A method and system for analyzing polling organization effects based on contrast rectification and quantitative evaluation

Abstract

The application discloses a kind of based on contrast rectification and quantitative evaluation's polling organization effect analysis method and its system, method includes: from multiple channels obtains each state election voting history data and each polling organization released polling data, the data of different sources are handled;Based on the polling organization rating framework constructed, the division of polling organization grade is carried out;According to the polling data, polling index value is calculated, and polling index system is constructed;By comparing and the same polling organization, multidimensional comparative analysis in multiple polling organizations, the deviation of the polling data is corrected;Comprehensive quantitative evaluation constructs organization rating model, and the final score and rating of each polling organization are calculated according to the polling organization grade.The method of the application plays a key role in the prediction of election results, realizes data-driven efficient modeling.

Description

Technical Field

[0001] This application of the present invention belongs to the field of data analysis and prediction technology, specifically, it relates to a method and system for analyzing the effects of polling organizations based on comparative correction and quantitative evaluation. Background Technology

[0002] Institutional effect rating is an important method for systematically evaluating the predictive effectiveness of polling organizations and their potential partisan bias. It quantifies the deviation between polling data released by an organization and historical election results, comprehensively calculating the error value and directionality of its predictions (e.g., bias towards the Democratic or Republican parties), thereby dynamically generating a rating result to objectively reflect the organization's predictive credibility, long-term stability, and potential bias characteristics. Based on the rating results, high-precision polling organizations can be selected, providing high-confidence data sources for election research, public opinion modeling, and prediction algorithm optimization, significantly improving the scientific rigor and interpretability of prediction results. Simultaneously, the rating conclusions can provide cross-cycle, multi-dimensional references for policymakers in governments, media, and research institutions, effectively enhancing the objectivity and foresight of judgments on major events in areas such as election prediction, public policy formulation, and social sentiment analysis.

[0003] Institutional effect rating faces numerous problems and challenges:

[0004] (1) The systematic and comprehensive nature of evaluation criteria. Different polling organizations have different survey methods, sample selections and preferences, and their prediction results often have certain biases. How to define a scientific and comprehensive evaluation criterion that can simultaneously reflect prediction accuracy, long-term stability and partisan bias, and how to combine other indicators (such as the number of surveys, sample size, data coverage, etc.) on the basis of prediction bias.

[0005] (2) Long-term prediction and stability issues of the model. Some institutions may perform well in a particular year, but their long-term performance is unstable, requiring the integration of data from multiple years and the introduction of weights to balance short-term and long-term performance. In addition, since the shift in partisan support involves many dynamic factors, traditional statistical prediction models are unable to capture and cope with such complex time-series changes.

[0006] (3) The authenticity and credibility of the data. Poll data may be inaccurate or incomplete due to institutional operation or sample bias. When rating institutional effects, it is necessary to consider whether the data has been certified by an authoritative body, whether the data collection process is open and transparent, and whether the sample is sufficiently representative (such as coverage of different races, ages, and geographical regions).

[0007] (4) Data scarcity and insufficient sample size. The limited number of polling organizations or the limited publicly available polling data in certain states or years may lead to insufficient samples, thus affecting the representativeness of the rating results. Therefore, when modeling ratings, it is necessary to avoid the performance of a small number of organizations having an excessive impact on the overall rating.

[0008] (5) The problem of interference from some external factors. The trend of party support is affected by many external factors, such as unforeseen events, policy changes, international relations, etc. These factors may have unexpected effects on the election results, increasing the uncertainty and risk of prediction.

[0009] Currently, common methods for rating the effectiveness of polling organizations involve analyzing data from specific years or regions, using simple statistical models to compare the errors between the poll data released by the organizations and the actual election results, thereby evaluating the poll results. This method has certain drawbacks: 1) Limited corrective capabilities: Evaluating organizations based on data or events within a specific scope does not fully utilize historical poll data from multiple years and regions, making it difficult to comprehensively and effectively assess the long-term performance and systematic biases of each organization; 2) Limited evaluation dimensions: Simple statistical models rely solely on a single indicator of prediction error for rating, ignoring factors such as historical performance, regional differences, and data transparency. This makes it difficult to quantify and compare the rating results, and the accuracy and objectivity of the results are also difficult to guarantee.

[0010] In summary, this application, in its work on rating the effects of polling organizations in election scenario prediction research, found that the core deficiency of existing technologies lies in the lack of a systematic corrective framework and dynamic quantitative evaluation mechanism. They rely on data from a single year or a specific region for analysis, failing to fully consider the cumulative effects of data from multiple years and regions, as well as the bias characteristics of each organization. They also ignore factors such as historical performance of organizations, regional differences, data transparency, and timeliness, lacking a comprehensive and systematic evaluation standard. This application, through research, finds that this deficiency can be addressed through a polling organization effect analysis method based on comparative corrective and quantitative evaluation. This method can provide a more objective and accurate reference for election research and prediction, effectively assess the long-term performance and systematic biases of each organization, and has a certain degree of universality in predicting major events such as elections, with broad practical significance in political analysis and decision support. Summary of the Invention

[0011] To address the shortcomings of existing rating methods, such as insufficient multi-dimensional correction capabilities, limited quantitative evaluation indicators, and limited cross-domain applicability, this invention provides a polling agency effect analysis method based on comparative correction and quantitative evaluation.

