A data redundancy encoding method based on deterministic network cooperative transmission

Abstract

The application relates to a data redundancy coding method based on a deterministic network cooperative transmission, comprising the following steps: calculating node-to-node cooperative transmission distortion, single data redundancy coding, multi-data redundancy coding, calculating a deterministic network system distortion, quantifying the correlation between redundancy coding parameters, updating the redundancy coding parameters, quantifying the relationship between different data redundancy coding parameters, determining optimal redundancy coding parameters, determining optimal redundancy coding parameters and the number of data packets according to the above steps, and designing a redundancy transmission system. The relationship between the number of packets, redundancy coding parameters and quantization coefficients during multi-packet redundancy transmission is derived through nonlinear fitting, the total distortion cost in the case is modeled, optimal parameters are selected for multi-packet redundancy transmission system design, and the link distortion caused by redundancy transmission can be minimized while high reliability is ensured.

Description

Technical Field

[0001] This application relates to the field of communication transmission, and specifically to a data redundancy coding method based on deterministic network cooperative transmission. Background Technology

[0002] With the booming development of new energy sources such as solar photovoltaics and electric vehicles, a new type of energy internet based on resources has become a promising technology. Deterministic networking technology can meet the networking requirements of deterministic power applications, providing highly reliable and low-latency connections. Through link redundancy and collaborative backup of multiple transmission links, it prevents service interruptions caused by network failures, packet loss, and other anomalies, improving the reliability of end-to-end user plane data transmission. Promoting the deep integration of deterministic networking technology and the energy internet provides a solid theoretical foundation for achieving real-time, high-quality, and highly reliable energy data transmission via deterministic networks.

[0003] To improve the reliability of deterministic network transmission, a reasonable data redundancy coding method needs to be designed. However, achieving this goal faces the following challenges: First, in multi-link cooperative transmission, the impact of the correlation between heterogeneous multimodal data on the distortion caused by data redundancy transmission needs to be considered. Second, in multi-link cooperative transmission, single-packet redundancy may not meet the distortion tolerance, often requiring multi-packet redundancy. Transmission under multi-packet redundancy becomes more complex, and quantifying the relationship between various parameters becomes an important issue. Finally, after redundancy coding of a packet, parameter updates are necessary, often based on sequence features. However, these features may have internal redundancy, thus requiring a feature extraction method for dimensionality reduction to meet the real-time processing requirements of deterministic networks. Summary of the Invention

[0004] The purpose of this application is to provide a data redundancy coding method based on deterministic network cooperative transmission. It adopts a heterogeneous data similarity matching strategy, quantifies the impact of other nodes on link loss during cooperative transmission, and derives the correlation between various influencing factors through nonlinear fitting, thereby simplifying the computational complexity of link loss during multi-packet redundant transmission.

[0005] To achieve the above objectives, this application provides the following technical solution:

[0006] This application provides a data redundancy coding method based on deterministic network cooperative transmission, including the following steps:

[0007] Step 1: Calculate the node-to-node cooperative transmission distortion D. The distortion in node-to-node cooperative transmission mainly includes the distortion D generated during redundant coding. c Distortion D caused by the correlation between heterogeneous multimodal data corr Error distortion caused by packet loss DEC And the potential error distortion D caused by the loss of the reference block. r ,

[0008] Step 2: Single Data Redundancy Encoding. A single data packet is used for cooperative redundancy transmission. When a block of the main data packet is lost, if that block is encoded with a single data redundancy packet, the redundancy packet can be decoded to replace the lost block. The distortion is calculated, and it is observed whether it meets the distortion tolerance. If it does, single data redundancy encoding is used for cooperative transmission; otherwise, steps 3-6 are executed.

[0009] Step 3: Multiple data redundancy coding. Multiple data packets are used for coordinated redundant transmission. When a portion of the main data packet is lost, different parts of the multiple redundant data packets are simultaneously aggregated to recover the current block.

[0010] Step 4: Calculate the distortion D of the deterministic network system total By comparing the overall distortion cost J of the deterministic network system when using multi-data redundancy coding with different quantities and redundancy coding parameters, the quantity and redundancy coding parameters with the minimum distortion cost are selected for multi-data redundancy coding.

