Carbon transaction electricity market simulation prediction system based on big data analysis

Abstract

The invention discloses a carbon transaction electricity market simulation prediction system based on big data analysis, which belongs to the technical field of carbon transaction and comprises a collaborative analysis module, a model module, an acquisition module, a simulation analysis module and a user combination analysis module. The collaborative analysis module is used for performing collaborative analysis and generating a collaborative prediction map; the model module is used for establishing a simulation model and obtaining a reference model, a platform side simulates and sets optimization materials, and the optimization materials comprise all collaborative prediction maps of simulation setting and corresponding prediction collection data and prediction real results; performing optimization adjustment on the reference model through the optimization material to obtain a simulation model; the acquisition module is used for acquiring prediction acquisition data; the simulation analysis module is used for performing simulation analysis and acquiring prediction acquisition data and a collaborative prediction map in real time; the prediction collection data and the collaborative prediction map are analyzed through a simulation model, and a comprehensive simulation result is obtained; and the user combination analysis module is used for carrying out analogue simulation analysis according to user requirements.

Description

Technical Field

[0001] The present invention belongs to the technical field of carbon trading, and in particular is a carbon trading electricity market simulation prediction system based on big data analysis. Background Art

[0002] In the current field of carbon trading power market simulation and prediction, although a variety of simulation models have been developed and applied to the prediction of market trends, these models generally have a significant limitation: they often ignore the importance of human experience and knowledge in the prediction process. Most of the existing simulation models are built based on historical data and mathematical algorithms. Although they can reflect the operating laws of the market to a certain extent, their prediction accuracy is often not ideal due to the lack of integration and utilization of human experience.

[0003] Power companies, carbon trading participants, policy makers and other market parties have a deep understanding and unique insights into various factors affecting the market based on their deep industry background and rich practical experience. They can make relatively accurate predictions of the future market trends based on their own experience and knowledge, especially for short-term predictions. However, the existing simulation prediction system does not make full use of this valuable prediction information, resulting in the prediction results may deviate from the actual market situation.

[0004] In addition, as a market participant, each user also has his or her own unique experience and judgment. However, current simulation models can usually only provide standardized prediction results and cannot be flexibly adjusted according to the user's personalized needs. This limits the user's ability to verify and optimize the prediction results based on their own experience, and also reduces the practicality of the simulation model and user satisfaction.

[0005] Based on this, in order to solve the above problems, the present invention provides a carbon trading electricity market simulation and prediction system based on big data analysis. Summary of the invention

[0006] In order to solve the problems existing in the above solutions, the present invention provides a carbon trading electricity market simulation and prediction system based on big data analysis.

[0007] The purpose of the present invention can be achieved through the following technical solutions:

[0008] A carbon trading power market simulation prediction system based on big data analysis, including collaborative analysis module, model module, acquisition module, simulation analysis module and user combination analysis module;

[0009] The collaborative analysis module is used to perform collaborative analysis. The platform sets the collaborative scope and identifies the collaborative parties within the collaborative scope; sets the influencing factors that have an impact on carbon trading forecasts; and sets corresponding influencing factor labels and factor coefficients corresponding to the influencing factor labels for the collaborative parties according to the influencing factors.

[0010] Acquire the prediction data of each collaborative party within the collaborative scope in real time, classify the prediction data according to the influencing factor labels corresponding to the collaborative parties, and obtain the factor prediction data corresponding to each influencing factor label;

[0011] The prediction data of each factor are summarized according to the influencing factor labels corresponding to each influencing factor to form a factor prediction data set corresponding to each influencing factor; the factor prediction data set of each influencing factor is analyzed to generate a collaborative prediction map.

[0012] Furthermore, the method for setting the factor coefficient of each influencing factor label includes:

[0013] The platform sets the initial coefficient and adjustment interval for the influencing factor label of each collaborative party; the initial coefficient and adjustment interval are marked as δ0 and [v, k] respectively;

[0014] Obtain the prediction verification data of each influencing factor label from the corresponding collaborating party in real time, wherein the prediction verification data includes the factor prediction data and the actual result, and calculate the prediction accuracy of each influencing factor label from the collaborating party in real time based on the prediction verification data;

[0015] Calculate the factor coefficient of the corresponding influencing factor label according to the formula δ=δ0+(k-v+1)σ+v-1;

[0016] Where: δ is the factor coefficient of the corresponding influencing factor label; δ0 is the initial coefficient; σ is the prediction accuracy of the corresponding influencing factor label; v and k are the two interval endpoints of the adjustment interval.