[0012] In a first aspect, embodiments of this application provide a method for analyzing the effects of polling organizations based on comparative bias correction and quantitative evaluation, the method comprising:

[0013] Election data preparation steps: Obtain historical voting data from various states and polling data released by various polling organizations from multiple sources, and process the data from different sources; classify polling organizations into levels based on the constructed polling organization rating framework; calculate polling indicator values ​​based on the polling data and construct a polling indicator system;

[0014] Correction steps for polling organizations: Based on the polling indicator system, through comparison of multiple polling organizations and multi-dimensional comparative analysis within the same polling organization, the bias of the polling data is corrected;

[0015] The steps for comprehensive quantitative assessment of institutional effects are as follows: Combining historical voting data, polling indicator system, and polling data bias, a comprehensive quantitative assessment is conducted to construct an institutional rating model. Based on the polling institution's level, the final score and rating of each polling institution are calculated.

[0016] In a specific embodiment of the present invention, the above-mentioned election data preparation steps include:

[0017] Election data fusion and processing steps: Collect historical voting results data from various states and publicly released polling data from various organizations through multiple channels, integrate data from different sources, and obtain processed polling data by cleaning, transforming, and verifying various data fields;

[0018] Institutional rating framework design steps: Based on the platform's institutional rating framework, determine the final rating categories and number distribution of polling institutions;

[0019] The rating index system was constructed and calculated as follows: the party support rate was calculated, the number of polls conducted by the polling agency in each state each year was counted, the error between the poll conducted by the agency in each state most recently before the election date and the actual party support rate for that year was calculated, the party bias of a single poll was calculated, and the agency's transparency was obtained. Based on the data obtained from the calculation, a rating index system for the polling agency was constructed.

[0020] In a specific embodiment of the present invention, the steps for constructing and calculating the above-mentioned rating index system include:

[0021] The party support rate was calculated and normalized by selecting the voting data of the two major parties for all poll data items and calculating the vote share in the form of percentages.

[0022] Statistics on the number of polls conducted in each state show how many times each polling organization conducts polls in each state each year.

[0023] Error calculation and standardization involve subtracting the vote share of one of the two major parties in the poll data from the actual vote share of one of the two major parties in the general election of the same year to obtain a percentage error value. The error results are then grouped by year and standardized separately. The sign of the error value indicates the party bias of the poll, and the absolute value indicates the magnitude of the error.

[0024] In a specific embodiment of the present invention, the above-mentioned polling organization's correction steps include:

[0025] The polling data and actual results from multiple polling organizations were corrected. The error value of each polling organization for each poll was calculated and standardized for different polling sessions.

[0026] The time-series correction of data from multiple elections by the same polling organization is used to compare the error of the same organization in different election years, calculate the error trend and score, and adopt a weighted calculation method, introducing a time decay coefficient to reflect the dynamic performance of the organization.

[0027] Correction of key regional data from the same polling organization, focusing on analyzing the polling performance of each organization in key regions, and calculating the error of each organization in key states separately;

[0028] Based on bias correction using key reference polling organizations, this study selects top-ranked organizations within key regions and compares them with election prediction polling organizations selected by authoritative websites. The Area Under the Curve (AUC) value is introduced as an auxiliary evaluation metric. This value reflects the model's comprehensive ability to distinguish between positive and negative samples and is commonly used to measure model performance in classification tasks. The introduction of AUC effectively avoids evaluation bias caused by factors such as class imbalance, further improving the overall evaluation metric system and helping to optimize the performance of the computational model from a more comprehensive perspective.

[0029] In a specific embodiment of the present invention, the above-mentioned comprehensive quantitative assessment step of the mechanism effect level includes:

[0030] Based on the calculated multi-year error values, the conversion error value is calculated, and the error value with year weight is calculated using the term decay coefficient.

[0031] The comprehensive quantitative assessment includes: error values ​​weighted by year, the number of surveys conducted by the organization, and the impact of the organization's transparency, which are used to calculate the final score for each polling organization.

[0032] In a specific embodiment of the present invention, the above-mentioned comprehensive quantitative assessment step of the mechanism effect level further includes:

[0033] The final scores of each polling organization were processed using standardization, and after being sorted, the final scores were divided into levels according to the different distribution ratios determined by the polling organization rating framework to obtain the final level of each polling organization.

[0034] The final ranking results will be corrected by polling organizations based on key reference points.

[0035] In a specific embodiment of the present invention, the final scores of each polling organization are calculated using the following method:

[0036]

[0037] Wherein, error represents the error value; converse_error represents the error conversion value; W year The term decay coefficient is represented by α and β, which represent the weighting coefficients for the number of institutional surveys and institutional transparency, respectively; total_poll_num represents the number of institutional surveys, and Transparency represents institutional transparency.

