Method and system for determining survey result of questionnaire

Abstract

The invention discloses a questionnaire survey result determination method and system, and relates to the technical field of data processing, and the method comprises the steps: obtaining filled questionnaires, and removing invalid questionnaires to obtain a target questionnaire; carrying out numerical coding on answers of all questions in the target questionnaire, carrying out reliability test and validity test on the target questionnaire after numerical coding, and carrying out feature selection on all questions in the target questionnaire after test by utilizing a Pearson's correlation coefficient and chi-square test; performing data integration on the target questionnaire after feature selection to obtain data features of the target questionnaire; inputting the data features of the target questionnaire into an investigation result prediction model of the questionnaire to obtain a predicted value of an investigation result of the questionnaire; according to the invention, the reliability and accuracy of the survey result of the questionnaire can be improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method and system for determining survey results of a questionnaire. Background Art

[0002] As an important data collection tool, questionnaires have gone through many years of development. From the earliest paper questionnaires to the widely used electronic questionnaires today, the form and method of questionnaires have continued to evolve. Traditional paper questionnaires are distributed by mail or face-to-face. Although they are convenient for reaching people who are not familiar with electronic devices, the data collection and collation process is time-consuming and laborious. Telephone questionnaires are another way to directly obtain feedback from respondents through telephone interviews. Although they can interact in real time, the labor cost is high. With the popularization of the Internet and e-mail, electronic questionnaires have become mainstream, which can collect data on a large scale efficiently and at low cost.

[0003] By systematically designing the questions in the questionnaire, the questionnaire can efficiently collect a large amount of data, help researchers understand the characteristics and needs of the target group, analyze the data collected by the questionnaire, determine the survey results of the questionnaire, provide valuable decision-making support, optimize products and services, and formulate scientific and reasonable strategies.

[0004] However, analyzing questionnaire data and determining the survey results is not an easy task. There are many difficulties and challenges. For example, there may be errors in the questionnaire questions, and the survey samples may be biased and cannot fully represent the target group, affecting the reliability of the questionnaire results. In addition, questionnaires usually contain multi-dimensional data, involving quantitative and qualitative issues, and the analysis and processing are complex. Although the existing numerical processing methods have achieved good results in many applications, there are still some deficiencies in processing questionnaire data, resulting in inaccurate predictions. Therefore, it is crucial to adopt scientific and reasonable numerical processing methods to achieve accurate questionnaire result predictions to improve the reliability of questionnaire results. Summary of the invention

[0005] The purpose of this application is to provide a method and system for determining the survey results of a questionnaire to solve the problem of low reliability and low accuracy of the survey results of the questionnaire.

[0006] To achieve the above objectives, this application provides the following solutions:

[0007] In a first aspect, the present application provides a method for determining a survey result of a questionnaire, comprising:

[0008] Obtain the completed questionnaires and eliminate invalid questionnaires from the completed questionnaires to obtain the target questionnaires; invalid questionnaires include questionnaires with contradictory answers to similar questions and questionnaires with ambiguous items selected for all questions;

[0009] Numerically encode the answers to all questions in the target questionnaire to obtain the numerically encoded target questionnaire;

[0010] Conduct reliability and validity tests on the numerically quantified target questionnaire to obtain the tested target questionnaire;

[0011] Using the Pearson correlation coefficient and chi-square test, feature selection is performed on all questions in the target questionnaire after the test to obtain the target questionnaire after feature selection;

[0012] Perform data integration on the target questionnaire after feature selection to obtain the data features of the target questionnaire;

[0013] The data features of the target questionnaire are input into the questionnaire survey result prediction model to obtain the predicted value of the questionnaire survey result; the questionnaire survey result prediction model is obtained by training the random forest model with training samples; the training samples include the sample values ​​of the data features of the questionnaire and the sample values ​​of the questionnaire survey result.

[0014] Optionally, the answers to all questions in the target questionnaire are numerically encoded to obtain a numerically encoded target questionnaire, which specifically includes:

[0015] Performing a first category determination on all questions in the target questionnaire to determine the first category to which each question belongs; the first category is a single-choice question or a multiple-choice question;

[0016] Based on the first category to which each question belongs, the corresponding numerical coding method is selected, and the answers corresponding to each question are numerically encoded according to the selected numerical coding method to obtain the numerically encoded target questionnaire; when the first category to which the question belongs is a single-choice question, the answer corresponding to the question is label-encoded; when the first category to which the question belongs is a multiple-choice question, the answer corresponding to the question is encoded 0-1.

[0017] Optionally, the expression for label encoding is:

[0018] y i =i,i∈{1,2,...,n};

[0019] Among them, y i Indicates the numerical label corresponding to the i-th option; n indicates the total number of options contained in a single-choice question;

[0020] The expression for 0-1 encoding is:

[0021]

[0022] Among them, y ijRepresents the binary code value of the jth option of the kth question; m is the total number of options contained in a multiple-choice question.