[0011] Step 5: Quantify the correlation between redundant coding parameters. The overall distortion of a deterministic network system is determined by the amount of redundancy and the redundant coding parameters. Select reasonable parameters to design a redundant system and quantify the correlation between redundant coding parameters through nonlinear fitting.

[0012] Step 6: Update the redundancy coding parameters. After completing the redundancy coding, select different sequence features. Based on the idea of ​​maximum correlation and minimum redundancy, select the sequence feature with the minimum redundancy, and then update the redundancy coding parameters based on this feature.

[0013] Step 7: Quantify the relationship between different data redundancy coding parameters, calculate the loss caused by different data redundancy coding, and determine the optimal number of data packets.

[0014] Step 8: Determine the optimal redundancy coding parameters. Based on the above steps, determine the optimal redundancy coding parameters and the number of data packets, and design a redundant transmission system accordingly.

[0015] The method for calculating the node-to-node cooperative transmission distortion D in step 1 is as follows:

[0016] Model the correlation coefficients between the data:

[0017]

[0018] In the formula Indicates the length of the data packet in bits. Indicates the priority of the data packet, and a and b are the PCA linear mapping coefficients. Let {x, y} represent data packets transmitted collaboratively on different links. {x, y} is the data block after PCA dimensionality reduction. Then, let's consider the interference S from other nodes during inter-node communication. ij Modeling:

[0019]

[0020] P t Let N be the transmission power of the sending node, N0 be the background noise, q be the path-loss exponent, and |Dis| be the distance between the sending and receiving nodes.

[0021] Then, the similarity coefficient between the two collaborating nodes is modeled:

[0022]

[0023] λ is used to adjust the similarity coefficient to the range of 0 to 1.

[0024] Finally, the distortion caused by the correlation of cooperative transmission is modeled as follows:

[0025]

[0026] In the formula D c The distortion caused by redundant encoding

[0027] Next, calculate the node-to-node cooperative transmission distortion D, as shown in the following formula:

[0028] D=(1-p)(D corr +D r )+p(D EC +D ECr )

[0029] In the formula D r It is a potential error distortion caused by the loss of the reference block, D EC It is the error distortion caused by packet loss, D ECr It is the potential error distortion caused by packet loss.

[0030] In step 4, the deterministic network system distortion D is calculated. total Specifically,

[0031] When multiple data packets are used for redundant encoding, the overall system distortion is as follows:

[0032]

[0033] In the formula, n is the number of data packets used.

[0034] The corresponding system distortion cost J at this time is as follows:

[0035]

[0036] In the formula D i The system distortion corresponding to the transmission of the i-th data packet is λ, where λ is the Lagrange daily number and R is the system distortion. i It is the total length of all packets when the i-th data packet is transmitted in a coordinated manner, and φ is the energy transfer function. The redundancy coding parameters are used when n redundant data packets are used to coordinate the transmission of the i-th data packet. Let be the code length of the i-th data packet. Let n be the code length of the redundant data packets.

[0037] The method for quantifying the relationship between different data packet redundancy coding parameters in step 7 is as follows:

[0038] First, different redundancy coding parameters are used to encode different feature sequences, and the redundancy coding parameters are obtained by observing the data. With quantization coefficient ΔQ i The following relationship exists:

[0039]

[0040] After redundant encoding, the feature sequence needs to be updated to quantize the code length of the i-th data packet. and the code length of n redundant data packets With quantization coefficient ΔQ i The relationship between them is as follows:

[0041]

[0042] In the formula, α and β are constants related to the characteristic sequence.

[0043] In step 6, the selection of sequence features with minimum redundancy based on the idea of ​​maximum correlation and minimum redundancy specifically involves:

[0044] First, we analyze the impact of each feature on the sequence evolution mode using information gain. Let the deterministic network contain N cooperative transmission paths between node pairs, denoted as A = {A1, A2, ... A}. N}, where the status of the i-th transmission path will be determined through the feature sequence. To express oneself Let represent its value in the j-th time unit, the total transmission delay of each path be T, and the final selected output sequence be S = {s1, s2, ..., sj}. T}, for feature A i In terms of its ability to reduce the uncertainty of output sequence prediction, g(S,A) i )for

[0045] g(S,Ai )=H(S)-H(S|A i )