[0017] Furthermore, the method for analyzing the factor prediction data set of each influencing factor includes:

[0018] Identify each factor prediction data in the factor prediction data set corresponding to the influencing factor, identify the prediction result corresponding to each factor prediction data, merge each prediction result to obtain each prediction merge class;

[0019] Identify the factor coefficients of each prediction result in the prediction merging class; mark each prediction result in the prediction merging class as i, i=1, 2, ..., n, n is a positive integer; mark the factor coefficient of each prediction result as δ i ;

[0020] According to the formula Calculate the classification value of each predicted merge class;

[0021] Where: PA is the classification value of the corresponding predicted merged class; e is a natural constant; δ i is the factor coefficient of the corresponding prediction result;

[0022] Calculate the classification proportion of each predicted combined class in the corresponding influencing factor according to the classification value of each predicted combined class;

[0023] A collaborative prediction map is generated based on each influencing factor and each prediction merged class and classification proportion corresponding to each influencing factor.

[0024] Furthermore, the method for merging the prediction results includes:

[0025] Step SA1: establishing an impact difference evaluation model, wherein the impact difference evaluation model is used to evaluate the impact difference value between each prediction result;

[0026] Step SA2: identifying each prediction result, analyzing each prediction result through the impact difference evaluation model, obtaining the impact difference value between each prediction result; sorting each impact difference value in ascending order, obtaining a difference sequence;

[0027] Step SA3: The two prediction results corresponding to the influence difference values ​​that are lower than the threshold value X1 and ranked first in the difference sequence are classified into one category, which is marked as the initial merged category;

[0028] Step SA4: Identify the prediction result with the smallest impact difference value between the prediction results in the initial merged class, and when the impact difference value is lower than the threshold value X1, assign the prediction result to the initial merged class;

[0029] Step SA5: looping step SA4 until no impact difference value is lower than the threshold value X1, marking the initial merged class as a predicted merged class, and removing the impact difference value corresponding to each prediction result in the predicted merged class from the difference sequence to obtain a new difference sequence;

[0030] Return to step SA3 until there is no influencing difference value in the difference sequence.

[0031] Furthermore, the method for establishing the impact difference assessment model in step SA1 includes:

[0032] Setting the influencing material data of each influencing factor based on the historical carbon trading data; performing an impact assessment based on each of the influencing material data to obtain the impact difference value between each of the influencing material data;

[0033] An impact difference evaluation model is established according to the impact material data and the impact difference values ​​between the impact material data.

[0034] The model module is used to establish a simulation model and obtain a benchmark model. The platform simulates and sets the optimization material, which includes the collaborative prediction maps of the simulation setting and the corresponding prediction collection data and prediction real results. The benchmark model is optimized and adjusted through the optimization material to obtain a simulation model.

[0035] The acquisition module is used to perform real-time data acquisition and obtain predicted acquisition data.

[0036] The simulation analysis module is used to perform simulation analysis, obtain prediction collection data and collaborative prediction maps in real time; analyze the prediction collection data and collaborative prediction maps through the simulation model to obtain comprehensive simulation results; and display the comprehensive simulation results to users.

[0037] The user combination analysis module is used to perform simulation analysis according to user needs, obtain the collaborative prediction map corresponding to the comprehensive simulation results in real time, and synchronously display the collaborative prediction map to the user;

[0038] The user sets corresponding combined analysis requirements according to the displayed comprehensive simulation results and collaborative prediction map; simulates and adjusts the collaborative prediction map according to the combined analysis requirements to obtain a collaborative simulation map;

[0039] The corresponding prediction collection data is obtained, and the prediction collection data and the collaborative simulation map are analyzed through the simulation model to obtain the user simulation result; and the user simulation result is displayed to the user.