[0023] Optionally, the digitized target survey questionnaire is subjected to reliability test and validity test to obtain a tested target survey questionnaire, which specifically includes:

[0024] Preprocessing the digitized target questionnaire to obtain a preprocessed target questionnaire;

[0025] The reliability of all questions in the preprocessed target questionnaire was tested using Cronbach's α coefficient to obtain the target questionnaire after reliability test.

[0026] The formula X=ΛF+∈ is used to perform validity test on all questions in the target questionnaire after reliability test to obtain the target questionnaire after test; where X is the observed variable vector, Λ is the factor loading matrix, F is the common factor vector, and ∈ is the error vector.

[0027] Optionally, the Cronbach's alpha coefficient is calculated as:

[0028]

[0029] Where α is Cronbach's α coefficient; N is the number of questions; is the average covariance between items; is the average of the variances between items.

[0030] Optionally, preprocessing the digitized target questionnaire to obtain a preprocessed target questionnaire specifically includes:

[0031] The mean filling method is used to process the missing values ​​of the answers to all questions in the digitized target questionnaire to obtain the target questionnaire after missing value processing;

[0032] The box plot method is used to detect outliers in the answers to all questions in the target questionnaire after missing value processing, and the target questionnaire after outlier detection is obtained;

[0033] Standardize the answers to all questions in the target questionnaire after outlier detection to obtain a standardized target questionnaire;

[0034] The standardized target questionnaire is normalized to obtain the preprocessed target questionnaire.

[0035] Optionally, a box plot method is used to detect outliers in all answers in the target questionnaire after missing value processing, including:

[0036] The formula IQR = Q3-Q1 is used to calculate the interquartile range of all answers in the target questionnaire after missing value processing; where Q1 is the first quartile; Q3 is the third quartile; IQR is the interquartile range;

[0037] Based on the quartiles and the interquartile range, it is determined whether each answer in the target questionnaire after the missing value processing meets the preset range, and the answers that do not meet the preset range are eliminated.

[0038] Optionally, the Pearson correlation coefficient and the chi-square test are used to perform feature selection on all questions in the tested target questionnaire to obtain a target questionnaire after feature selection, specifically including:

[0039] Performing a second category determination on all questions in the tested target questionnaire to determine the second category to which the question belongs; the second category is a regression problem or a classification problem;

[0040] When it is determined that the second category to which the question belongs is a regression problem, the correlation between the question score and the target variable is calculated using the Pearson correlation coefficient, and the question whose correlation meets the first preset threshold is selected; the target variable is a sample value of the output result of the target questionnaire;

[0041] When it is determined that the second category to which the question belongs is a classification problem, the chi-square test is used to calculate the chi-square statistic between the question and the target variable, and the questions whose chi-square statistic meets the second preset threshold are selected to complete the feature selection and obtain the target questionnaire after feature selection.

[0042] Optionally, the training process of the questionnaire result prediction model specifically includes:

[0043] Input the training samples into the random forest model to obtain sample prediction values ​​of the questionnaire results;

[0044] A loss function is constructed according to the sample prediction value of the questionnaire results and the sample value of the questionnaire results, and the model parameters of the random forest model are adjusted according to the loss function to obtain a questionnaire result prediction model.

[0045] In a second aspect, the present application provides a system for determining a questionnaire result. The system for determining a questionnaire result is based on any one of the methods for determining a questionnaire result. The system for determining a questionnaire result comprises:

[0046] The target questionnaire determination unit is used to obtain the filled questionnaires and eliminate invalid questionnaires from the filled questionnaires to obtain the target questionnaires; invalid questionnaires include questionnaires with contradictory answers to similar questions and questionnaires with ambiguous items selected for all questions;

[0047] A unit for determining a target questionnaire after digitization is used to digitize the answers to all questions in the target questionnaire to obtain a target questionnaire after digitization;

[0048] The unit for determining the target survey questionnaire after verification is used to conduct reliability and validity tests on the digitized target survey questionnaire to obtain the target survey questionnaire after verification;

[0049] The target questionnaire determination unit after feature selection is used to perform feature selection on all questions in the target questionnaire after the test by using the Pearson correlation coefficient and the chi-square test to obtain the target questionnaire after feature selection;

[0050] A data feature determination unit, used to integrate the data of the target survey questionnaire after feature selection to obtain the data features of the target survey questionnaire;

[0051] The questionnaire output result determination unit is used to input the data features of the target questionnaire into the questionnaire survey result prediction model to obtain the predicted value of the questionnaire survey result; the questionnaire survey result prediction model is obtained by training the random forest model with training samples; the training samples include the sample values ​​of the data features of the questionnaire and the sample values ​​of the questionnaire survey result.