[0046] When a certain feature is added as a reference, the entropy H(S|A) obtained from the partitioning is... i The smaller the value of g(S,A), the lower the level of uncertainty. i The larger the information gain, the greater the contribution of the added element. Therefore, the features are ranked according to their information gain, and the features with larger information gains are selected. Then, a correlation analysis is performed on the selected features, i.e., the Pearson correlation coefficient is calculated as follows:

[0047]

[0048] Where cov(A) i A j ) is A i and A j The covariance, σ(A) i ) is A i If the correlation coefficient between multiple features is large, it indicates that there is redundancy among these features. The feature with the highest information gain should be retained. Finally, the selected feature sequence has the greatest correlation with the output sequence, and its subset has the minimum redundancy.

[0049] Compared with the prior art, the beneficial effects of this application are:

[0050] 1. By combining PCA theory to model the correlation distortion between heterogeneous multimodal data and incorporating this part into the calculation of cooperative transmission distortion between node pairs, the algorithm comprehensively considers the impact of sensing other node pairs on the redundant transmission mechanism under cooperative transmission. This serves as the main basis for subsequent parameter tuning of the redundancy mechanism. It is an intelligent algorithm that considers the potential cognition and perception capabilities of deterministic networks.

[0051] 2. By continuously updating the redundancy coding parameters through a feature selection method based on the idea of ​​maximum correlation and minimum redundancy, multi-packet redundancy coding under cooperative transmission is achieved, which helps to simplify the computational complexity of link loss in multi-packet redundant transmission. The relationship between the number of packets, redundancy coding parameters, and quantization coefficients in multi-packet redundant transmission is derived through nonlinear fitting, and the total distortion cost under this condition is modeled. Optimal parameters are selected for the design of the multi-packet redundant transmission system, which helps to minimize link distortion caused by redundant transmission while ensuring high reliability. Attached Figure Description

[0052] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 This is an architecture diagram of this application;

[0054] Figure 2 This is a flowchart of the method in this application. Detailed Implementation

[0055] The technical solutions of the embodiments of this application will now be described with reference to the accompanying drawings. It should be noted that similar reference numerals and letters in the following drawings indicate similar items; therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0056] The terms “comprising,” “including,” or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase “comprising one…” does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0057] The following is combined Figures 1 to 2 The present invention specifically describes a data redundancy coding method based on deterministic network cooperative transmission, comprising the following specific steps:

[0058] Step 1: Calculate the node-to-node cooperative transmission distortion D. The distortion in node-to-node cooperative transmission mainly includes the distortion D generated during redundant coding. c Distortion D caused by the correlation between heterogeneous multimodal data corr Error distortion caused by packet loss D EC And the potential error distortion D caused by the loss of the reference block. r .

[0059] The method for calculating node-to-node cooperative transmission distortion is as follows:

[0060] First, when modeling collaborative transmission, the distortion caused by the correlation between heterogeneous multimodal data is considered. The following is the method for calculating this distortion:

[0061] Model the correlation coefficients between the data:

[0062]

[0063] In the formula Indicates the length of the data packet. Indicates the priority of the data packet, and a and b are the PCA linear mapping coefficients. This represents data packets transmitted collaboratively on different links, where {x,y} is a data block after PCA dimensionality reduction.

[0064] Then, regarding the interference S from other nodes during inter-node communication between collaborative nodes. ij Modeling:

[0065]

[0066] P t Let N be the transmission power of the sending node, N0 be the background noise, q be the path-loss exponent, and |Dis| be the distance between the sending and receiving nodes.

[0067] Then, the similarity coefficient between the two collaborating nodes is modeled:

[0068]

[0069] λ is used to adjust the similarity coefficient to the range of 0 to 1.

[0070] Finally, the distortion caused by the correlation of cooperative transmission is modeled as follows:

[0071]

[0072] In the formula D c This refers to the distortion that occurs during redundant encoding.

[0073] Next, calculate the node-to-node cooperative transmission distortion D, as shown in the following formula:

[0074] D=(1-p)(D corr +D r )+p(D EC +D ECr )

[0075] In the formula D r It is a potential error distortion caused by the loss of the reference block, D EC It is the error distortion caused by packet loss, D ECr It is the potential error distortion caused by packet loss.