[0038] Secondly, this application provides a polling agency effect analysis system based on comparative bias correction and quantitative evaluation, employing the polling agency effect analysis method based on comparative bias correction and quantitative evaluation as described above. The system includes:

[0039] Election data preparation module: Acquire historical voting data from various states and polling data released by various polling organizations from multiple channels, and process data from different sources; classify polling organizations into levels based on the constructed polling organization rating framework; calculate polling indicator values ​​based on polling data, and construct a polling indicator system;

[0040] Polling data correction module: Based on the polling indicator system, it corrects polling data bias by comparing multiple polling organizations and conducting multi-dimensional comparative analysis within the same polling organization.

[0041] The comprehensive quantitative assessment module for institutional effect levels combines historical voting data, polling indicator systems, and polling data bias to construct an institutional rating model. Based on the polling institution's level, the final score and rating of each polling institution are calculated.

[0042] Thirdly, embodiments of this application provide a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the aforementioned polling agency effect correction and quantitative evaluation steps.

[0043] Fourthly, embodiments of this application provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the polling agency effect correction and quantitative evaluation steps as described above.

[0044] The present invention designs a polling agency effect analysis method based on comparative correction and quantitative evaluation, including the following steps: 1) election data fusion processing; 2) agency rating framework design; 3) rating indicator system construction and calculation; 4) polling agency correction through multi-dimensional comparative analysis; 5) comprehensive quantitative evaluation of agency effect level.

[0045] Compared with existing technologies, it has the following outstanding advantages:

[0046] This invention constructs a comprehensive, multi-dimensional institutional rating model that comprehensively considers the influence of institutional bias levels, poll frequency, transparency, and term decay, reducing reliance on data from a single year or specific region. Through multi-dimensional comparative correction methods, it can effectively assess the long-term performance of each institution, improving the accuracy and objectivity of polling agency rating results and providing a reliable basis for further forecasting task selection. It solves the problems of insufficient multi-dimensional correction capabilities, limited quantitative evaluation indicators, and limited cross-domain applicability in existing rating methods. This invention uses polling data from the 2008, 2012, 2016, and 2020 US presidential elections as the data source for modeling, playing a crucial role in predicting the 2024 election results and achieving efficient data-driven modeling. Attached Figure Description

[0047] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0048] Figure 1 This is a schematic diagram of the polling agency effect analysis method based on comparative bias correction and quantitative evaluation of the present invention.

[0049] Figure 2 This is a schematic diagram of the method flow of an embodiment of the present invention;

[0050] Figure 3 This is a flowchart illustrating the construction process of the comprehensive quantitative assessment agency effect rating model in this invention.

[0051] Figure 4 This is a schematic diagram of a polling agency effect analysis system based on comparative bias correction and quantitative evaluation, as described in an embodiment of the present invention.

[0052] Figure 5 This is a schematic diagram of the computer hardware of the present invention. Detailed Implementation

[0053] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0054] It should also be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0055] It should also be understood that, in various embodiments of the present invention, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0056] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0057] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0058] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0059] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0060] To make the above-mentioned features and effects of the present invention clearer and easier to understand, specific embodiments are described below in conjunction with the accompanying drawings. This specification discloses one or more embodiments incorporating the features of the present invention. The disclosed embodiments are merely illustrative. The scope of protection of the present invention is not limited to the disclosed embodiments, but is defined by the appended claims.

[0061] The following are system embodiments corresponding to the above method embodiments. This embodiment can be implemented in conjunction with the above embodiments. The relevant technical details mentioned in the above embodiments are still valid in this embodiment, and will not be repeated here to reduce repetition. Accordingly, the relevant technical details mentioned in this embodiment can also be applied to the above embodiments.

[0062] This invention aims to provide a method for analyzing the effects of polling organizations based on comparative bias correction and quantitative evaluation. The method systematically analyzes the polling results and actual results from various organizations in different states over the years, calculating biases across organizations, time periods, and regions. It performs multi-dimensional comparative bias correction, comprehensively considering factors such as bias level, number of surveys, transparency, and session decay, establishing a multi-dimensional comparative bias correction and comprehensive quantitative evaluation rating model. This model can quantify the overall score and rating of organizations, providing a reliable basis for predictive models.

[0063] The institutional effect rating modeling method proposed in this invention includes the following steps:

[0064] 1. Election Data Integration and Processing. This involves acquiring historical voting results data from various states and publicly released polling data from various organizations, integrating the data from different sources, and performing cleaning, transformation, and verification operations on various data fields to create standardized and accurate data, thus providing a more precise basis for analysis.

[0065] 2. Institutional Rating Framework Design. This invention references the rating framework of FiveThirtyEight (a media platform focused on data analysis and prediction), dividing institutions into four levels: A, B, C, and D, with the distribution ratios of the four levels set at 15%, 35%, 35%, and 15%, respectively.

[0066] 3. Construction and Calculation of Rating Indicator System. This involves constructing the indicator system required for the institutional rating model and calculating the values ​​of each indicator based on polling data. Key indicators include party support rate, number of polls conducted in each state, party bias, and the institution's level of bias.