[0040] Furthermore, when the user simulation result does not meet the user's requirements, the user adjusts the combined analysis requirements and re-simulates the analysis until the user simulation result meets the user's requirements.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] By combining the forecast information of various market parties such as power companies, carbon trading participants, and policy makers, as well as the personalized adjustments of each user based on their own experience, the prediction accuracy and reliability of simulation can be significantly improved, which can better reflect the real situation and potential trends of the market than relying solely on model predictions; allowing users to arbitrarily combine and adjust the influencing factors in the simulation model according to their own experience and needs; this user-defined function greatly enhances the flexibility and practicality of the simulation model, allowing the model to adapt to the needs of different users and different scenarios; by providing two simulation modes, it provides users with the possibility of analyzing the market from different angles; users can choose to ignore certain influencing factors as needed, or weight the influencing factors to obtain prediction results that better meet their own needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be 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.

[0044] Figure 1 This is a principle block diagram of the present invention. DETAILED DESCRIPTION

[0045] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] like Figure 1 As shown, the carbon trading power market simulation prediction system based on big data analysis includes a collaborative analysis module, a model module, a collection module, a simulation analysis module and a user combination analysis module;

[0047] The collaborative analysis module is used to analyze the forecast data of various influencing factors predicted by various market parties such as power companies, carbon trading participants and policy makers; the power companies, carbon trading participants and policy makers are regarded as collaborative parties; the platform sets the collaborative scope, that is, which collaborative parties' forecast data will be collected in the future. Generally, the platform needs to negotiate with each market party to determine, so as to avoid the uncertainty of directly obtaining the forecast data of each market party from the Internet and other channels, and there can be multiple collaborative parties of the same type; various influencing factors that have an impact on carbon trading forecasts, such as policy adjustments, technological progress, market demand, etc. are set; according to the actual situation of each collaborative party within the collaborative scope, a corresponding influencing factor label is set for each collaborative party, indicating that the collaborative party has predictive representativeness for the influencing factors corresponding to each marked influencing factor label; and the platform can set a corresponding factor coefficient for each collaborative party's prediction of each influencing factor label, or can intelligently set a corresponding factor coefficient based on an existing analysis algorithm. The factor coefficient is used to indicate the accuracy of its prediction of the influencing factor. The larger the factor coefficient, the higher the prediction accuracy;

[0048] Obtain the forecast data of each collaborative party within the collaborative scope in real time, classify the forecast data according to the influencing factor labels corresponding to the collaborative parties, and obtain the factor forecast data corresponding to each influencing factor label;

[0049] The prediction data of each factor are summarized according to the labels corresponding to each influencing factor to form a factor prediction data set corresponding to each influencing factor; the factor prediction data set of each influencing factor is analyzed to generate a collaborative prediction map for the prediction of each influencing factor. The collaborative prediction map integrates the prediction data sets of each factor to understand the prediction results of each influencing factor and the proportion of each prediction result.

[0050] In one embodiment, a method for setting factor coefficients of each influencing factor label of each collaborative party includes:

[0051] The platform sets an initial coefficient and adjustment interval for each collaborator based on its understanding of each collaborator. The initial coefficient is a factor coefficient that is set in advance without considering subsequent predictions. The adjustment interval is preset by the platform, regardless of subsequent predictions. Its impact on the initial coefficient can only change within this interval. For example, it is generally set to [-0.2, 0.5], which means that the initial coefficient can only be reduced by 0.2 at the lowest and increased by 0.5 at the highest. It is mainly set based on the platform's recognition of the collaborator. For example, if its ability and status are more important in this direction, its initial coefficient is higher, the lower limit of the interval is smaller, and the upper limit of the interval is higher. The initial coefficient and adjustment interval are marked as δ0 and [v, k] respectively.

[0052] Obtain the prediction verification data of the influencing factors corresponding to each influencing factor label from the corresponding collaborating party in real time. The prediction verification data includes the factor prediction data and the actual results. The prediction accuracy of each influencing factor label by the collaborating party is counted in real time based on the prediction verification data.

[0053] Calculate the factor coefficient according to the formula δ=δ0+(k-v+1)σ+v-1;

[0054] Where: δ is the factor coefficient of the corresponding influencing factor label; δ0 is the initial coefficient; σ is the prediction accuracy of the corresponding influencing factor label; v and k are the two interval endpoints of the adjustment interval.