[0076] Step 2: Single Data Redundancy Encoding. Cooperative redundancy transmission is achieved using a single data packet. When a block of the main data packet is lost, if that block is encoded with a single data redundancy packet, the redundancy packet can be decoded to replace the lost block. For example... Figure 1As shown in (a). Calculate the distortion at this point and observe whether it meets the distortion tolerance. If it does, use single data redundancy coding for cooperative transmission; if it does not, proceed to steps 3-6.

[0077] The distortion corresponding to single-data packet redundancy encoding is as follows:

[0078] D=(1-p)(D corr +D r )+p(1-p)(D dif +D r_r )+pp(D EC +D ECr )

[0079] In the formula, D dif For redundant coding parameters, D r_r Potential error distortion caused by data block loss.

[0080] Step 3: Multiple Data Redundancy Coding. Multiple data packets are used for coordinated redundant transmission. When a portion of the main data packet is lost, different parts of the multiple redundant data packets are simultaneously aggregated to recover the current block, such as... Figure 1 As shown in (b).

[0081] Step 4: Calculate the distortion D of the deterministic network system total By comparing the overall distortion cost J of the deterministic network system when using multi-data redundancy coding with different quantities and redundancy coding parameters, the quantity and redundancy coding parameters with the minimum distortion cost are selected for multi-data redundancy coding.

[0082] When multiple data packets are used for redundant encoding, the overall system distortion is as follows:

[0083]

[0084] In the formula, n is the number of data packets used. The redundancy coding parameters for the nth data packet. Potential error distortion caused by the loss of the nth data block.

[0085] The corresponding system distortion cost J at this time is as follows:

[0086]

[0087] In the formula D i The system distortion corresponding to the transmission of the i-th data packet is λ, where λ is the Lagrange daily number and R is the system distortion. i It is the total length of all packets during the cooperative transmission of the i-th data packet. φ is the energy transfer function. The redundancy coding parameters are used when n redundant data packets are used to coordinate the transmission of the i-th data packet. Let be the code length of the i-th data packet. Let n be the code length of the redundant data packets.

[0088] Step 5: Quantify the correlation between redundant coding parameters. As can be seen from Step 4, the overall distortion of a deterministic network system is determined by the amount of redundancy and the redundant coding parameters. Therefore, it is necessary to select reasonable parameters to design a redundant system and to quantify the correlation between redundant coding parameters through nonlinear fitting.

[0089] First, different redundancy coding parameters are used to encode different feature sequences, and the redundancy coding parameters are obtained by observing the data. With quantization coefficient ΔQ i The following relationship exists:

[0090]

[0091] After redundant encoding, the feature sequence needs to be updated to quantize the code length of the i-th data packet. and the code length of n redundant data packets With quantization coefficient ΔQ i The relationship between them is as follows:

[0092]

[0093] In the formula, α and β are constants related to the characteristic sequence.

[0094] Step 6: Update Redundancy Coding Parameters. After completing redundancy coding, different sequence features need to be selected. Based on the idea of ​​maximum correlation and minimum redundancy, the sequence feature with the minimum redundancy is selected, and then the redundancy coding parameters are updated based on this feature.

[0095] First, we analyze the impact of each feature on the sequence evolution mode using information gain. Let the deterministic network contain N cooperative transmission paths between node pairs, denoted as A = {A1, A2, ..., A...}. N The status of the i-th transmission path will be determined through a feature sequence. To express oneself Let represent its value in the j-th time unit, the total transmission delay of each path be T, and the final selected output sequence be S = {s1, s2, ..., sj}. T For feature A i In terms of its ability to reduce the uncertainty of output sequence prediction, g(S,A) i )for

[0096] g(S,A i )=H(S)-H(S|A i )

[0097] When a certain feature is added as a reference, the entropy H(S|A) obtained from the partitioning is... i The smaller the value of g(S,A), the lower the level of uncertainty. i The larger the information gain, the greater the contribution of that element. Therefore, features are ranked according to information gain, and features with larger information gains are selected. Then, correlation analysis is performed on the selected features, i.e., the Pearson correlation coefficient is calculated as follows:

[0098]

[0099] Where cov(A) i A j ) is A i and A j The covariance, σ(A) i ) is A i If the correlation coefficients among multiple features are large, it indicates redundancy among these features, and the feature with the highest information gain should be retained. Finally, the selected feature sequence has the highest correlation with the output sequence, and its subset has the lowest redundancy.