[0067] 4. Multi-dimensional comparative analysis for polling agency bias correction. This invention corrects the bias of polling agencies and improves the accuracy of ratings through multi-dimensional comparative analysis, mainly including the following dimensions: 1) Correction of polling data and actual results from multiple polling agencies; 2) Correction of data time sequence from multiple polling agencies; 3) Correction of data from key regions from the same polling agency; 4) Correction of polling agency bias based on key reference systems.

[0068] 5. Comprehensive Quantitative Assessment of Institutional Effect Level. This invention comprehensively considers factors such as the bias level (historical error) of polling organizations, the number of polls conducted by the organization, the transparency of the organization's data, and the frequency decay (timeliness of polls) to construct a rating model through comprehensive quantitative assessment. It calculates the final score and rating of each organization, providing reference institutional data for subsequent prediction tasks.

[0069] The methods of this application will be described in detail below with reference to specific embodiments:

[0070] Example 1

[0071] This invention proposes a method for analyzing the effects of polling organizations based on comparative bias correction and quantitative evaluation. The overall process of the method is as follows: Figure 1 and Figure 2 As shown. This institutional effect rating method mainly includes five parts: election data fusion and processing, institutional rating framework design, rating indicator system construction and calculation, multi-dimensional comparative analysis for polling institution bias correction, and comprehensive quantitative assessment of institutional effect levels. Specifically, this application provides a polling institution effect analysis method based on comparative bias correction and quantitative assessment, the method including:

[0072] Election data preparation step 10: Obtain historical voting data from various states and polling data released by various polling organizations from multiple channels, and process the data from different sources; classify the polling organizations according to the constructed polling organization rating framework; calculate the polling index values ​​based on the polling data and construct a polling index system;

[0073] In a specific embodiment of the present invention, the above-mentioned election data preparation step 10 includes: step 1, step 2, and step 3, specifically as follows:

[0074] Election data fusion and processing step 1: Collect historical voting results data from various states and publicly released polling data from various organizations through multiple channels, integrate data from different sources, and obtain processed polling data by cleaning, transforming and verifying various data fields;

[0075] Specifically, in the specific embodiments of the present invention, election data fusion processing involves collecting historical voting results data from various states and publicly released polling data from various organizations through channels such as the U.S. National Archives and the official websites of relevant polling agencies. Data from different sources is integrated, and standardized and accurate data is formed by cleaning, transforming, and verifying various data fields.

[0076] The goal of Step 1 is to obtain accurate and reliable polling data and historical voting data from multiple channels, and extract the data items required by this invention. Each data item must include complete information such as the year of the poll, region (state name), organization name, poll time (specific date), scale (number of voters participating in the poll), number of votes from the Democratic Party, number of votes from the Republican Party, and number of votes from third-party parties (if any).

[0077] More specifically, the data supporting this invention includes multiple data sources, primarily the National Archives and relevant authoritative polling websites (such as 270towin, FiveThirtyEight, and realclearpolling). Detailed information about these websites is shown in Table 1, and the data field information is shown in Table 2. According to one embodiment of this invention, the polling data used includes data from the 2008, 2012, 2016, and 2020 US presidential elections, and primarily focuses on analyzing data from key regions, namely the seven swing states (Pennsylvania, Nevada, Arizona, Georgia, Michigan, North Carolina, and Wisconsin).

[0078] Table 1

[0079]

[0080] Table 2

[0081]

[0082] Step 2 of the institutional rating framework design: Based on the platform's institutional rating framework, determine the types and number distribution of the final rated polling organizations;

[0083] Specifically, in the specific embodiments of the present invention, the institutional rating framework design refers to the institutional rating framework design of the FiveThirtyEight platform to determine the final rating categories and quantity distribution.

[0084] The objective of step 2 is to design the final rating framework given by the present invention. According to one embodiment of the present invention, referring to the institutional rating model of the FiveThirtyEight platform for political election prediction analysis, institutions are divided into four levels: A, B, C, and D, and distributed according to the proportion of the four levels being 15%, 35%, 35%, and 15%, respectively.

[0085] The third step in the construction and calculation of the rating index system is as follows: calculate the party support rate, count the number of polls conducted by the polling agency in each state each year, calculate the error between the poll conducted by the agency in each state most recent to the election date and the actual party support rate for that year, calculate the party bias of a single poll, and obtain the agency's transparency. Based on the data obtained from the calculation, a rating index system for the polling agency is constructed.

[0086] Specifically, in the specific embodiments of this invention, the rating index system is constructed and calculated. To ensure the comprehensiveness and accuracy of the rating, this invention constructs a complete rating index system to provide data support for building an institutional rating model. This system mainly includes calculating party support rates, counting the number of polls conducted by polling agencies in each state each year, calculating the error between the poll conducted by the agency in each state most recently before the election date and the actual party support rate for that year, the party bias of a single poll, and obtaining institutional transparency, etc. The corresponding data results need to be standardized.