[0055] In one embodiment, the factor prediction data set of each influencing factor is analyzed based on existing technologies.

[0056] In one implementation, a method for analyzing a factor prediction data set of each influencing factor includes:

[0057] Identify the prediction data of each factor in the factor prediction data set corresponding to the influencing factor, identify the prediction results corresponding to each factor prediction data, merge each prediction result to obtain each prediction merge class, and identify the factor coefficient corresponding to each prediction result in the prediction merge class; mark each prediction result in the prediction merge class as i, i = 1, 2, ..., n, n is a positive integer; mark the factor coefficient corresponding to each prediction result as δi ;

[0058] According to the formula Calculate the classification value corresponding to each predicted merge class;

[0059] Where: PA is the classification value of the corresponding predicted merged class; e is a natural constant; δ i is the factor coefficient of the corresponding prediction result;

[0060] The classification proportion of each prediction combined class in the corresponding influencing factor is calculated according to the classification value of each prediction combined class; that is, the proportion of the classification value of the corresponding prediction combined class to the sum of the classification values ​​of each prediction combined class corresponding to the influencing factor;

[0061] A collaborative prediction map is generated based on various influencing factors and the prediction merging categories and classification proportions corresponding to each influencing factor; it can be generated based on a fan chart, etc.

[0062] In one embodiment, the prediction results are merged based on existing technologies, such as merging according to the similarity between the prediction results. The platform may also preset the range of variation results of each influencing factor and then directly perform matching and classification.

[0063] In one embodiment, the prediction results are combined as follows:

[0064] Step SA1: Based on the historical carbon trading data and various influencing factors, various influencing material data of each influencing factor are determined, which are set according to the prediction results of the prediction data of the collaborative party, so as to realize the reverse deduction of the corresponding influencing material data after the actual results are clarified; impact assessment is performed according to each influencing material data, and the impact difference value between each influencing material data is determined. The impact difference value is set according to the difference in the impact of the two on the simulation results. If the two influencing material data have the same impact on the simulation results, the impact difference value is 0, which is determined according to the actual historical carbon trading data and combined with the corresponding control variable method; an impact difference assessment model is established according to each influencing material data and the impact difference value between each influencing material data. The impact difference assessment model is used for subsequent impact difference analysis between each prediction result and evaluation of the impact difference values ​​between each other; specifically, the impact difference assessment model can be established in combination with existing intelligent technology, modeling technology, etc. For example, when comprehensive influencing material data is available, it can be established in combination with a matching algorithm, and then matching analysis can be performed; it can also be established based on neural networks such as CNN networks or DNN networks, and a training set is established based on the above data for training.

[0065] Step SA2: Identify each prediction result, analyze each prediction result through the impact difference evaluation model, and obtain the impact difference value between each prediction result; sort each impact difference value in ascending order to obtain a difference sequence;

[0066] Step SA3: The prediction results corresponding to the influence difference values ​​whose influence difference values ​​are lower than the threshold value X1 and ranked first in the difference sequence are classified into one category, which is marked as the initial merged category;

[0067] Step SA4: Identify the prediction result with the smallest impact difference value between the prediction results in the initial merged class. When the corresponding impact difference value is lower than the threshold value X1, assign the corresponding prediction result to the initial merged class and proceed to step SA5; when the corresponding impact difference value is not lower than the threshold value X1, proceed to step SA5;

[0068] Step SA5: Loop step SA4 until no impact difference value is lower than the threshold X1, mark the initial merge class as the predicted merge class, and remove the impact difference value corresponding to each prediction result in the predicted merge class from the difference sequence to obtain a new difference sequence, and return to step SA3 until there is no impact difference value in the difference sequence.

[0069] The model module is used to establish a simulation model for simulating and predicting the carbon trading electricity market and to set a benchmark model. The benchmark model can be established by the platform based on existing simulation model establishment technology, such as based on a machine learning algorithm; or an existing simulation model that can be learned and adjusted can be directly selected as the benchmark model.