[0100] Step 7: Quantify the relationship between different data redundancy coding parameters and calculate the loss caused by different data redundancy coding. For multiple data packet redundancy modes, different coding parameters need to be used for data blocks at different locations. At this time, it is necessary to quantify the relationship between different data packets and redundancy coding parameters. As time evolves, the system distortion (loss) caused by different data packets after redundancy coding is also different. Therefore, it is necessary to quantify the loss caused by redundancy coding of different data packets to determine the optimal number of data packets.

[0101] The correlation between the quantization coefficients of different redundant packets is as follows:

[0102] ΔQ i+1 ≈ΔQ i (p -1 / 2 -1)

[0103] In the formula, i is the redundant packet number.

[0104] When using redundant transmission of n data packets, the difference in distortion cost caused by the nth data packet is:

[0105]

[0106] When ΔJ > 0, the nth redundant frame can be encoded. A larger n indicates that more redundant packets are needed. Sequences with complex structures that evolve rapidly and are more prone to packet loss require more redundant packets. The initial range M for the final determination of the number of redundant packets is then established.

[0107] Step 8: Determine the optimal redundancy coding parameters. Based on the steps above, determine the optimal redundancy coding parameters and the number of data packets, and design a redundant transmission system accordingly.

[0108] Considering the distortion costs when using [M-1,M] redundant packets and different encoding parameters, the encoding parameters and the number of redundant packets with the minimum distortion cost are finally selected as the final confirmed encoding parameters for encoding the current data packet.

[0109] Compared to existing technologies, this invention proposes a data redundancy coding method based on deterministic network cooperative transmission. This invention models the similarity between heterogeneous multimodal data and calculates the distortion generated by cooperative transmission. Combining packet loss rate and redundancy coding distortion, a method for calculating the cooperative transmission distortion between node pairs is designed. According to this method, the distortion under single-packet redundancy is calculated to see if it meets the requirements. If the single-packet redundancy does not meet the requirements, multi-packet redundancy cooperative transmission is adopted. Then, the relationship between the number of packets, coding parameters, and quantization coefficients is modeled through nonlinear fitting, and the parameter set corresponding to the minimum system distortion cost is selected to design the redundancy coding mechanism. Finally, after the redundant transmission of a certain packet ends, a feature selection method is designed based on the idea of ​​maximum correlation and minimum redundancy combined with information gain. The feature sequence with the highest information gain is retained for updating the coding parameters, which are then updated for the redundant transmission of the next data packet.

[0110] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A data redundancy coding method based on deterministic network cooperative transmission, characterized in that, Includes the following steps: Step 1: Calculate node-to-node cooperative transmission distortion Distortion during node-to-node collaborative transmission mainly includes distortion generated during redundant coding. Distortion caused by the correlation between heterogeneous multimodal data Error distortion caused by packet loss And the potential error distortion caused by the loss of reference blocks. , Step 2: Single Data Redundancy Encoding. A single data packet is used for cooperative redundancy transmission. When a block of the main data packet is lost, if the lost block is encoded with a single data redundancy packet, the redundancy packet can be decoded to replace the lost block. The distortion is calculated, and it is observed whether it meets the distortion tolerance. If it does, single data redundancy encoding is used for cooperative transmission; otherwise, steps 3-6 are executed. Step 3: Multiple data redundancy coding. Multiple data packets are used for coordinated redundant transmission. When a portion of the main data packet is lost, different parts of the multiple redundant data packets are simultaneously aggregated to recover the current block. Step 4: Calculate the distortion of the deterministic network system By comparing the overall distortion cost of deterministic network systems when using multi-data redundancy coding with different amounts and redundancy coding parameters, this study investigates the impact of different redundancy coding methods. Choose the number and redundancy coding parameters with the minimum distortion cost for multi-data redundancy coding. Step 5: Quantify the correlation between redundant coding parameters. The overall distortion of a deterministic network system is determined by the amount of redundancy and the redundant coding parameters. Select reasonable parameters to design a redundant system and quantify the correlation between redundant coding parameters through nonlinear fitting. Step 6: Update the redundancy coding parameters. After completing the redundancy coding, select different sequence features. Based on the idea of ​​maximum correlation and minimum redundancy, select the sequence feature with the minimum redundancy, and then update the redundancy coding parameters based on this feature. Step 7: Quantify the relationship between different data redundancy coding parameters, calculate the loss caused by different data redundancy coding, and determine the optimal number of data packets. Step 8: Determine the optimal redundancy coding parameters. Based on the above steps, determine the optimal redundancy coding parameters and the number of data packets, and design a redundant transmission system accordingly.