[0087] The goal of step 3 is to calculate and obtain the various rating metrics needed to build the model, mainly including party support rate, number of polls in each state, party bias, and error.

[0088] Step 4 of the polling organization correction process: Based on the polling indicator system, through comparison of multiple polling organizations and multi-dimensional comparative analysis within the same polling organization, the bias of the polling data is corrected.

[0089] Specifically, in the specific embodiments of this invention, the multi-dimensional comparative analysis of polling agency bias correction involves comparative correction across multiple agencies, time periods, and regions. This mainly includes bias correction of polling data and actual results from multiple polling agencies, time-series correction of data from the same polling agency across multiple sessions, and data correction of the same polling agency in key regions. In addition, it also includes polling agency bias correction based on key reference systems. That is, after the ranking is determined, agencies with the highest ranking scores in each state are selected as reference agencies for predicting the voting situation in that state in 2024, and compared with the list of reference agencies for election prediction work in each state provided by authoritative websites to verify the effectiveness of the model.

[0090] The goal of step 4 is to correct the biases of polling agencies through multi-dimensional comparative analysis, thereby improving the accuracy of the ratings.

[0091] Step 5 of the comprehensive quantitative assessment of institutional effect level: Combining historical voting data, polling indicator system and polling data deviation, a comprehensive quantitative assessment is conducted to construct an institutional rating model. Based on the polling institution's level, the final score and rating of each polling institution are calculated.

[0092] Specifically, in the specific embodiments of the present invention, the level of institutional effect is comprehensively and quantitatively evaluated. This evaluation takes into account factors such as the bias level (historical error) of the polling organization, the number of polls conducted by the organization, the transparency of the organization's data, and the decay of poll sessions (timeliness of polls). The evaluation constructs a rating model and calculates the final score and rating of each organization.

[0093] In a specific embodiment of the present invention, step 3 of the above-mentioned rating index system construction and calculation includes:

[0094] Step 301: Calculation and normalization of party support rates. Select the voting data of the two major parties from all poll data items and calculate the vote share in percentage form.

[0095] In a specific embodiment of the present invention, party support rate calculation and normalization are performed. For all poll data entries, only the voting data of the Democratic Party (D) and the Republican Party (R) are retained (without considering third parties or other parties), and the vote share is calculated as a percentage, ensuring that the sum of the two is 100%. The goal of this step is to simplify the error calculation, so that the error can be calculated and the party bias can be determined based on the data of only one party.

[0096] According to one embodiment of the present invention, suppose that in 2020, the poll results of a certain polling organization A in state B were "Democratic_polls = 45.8, Republican_polls = 49.3". After normalization, the result is "Democratic_polls = 48.16%, Republican_polls = 51.84%", and Democratic_polls + Republican_polls = 100%. The actual vote results in state B in 2020 were "Democratic_votes = 1,034,707, Republican_votes = 1,230,111, Other_votes = 28,657, Total_votes = 2,293,475". After normalization, the result is "Democratic_votes = 45.69%, Republican_votes = 54.31%".

[0097] Step 302: Statistics on the number of polls conducted in each state, including the number of times each polling organization conducts polls in each state each year;

[0098] In a specific embodiment of this invention, the number of polls conducted in each state is statistically analyzed to determine the number of polls conducted annually by each polling organization in each state (poll_num). According to one embodiment of the invention, the more polls a polling organization conducts and the larger the scale of its polls, the higher its score should be. For example, if two polling organizations have roughly the same error rate, the organization that conducts more polls will have a relatively higher score and a higher ranking. Finally, the total number of polls conducted by each organization in each state and each year is summed to obtain the total number of surveys conducted by each organization in the seven swing states (total_poll_num).

[0099] Step 303: Error Calculation and Standardization. The percentage error value is obtained by subtracting the vote share of one of the two major parties in the poll data from the actual vote share of one of the two major parties in the general election of the same year. The error results are grouped by year and standardized separately. The sign of the error value indicates the party bias of the poll, and the absolute value indicates the magnitude of the error.

[0100] In a specific embodiment of the present invention, error calculation and standardization are performed. The percentage error is obtained by subtracting the D Party's vote share from the poll data from the actual D Party vote share in the general election of the same year. The sign of this value indicates the partisan bias of the poll, and the absolute value indicates the magnitude of the error. The error results are grouped according to year, and the data from different years are standardized separately, with the error standardized to the [0,1] interval.

[0101] According to one embodiment of the present invention, the error is calculated by selecting the poll data from each state most recent to the election date each year; therefore, the date field needs to be sorted in reverse order. Based on the party support rate normalization step described in S301, the error can be calculated using only the data from one party. This invention selects the Democratic Party as the calculation standard, using the difference between the poll data (Democratic_polls) and the actual data (Democratic_votes) to obtain the error. The specific calculation method is as follows:

[0102] Error=Democratic_polls-Democratic_votes (1)

[0103] According to one embodiment of the present invention, after obtaining the errors for each year, it is necessary to standardize the error results for each state in the same year. The present invention standardizes the error data after taking the absolute value to the [0,1] interval, and for each error data e...i The standardized value e' i The calculation is as follows:

[0104]

[0105] Among them, e min =min(e1,e2,…,e n ), representing the minimum error data of each institution in the same year, e max =max(e1,e2,…,e n ), which represents the maximum value of error data for each institution in the same year.

