidentiality strength index and the transmission and storage sensitivity of each energy data sequence; take the product of the impact sensitivity coefficient of each energy data sequence and the sum as the data sensitivity recognition degree of each energy data sequence; The construction of a true energy sensitive data sequence based on data sensitivity recognition is specifically as follows: All energy data sequences are regarded as nodes of an undirected graph, and the data sensitivity recognition differences between energy data sequences are regarded as weights between nodes to construct a weighted undirected graph; the weighted undirected graph is used as the input of the Markov graph clustering algorithm, and the output is each cluster; each energy data sequence in the cluster with the largest mean data sensitivity recognition is regarded as a candidate energy sensitive data sequence, and the candidate energy sensitive data sequence with the largest data sensitivity recognition is regarded as a benchmark sequence, and the Pearson correlation coefficient between each candidate energy sensitive data sequence and the benchmark sequence is calculated respectively, and all Pearson correlation coefficients are used as the transaction set input of the Apriori algorithm, and the output is a frequent item set; the candidate energy sensitive data sequence in the frequent item set is regarded as the true energy sensitive data sequence.

2. The method for intelligently identifying energy sensitive data based on association mining according to claim 1, characterized in that: The method of obtaining a strongly correlated data sequence of each energy data sequence according to the strength of the correlation between the energy data sequences is specifically as follows: The correlation between the b-th energy data sequence and each energy data sequence is obtained by a grey correlation analysis algorithm; the segmentation threshold of all the correlations is obtained by an Otsu threshold segmentation algorithm; and the energy data sequence whose correlation with the b-th energy data sequence is greater than or equal to the segmentation threshold is regarded as a strongly correlated data sequence of the b-th energy data sequence.

3. The method for intelligently identifying energy sensitive data based on association mining according to claim 1, characterized in that: The LOF anomaly score value of each energy data sequence key length is obtained according to the length of the energy data sequence key and the LOF anomaly detection algorithm, specifically: The length of each energy data sequence key is counted, and all energy data sequence key lengths are input into the LOF anomaly detection algorithm. The algorithm output is the LOF anomaly score value of each energy data sequence key length.

Images