2. The data redundancy coding method based on deterministic network cooperative transmission according to claim 1, characterized in that, Step 1 calculates node-to-node cooperative transmission distortion. The calculation method is as follows: Model the correlation coefficients between the data: ， In the formula Indicates the length of the data packet in bits. Indicates the priority of the data packets. For PCA linear mapping coefficients, This refers to data packets that are transmitted collaboratively on different links. It is the data block after PCA dimensionality reduction, and then it addresses the interference from other nodes during inter-node communication in the collaborative process. Modeling: ， The transmission power of the sending node. The background noise is represented by q, and the path loss exponent is represented by q. The distance between the sending node and the receiving node. Then, the similarity coefficient between the two collaborating nodes is modeled: ， This is used to adjust the similarity coefficient to the range of 0 to 1. Finally, the distortion caused by the correlation of cooperative transmission is modeled as follows: ， In the formula The distortion caused by redundant encoding Secondly, calculate the node-to-node collaborative transmission distortion. As shown in the following formula: ， In the formula This is a potential error distortion caused by the loss of the reference block. It is an error distortion caused by packet loss. It is the potential error distortion caused by packet loss.

3. The data redundancy coding method based on deterministic network cooperative transmission according to claim 1, characterized in that, Step 4 involves calculating the distortion of the deterministic network system. Specifically, When multiple data packets are used for redundant encoding, the overall system distortion is as follows: ， In the formula The number of data packets used. The corresponding system distortion cost at this time as follows: ， In the formula It is the transmission of the first The system distortion corresponding to each data packet For Lagrange daily numbers, It is the first The total length of all packets when data packets are transmitted collaboratively. Let be the energy transfer function. For use The redundant data packet coordinates the first Redundant coding parameters during data packet transmission For the first The code length of each data packet. for The code length of a redundant data packet.

4. The data redundancy coding method based on deterministic network cooperative transmission according to claim 1, characterized in that... include: The method for quantifying the relationship between different data packet redundancy coding parameters in step 7 is as follows: First, different redundancy coding parameters are used to encode different feature sequences, and the redundancy coding parameters are obtained by observing the data. With quantization coefficient The following relationship exists: ， After redundant encoding, the feature sequence needs to be updated, and the first quantization step needs to be performed. With quantization coefficient The relationship between them is as follows: ， In the formula and It is a constant related to the characteristic sequence.

5. A data redundancy coding method based on deterministic network cooperative transmission according to claim 1, characterized in that... include: In step 6, the selection of sequence features with minimum redundancy based on the idea of ​​maximum correlation and minimum redundancy specifically involves: First, we analyze the impact of each feature on the sequence evolution mode using information gain, assuming that the deterministic network contains... The set of collaborative transmission paths between nodes is denoted as . , among which, the The status of each transmission path will be determined through characteristic sequences. To express oneself, Indicates that it is in the first The values ​​within each time unit, the overall transmission delay for each path is The final selected output sequence is For features In terms of its ability to reduce the uncertainty of output sequence prediction for ， When a certain feature is added as a reference, the entropy obtained from the partitioning is... The smaller the value, the better the ability to reduce uncertainty. The larger the information gain, the greater the contribution of a particular feature. Therefore, features are ranked according to information gain, and features with larger information gains are selected. Then, correlation analysis is performed on the selected features, i.e., the Pearson correlation coefficient is calculated as follows: ， in yes and covariance, for If the correlation coefficient between multiple features is large, it indicates that there is redundancy among these features. The feature with the highest information gain should be retained. Finally, the selected feature sequence has the greatest correlation with the output sequence, and its subset has the minimum redundancy.

Images