[0106] According to one embodiment of the present invention, the sign of the error can be used to determine the partisan bias of a single poll, specifically defined as:

[0107]

[0108] Where D stands for Democratic Party and R stands for Republican Party.

[0109] In a specific embodiment of the present invention, the above-mentioned polling organization correction step 4 includes:

[0110] Step 401: Correcting the poll data and actual results from multiple polling organizations. Based on the calculation of the error value of each poll by each polling organization, standardization is performed between different sessions.

[0111] In a specific embodiment of the present invention, multiple polling organizations correct the polling data and actual results by performing step S303, which calculates the error of each polling organization for each polling session, and performs standardization processing between different sessions to ensure the comparability of the data.

[0112] Step 402: Time-series correction of data from multiple elections by the same polling organization. Compare the errors of the same organization in different election years, calculate the error trend and score, use a weighted calculation method, and introduce a time decay coefficient to reflect the dynamic performance of the organization.

[0113] In a specific embodiment of the present invention, the time-series correction of data from the same polling organization across multiple elections is performed. The errors of the same organization in different election years are compared, and the error trend and score are calculated. According to one embodiment of the present invention, a weighted calculation method is adopted, and a time decay coefficient is introduced to strengthen the weight of recent data, so that the latest data has a greater impact on the rating, while still maintaining the reference value of historical data and reflecting the dynamic performance of the organization.

[0114] Step 403: Correction of key regional data from the same polling organization, focusing on analyzing the polling performance of each organization in key regions and calculating the error of each organization in key states separately;

[0115] In a specific embodiment of the present invention, the data correction for the same polling organization in key regions focuses on analyzing the polling performance of each organization in key regions (such as swing states) and separately calculates the error of each organization in key states, so as to improve the accuracy of the assessment of organizations with uneven regional coverage.

[0116] Step 404: Correcting bias based on key reference polling organizations. Within key regions, select top-ranked polling organizations and compare them with election prediction polling organizations selected by authoritative websites. Introduce the AUC (Area Under the Curve) value as an auxiliary evaluation metric. This value reflects the model's comprehensive ability to distinguish between positive and negative samples and is commonly used to measure model performance in classification tasks. Introducing the AUC value can effectively avoid evaluation bias caused by factors such as class imbalance, further improve the overall evaluation metric system, and help optimize the performance of the computational model from a more comprehensive perspective.

[0117] In a specific embodiment of this invention, bias correction is based on polling organizations with a key reference system. Within a key region, top-ranked organizations are selected as a list of organizations that can be used as references in that state, and compared with polling organizations selected by relevant authoritative websites that can be used for 2024 election predictions. To more comprehensively evaluate the model's classification performance, the AUC (Area Under the Curve) value is used as an auxiliary evaluation metric. The AUC value is a common model evaluation metric in machine learning tasks, referring to the area under the Receiver Operating Characteristic Curve (ROC), used to reflect the model's comprehensive ability to distinguish between positive and negative samples at different classification thresholds. The ROC curve, with the False Positive Rate on the horizontal axis and the True Positive Rate on the vertical axis, depicts the model's classification performance at various decision thresholds. The AUC value is obtained by calculating the area under this curve, and its range is [0,1]. The closer the AUC value is to 1, the stronger the model's ability to distinguish between positive and negative samples, and the better its overall performance.

[0118] In a specific embodiment of the present invention, step 5 of the above-mentioned comprehensive quantitative assessment of the mechanism effect level includes:

[0119] Based on the calculated multi-year error values, the conversion error value is calculated, and the error value with year weight is calculated using the term decay coefficient.

[0120] The comprehensive quantitative assessment includes: error values ​​weighted by year, the number of surveys conducted by the organization, and the impact of the organization's transparency, which are used to calculate the final score for each polling organization;

[0121] The final scores of each polling organization were processed using standardization, and after being sorted, the final scores were divided into levels according to the different distribution ratios determined by the polling organization rating framework to obtain the final level of each polling organization.

[0122] The final ranking results will be corrected by polling organizations based on key reference points.

[0123] In a specific embodiment of the present invention, the final scores of each polling organization are calculated using the following method:

[0124]

[0125] Wherein, error represents the error value; converse_error represents the error conversion value; W year The term decay coefficient is represented by α and β, which represent the weighting coefficients for the number of institutional surveys and institutional transparency, respectively; total_poll_num represents the number of institutional surveys, and Transparency represents institutional transparency.

[0126] The goal of S5 is to comprehensively consider factors such as the bias level of polling organizations (historical error), the number of polls conducted by the organization, the transparency of the organization's data, and the decay of polling sessions (timeliness of polling), to construct a rating model through comprehensive quantitative evaluation, and to calculate the final score and rating for each organization. The specific steps are as follows: Figure 3 express.