[0070] Based on the above-mentioned method for generating collaborative prediction maps, simulation settings are optimized materials, the optimized materials are the collaborative prediction maps of the simulation settings and the corresponding prediction collection data and prediction real results, the prediction collection data are the input data of the useful simulation analysis of the corresponding simulation settings, and the input data of the benchmark model can be set for setting; the prediction real result is the real carbon trading result corresponding to the prediction collection data; the benchmark model is optimized and adjusted by optimizing the materials until it meets the requirements; the benchmark model is marked as a simulation model.

[0071] In other embodiments, the simulation model may be directly established once by the platform, eliminating the optimization step.

[0072] The acquisition module is used to collect data, specifically to collect data according to the simulation requirements of the original benchmark model, and to collect data using the supporting existing technology to obtain predicted collection data.

[0073] The simulation analysis module is used to perform simulation analysis and obtain prediction collection data and collaborative prediction maps in real time; the prediction collection data and collaborative prediction maps are analyzed through the simulation model to obtain corresponding simulation analysis results. In order to distinguish them from subsequent simulation results, the above simulation analysis results are marked as comprehensive simulation results; the obtained comprehensive simulation results are displayed to the user.

[0074] The user combination analysis module is used to perform targeted simulation analysis according to the user's personalized prediction needs, obtain the collaborative prediction map corresponding to the comprehensive simulation results in real time, and synchronously display the collaborative prediction map to the user;

[0075] The user determines his / her combined analysis needs based on his / her own experience and collaborative forecasting map. If he / she can ignore the influencing factors of corresponding policies, technologies, etc. in the short term, he / she can set the corresponding combined analysis needs and remove the corresponding influencing factors. The collaborative forecasting map is simulated and adjusted according to the combined analysis needs, that is, the collaborative forecasting map is adjusted according to the adjustment of each influencing factor in the combined analysis needs, and the adjusted collaborative forecasting map is marked as a collaborative simulation map.

[0076] Obtain the corresponding prediction collection data, analyze the prediction collection data and the collaborative simulation map through the simulation model, and obtain the user simulation results; display the user simulation results to the user; subsequent users can also continue to set combined analysis requirements for simulation analysis, and the simulation results in the process can be saved for subsequent users to review.

[0077] The above formulas are all calculated by removing dimensions and taking numerical values. The formula is a formula that is closest to the actual situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained by simulating a large amount of data.

[0078] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. Carbon trading power market simulation prediction system based on big data analysis, characterized by: It includes collaborative analysis module, model module, acquisition module, simulation analysis module and user combination analysis module; The collaborative analysis module is used to perform collaborative analysis. The platform sets the collaborative scope and identifies the collaborative parties within the collaborative scope; sets the influencing factors that have an impact on carbon trading forecasts; and sets corresponding influencing factor labels and factor coefficients corresponding to the influencing factor labels for the collaborative parties according to the influencing factors. Acquire the prediction data of each collaborative party within the collaborative scope in real time, classify the prediction data according to the influencing factor labels corresponding to the collaborative parties, and obtain the factor prediction data corresponding to each influencing factor label; Summarize the prediction data of each factor according to the influencing factor labels corresponding to each influencing factor to form a factor prediction data set corresponding to each influencing factor; analyze the factor prediction data set of each influencing factor to generate a collaborative prediction map; The model module is used to establish a simulation model, obtain a benchmark model, and simulate and set optimization materials by the platform. The optimization materials include various collaborative prediction graphs set by the simulation, as well as corresponding prediction collection data and prediction real results; The benchmark model is optimized and adjusted by using the optimization material to obtain a simulation model; The acquisition module is used to perform real-time data acquisition and obtain predicted acquisition data; The simulation analysis module is used to perform simulation analysis and obtain prediction collection data and collaborative prediction maps in real time; Analyze the prediction collection data and the collaborative prediction map through the simulation model to obtain a comprehensive simulation result; Displaying the comprehensive simulation results to a user; The user combination analysis module is used to perform simulation analysis according to user needs, obtain the collaborative prediction map corresponding to the comprehensive simulation results in real time, and synchronously display the collaborative prediction map to the user; The user sets corresponding combined analysis requirements according to the displayed comprehensive simulation results and collaborative prediction map; simulates and adjusts the collaborative prediction map according to the combined analysis requirements to obtain a collaborative simulation map; Obtain corresponding prediction collection data, analyze the prediction collection data and the collaborative simulation map through the simulation model, and obtain user simulation results; Display the user simulation results to the user.