[0127] According to one embodiment of the present invention, the model first needs to consider the error index (Note: the error referred to below is the absolute value error, Error≥0). In step S33 above, the errors of the mechanism for the four years of 2008, 2012, 2016, and 2020 have been calculated and represented by 2008_error, 2012_error, 2016_error, and 2020_error respectively (Note: the errors here have been standardized, 0≤error≤1).

[0128] According to one embodiment of the present invention, since a smaller error value is better, and in order to use the weighted summation method described in step S402, it needs to be transformed. The processing method of the present invention is to first calculate the converse_error (error transformation value), and then introduce a degree attenuation coefficient W. yearThis is used to calculate an error value (W_error) with year weighting, which strengthens the weight of recent data, making the latest data have a greater impact on the rating, while still maintaining the reference value of historical data and reflecting the dynamic performance of the institution.

[0129] According to one embodiment of the present invention, the comprehensive quantitative assessment also needs to take into account the impact of the number of institutional surveys (total_poll_num) and institutional transparency (Transparency: a constant term, referring to the transparency score published on the FiveThirtyEight platform) in order to calculate the final score.

[0130] According to one embodiment of the present invention, the specific calculation method is as follows:

[0131]

[0132] In formula (5), W year The year weight (term decay coefficient) is represented by the weight values ​​for 2020, 2016, 2012, and 2008, which are {0.4, 0.3, 0.2, 0.1} respectively. In formula (6), α and β represent the weight coefficients for the number of institutional surveys and institutional transparency, respectively. According to an embodiment of the present invention, α = 0.01 and β = 0.05 are set.

[0133] Then, the scores of each institution are standardized in the same way as mentioned in step S303. After sorting from largest to smallest, the scores are divided into grades A, B, C and D with a distribution ratio of 15%, 35%, 35% and 15% respectively, to obtain the final grade of each institution.

[0134] According to one embodiment of the present invention, after obtaining the results, a bias correction based on key reference systems of polling organizations needs to be performed in step S404. The present invention selects the top-rated organizations in each of the seven swing states as reference organizations for predicting the 2024 presidential election in those states. Compared with the reference organizations given by relevant authoritative websites, the AUC value can reach approximately 0.8523. At the same time, relevant authoritative experts have also affirmed the results obtained by the present invention, providing a favorable reference basis for subsequent prediction work.

[0135] As described above, the method of the present invention can be implemented well.

[0136] Example 2

[0137] like Figure 4 As shown in the embodiments of this application, a polling agency effect analysis system based on comparative bias correction and quantitative evaluation is provided. The system employs the polling agency effect analysis method based on comparative bias correction and quantitative evaluation as described above, and includes:

[0138] Election Data Preparation Module 60: Obtain historical voting data from various states and polling data released by various polling organizations from multiple channels, and process data from different sources; classify polling organizations into different levels based on the constructed polling organization rating framework; calculate polling indicator values ​​based on polling data, and construct a polling indicator system;

[0139] Polling organization bias correction module 70: Based on the polling indicator system, it corrects polling data bias by comparing multiple polling organizations and conducting multi-dimensional comparative analysis within the same polling organization.

[0140] Module 80: Comprehensive quantitative assessment of institutional effect level: Combining historical voting data, polling indicator system and polling data deviation, a comprehensive quantitative assessment is constructed to build an institutional rating model. Based on the polling institution level, the final score and rating of each polling institution are calculated.

[0141] Example 3

[0142] This application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the polling agency effect analysis method based on comparative bias correction and quantitative evaluation.

[0143] Example 4

[0144] This application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps of the polling agency effect analysis method based on comparative bias correction and quantitative evaluation as described above.

[0145] In addition, combined Figure 1 The polling agency effect analysis method based on comparative bias correction and quantitative evaluation described in this application embodiment can be implemented by electronic devices, such as computer devices. Figure 5 This is a schematic diagram of the hardware structure of a computer device according to an embodiment of this application.

[0146] In some embodiments, the computer device may further include a communication interface 83 and a bus 80. For example, Figure 5 As shown, the processor 81, memory 82, and communication interface 83 are connected through bus 80 and complete communication with each other.

[0147] Specifically, the processor 81 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0148] The memory 82 can be used to store or cache various data files that need to be processed and / or communicated, as well as possible computer program instructions executed by the processor 81.

[0149] The processor 81 reads and executes computer program instructions stored in the memory 82 to implement any of the polling agency effect analysis methods based on comparative bias correction and quantitative evaluation in the above embodiments.

[0150] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0151] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for analyzing the effects of polling organizations based on comparative bias correction and quantitative evaluation, characterized in that, The method includes: Election data preparation steps: Obtain historical voting data from various states and polling data released by various polling organizations from multiple channels, and process the data from different sources; classify polling organizations into different levels based on the constructed polling organization rating framework; calculate polling index values ​​based on the polling data, and construct a polling index system; Correction steps for polling organizations: Based on the aforementioned polling indicator system, the bias in the polling data is corrected through comparisons of multiple polling organizations and multi-dimensional comparative analysis within the same polling organization. The steps for comprehensive quantitative assessment of institutional effect level are as follows: Combining historical voting data, the polling indicator system, and the polling data deviation, a comprehensive quantitative assessment is conducted to construct an institutional rating model. Based on the polling institution level, the final score and rating of each polling institution are calculated.