2. The carbon trading electricity market simulation prediction system based on big data analysis according to claim 1 is characterized in that: The method for setting the factor coefficient of each influencing factor label includes: The platform sets the initial coefficient and adjustment interval for the influencing factor label of each collaborative party; the initial coefficient and adjustment interval are marked as δ0 and [v, k] respectively; Obtain the prediction verification data of each influencing factor label from the corresponding collaborating party in real time, wherein the prediction verification data includes the factor prediction data and the actual result, and calculate the prediction accuracy of each influencing factor label from the collaborating party in real time based on the prediction verification data; According to the formula δ=δ0+(k-v+1) σ +v-1 calculates the factor coefficient of the corresponding influencing factor label; Where: δ is the factor coefficient of the corresponding influencing factor label; δ0 is the initial coefficient; σ is the prediction accuracy of the corresponding influencing factor label; v and k are the two interval endpoints of the adjustment interval.

3. The carbon trading power market simulation prediction system based on big data analysis according to claim 2 is characterized in that: Methods for analyzing the factor prediction data set of each influencing factor include: Identify each factor prediction data in the factor prediction data set corresponding to the influencing factor, identify the prediction result corresponding to each of the factor prediction data, merge the prediction results to obtain each prediction merge class; calculate the classification value of each prediction merge class; Calculate the classification proportion of each predicted combined class in the corresponding influencing factor according to the classification value of each predicted combined class; A collaborative prediction map is generated based on each influencing factor and each prediction merged class and classification proportion corresponding to each influencing factor.

4. The carbon trading electricity market simulation prediction system based on big data analysis according to claim 3 is characterized in that: Methods for combining prediction results include: Step SA1: establishing an impact difference evaluation model, wherein the impact difference evaluation model is used to evaluate the impact difference value between each prediction result; Step SA2: identifying each prediction result, analyzing each prediction result through the impact difference evaluation model, obtaining the impact difference value between each prediction result; sorting each impact difference value in ascending order, obtaining a difference sequence; Step SA3: The two prediction results corresponding to the influence difference values ​​that are lower than the threshold value X1 and ranked first in the difference sequence are classified into one category, which is marked as the initial merged category; Step SA4: Identify the prediction result with the smallest impact difference value between the prediction results in the initial merged class, and when the impact difference value is lower than the threshold value X1, assign the prediction result to the initial merged class; Step SA5: looping step SA4 until no impact difference value is lower than the threshold value X1, marking the initial merged class as a predicted merged class, and removing the impact difference value corresponding to each prediction result in the predicted merged class from the difference sequence to obtain a new difference sequence; Return to step SA3 until there is no influencing difference value in the difference sequence.

5. The carbon trading power market simulation prediction system based on big data analysis according to claim 4 is characterized in that: The method for establishing the impact difference assessment model in step SA1 includes: Setting the influencing material data of each influencing factor based on the historical carbon trading data; performing an impact assessment based on each of the influencing material data to obtain the impact difference value between each of the influencing material data; An impact difference evaluation model is established according to the impact material data and the impact difference values ​​between the impact material data.

6. The carbon trading electricity market simulation prediction system based on big data analysis according to claim 1 is characterized in that: When the user simulation result does not meet the user's requirements, the user adjusts the combined analysis requirements and re-simulates the analysis until the user simulation result meets the user's requirements.

7. The carbon trading electricity market simulation prediction system based on big data analysis according to claim 3 is characterized in that: Methods for calculating classification values ​​for each predicted merge class include: Identify the factor coefficients of each prediction result in the prediction merging class; mark each prediction result in the prediction merging class as i, i=1, 2, ..., n, n is a positive integer; mark the factor coefficient of each prediction result as δ i ; According to the formula Calculate the classification value of each predicted merge class; Where: PA is the classification value of the corresponding predicted merged class; e is a natural constant; δ i is the factor coefficient of the corresponding prediction result.

Images