2. The polling agency effect analysis method based on comparative bias correction and quantitative evaluation according to claim 1, characterized in that, The election data preparation steps include: Election data fusion and processing steps: Collect historical voting results data from various states and publicly released polling data from various organizations through multiple channels, integrate data from different sources, and obtain processed polling data by cleaning, transforming, and verifying various data fields; Institutional rating framework design steps: Based on the platform's institutional rating framework, determine the final rating categories and number distribution of polling institutions; The rating index system was constructed and calculated as follows: the party support rate was calculated, the number of polls conducted by the polling agency in each state each year was counted, the error between the poll conducted by the agency in each state most recently before the election date and the actual party support rate for that year was calculated, the party bias of a single poll was calculated, and the agency's transparency was obtained. Based on the data obtained from the calculation, a rating index system for the polling agency was constructed.

3. The polling agency effect analysis method based on comparative bias correction and quantitative evaluation according to claim 2, characterized in that, The steps for constructing and calculating the rating indicator system include: The party support rate was calculated and normalized by selecting the voting data of the two major parties for all poll data items and calculating the vote share in the form of percentages. Statistics on the number of polls conducted in each state show how many times each polling organization conducts polls in each state each year. Error calculation and standardization involve subtracting the vote share of one of the two major parties in the poll data from the actual vote share of one of the two major parties in the general election of the same year to obtain a percentage error value. The error results are then grouped by year and standardized separately. The sign of the error value indicates the party bias of the poll, and the absolute value indicates the magnitude of the error.

4. The polling agency effect analysis method based on comparative bias correction and quantitative evaluation according to claim 3, characterized in that, The corrective measures taken by the polling organization include: The polling data and actual results from multiple polling organizations are corrected by calculating the error value of each poll conducted by each polling organization for each period, and standardization is performed separately for different periods. The time-series correction of data from multiple elections by the same polling organization is used to compare the error of the same organization in different election years, calculate the error trend and score, and adopt a weighted calculation method, introducing a time decay coefficient to reflect the dynamic performance of the organization. Correction of key regional data from the same polling organization, focusing on analyzing the polling performance of each organization in key regions, and calculating the error of each organization in key states separately; Based on the bias correction of polling organizations with key reference points, the top-ranked organizations in key regions are selected and compared with polling organizations that predict elections selected by relevant authoritative websites. The AUC auxiliary evaluation index is introduced to improve the overall evaluation index system and optimize the performance of the calculation model.

5. The polling agency effect analysis method based on comparative bias correction and quantitative evaluation according to claim 2, characterized in that, The steps for the comprehensive quantitative assessment of the institutional effect level include: Based on the calculated error values ​​over many years, the conversion error value is calculated, and the error value with year weight is calculated using the term decay coefficient. The comprehensive quantitative assessment includes: the error value with year weighting, the number of surveys conducted by the organization, and the impact of the organization's transparency, to calculate the final score for each polling organization.

6. The polling agency effect analysis method based on comparative bias correction and quantitative evaluation according to claim 5, characterized in that, The comprehensive quantitative assessment steps for the institutional effect level also include: The final scores of each polling organization are processed using the standardization method, and after being sorted, the final scores are divided into different levels according to the different distribution ratios determined by the polling organization rating framework to obtain the final level of each polling organization. The final ranking results will be corrected by polling organizations based on key reference points.

7. The polling agency effect analysis method based on comparative bias correction and quantitative evaluation according to claim 6, characterized in that, The final scores for each polling organization are calculated using the following method: Wherein, error represents the error value; converse_error represents the error conversion value; W year The term decay coefficient is represented by α and β, which represent the weighting coefficients for the number of institutional surveys and institutional transparency, respectively; total_poll_num represents the number of institutional surveys, and Transparency represents institutional transparency.

8. A polling agency effect analysis system based on comparative bias correction and quantitative evaluation, employing the polling agency effect analysis method based on comparative bias correction and quantitative evaluation as described in any one of claims 1-7, characterized in that, The system includes: Election data preparation module: Acquire historical voting data from various states and polling data released by various polling organizations from multiple channels, and process data from different sources; classify polling organizations into levels based on the constructed polling organization rating framework; calculate polling index values ​​based on the polling data, and construct a polling index system; Polling data correction module: Based on the polling indicator system, the module corrects the bias in the polling data by comparing multiple polling organizations and conducting multi-dimensional comparative analysis within the same polling organization. The comprehensive quantitative assessment module for institutional effect level combines historical voting data, the polling indicator system, and the polling data deviation to construct an institutional rating model. Based on the polling institution's level, the final score and rating of each polling institution are calculated.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the polling agency effect analysis method based on contrast correction and quantitative evaluation as described in any one of claims 1-7.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the polling agency effect analysis method based on comparative bias correction and quantitative evaluation as described in any one of claims 1 to 7.

Images