[0052] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0053] The present application provides a method and system for determining the survey results of a questionnaire, which obtains data features of a target questionnaire by successively performing invalid questionnaire elimination, numerical processing, reliability test and validity test, feature selection and data integration on the completed questionnaires, so that the data features can better reflect the real features of the questionnaire, and then inputs the data features into a survey result prediction model of the questionnaire to obtain the predicted value of the survey result of the questionnaire, so that the survey result of the questionnaire is more accurate and reliable. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0055] Figure 1 This is an application environment diagram of a method for determining a survey result of a questionnaire in an embodiment of the present application;

[0056] Figure 2A flowchart of a method for determining survey results of a questionnaire provided in an embodiment of the present application;

[0057] Figure 3 A visualization diagram of a questionnaire result provided in an embodiment of the present application;

[0058] Figure 4 Another visualization diagram of questionnaire results provided in one embodiment of the present application;

[0059] Figure 5 Another visualization diagram of questionnaire results provided in one embodiment of the present application;

[0060] Figure 6 Another visualization diagram of questionnaire results provided in one embodiment of the present application;

[0061] Figure 7 A schematic diagram of the functional modules of a system for determining questionnaire results provided in one embodiment of the present application.

[0062] Reference numerals:

[0063] Terminal-102, server-104, target questionnaire determination unit-1, digitized target questionnaire determination unit-2, verified target questionnaire determination unit-3, target questionnaire determination unit-4 after feature selection, data feature determination unit-5, questionnaire output result determination unit-6. DETAILED DESCRIPTION

[0064] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0065] Existing numerical processing methods have achieved good results in many applications, but there are still some shortcomings when processing questionnaire data, including problems with the order of categorical data, high dimensionality caused by one-hot encoding, ignoring the correlation and contextual information between different questions in the questionnaire, inefficient time and space on large-scale data sets, and difficulty in automatically adapting to different types of questionnaire data. These problems lead to inaccurate analysis results and information loss, increasing the complexity and cost of data processing. In addition, improper data processing can lead to inaccurate analysis results, which in turn affects decision-making and research conclusions. Improperly handled missing values ​​and outliers can introduce bias and distort results; improperly processed categorical data and high-dimensional data can affect the performance of machine learning models and lead to inaccurate predictions; incorrect encoding and data conversion can lead to misleading conclusions, affecting decision-making and strategy formulation. In addition, simplified or incorrect processing methods may lead to the loss of valuable information and reduce the comprehensiveness and depth of analysis.

[0066] The purpose of this application is to provide a method and system for determining the survey results of a questionnaire, aiming to improve the reliability and accuracy of the survey results of the questionnaire.

[0067] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0068] The method for determining the survey result of a questionnaire provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set up separately, or it can be integrated on the server 104, or it can be placed on the cloud or other servers. The terminal 102 can send the questionnaire to be processed to the server 104. After the server 104 receives the questionnaire to be processed, the server 104 determines the survey results for the questionnaire to be processed and obtains the predicted value of the survey results of the questionnaire. The server 104 can feed back the predicted value of the survey results of the obtained questionnaire to the terminal 102. In addition, in some embodiments, the method for determining the survey results of the questionnaire can also be implemented separately by the server 104 or the terminal 102, such as the terminal 102 can directly determine the survey results for the questionnaire to be processed, or the server 104 can obtain the questionnaire to be processed from the data storage system, and determine the survey results of the questionnaire for the questionnaire to be processed.

[0069] The terminal 102 may be, but is not limited to, various desktop computers, laptop computers, smart phones, tablet computers, IoT devices, and portable wearable devices. The IoT devices may be smart speakers, smart TVs, smart air conditioners, smart vehicle-mounted devices, etc. The portable wearable devices may be smart watches, smart bracelets, head-mounted devices, etc. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers, or may be a cloud server.

[0070] In an exemplary embodiment, Figure 2 As shown, a method for determining the survey results of a questionnaire is provided. The method is executed by a computer device, and can be executed by a computer device such as a terminal or a server alone, or by a terminal and a server together. In the embodiment of the present application, the method is applied to Figure 1 The server 104 in the example is used as an example to illustrate, including the following steps S1 to S6:

[0071] Step S1, obtain the completed questionnaires, and eliminate invalid questionnaires from the completed questionnaires to obtain the target questionnaire; invalid questionnaires include questionnaires with contradictory answers to similar questions and questionnaires with ambiguous items selected for all questions.

[0072] Step S2, numerically encode the answers to all questions in the target questionnaire to obtain a numerically encoded target questionnaire.

[0073] Step S3, performing reliability test and validity test on the digitized target questionnaire to obtain the tested target questionnaire.

[0074] Step S4, using the Pearson correlation coefficient and the chi-square test, perform feature selection on all questions in the tested target questionnaire to obtain the target questionnaire after feature selection.

[0075] Step S5, integrating the data of the target questionnaire after feature selection to obtain the data features of the target questionnaire.

[0076] Step S6, inputting the data features of the target questionnaire into the survey result prediction model of the questionnaire to obtain the predicted value of the survey result of the questionnaire; the survey result prediction model of the questionnaire is obtained by training the random forest model with training samples; the training samples include sample values ​​of the data features of the questionnaire and sample values ​​of the survey results of the questionnaire.

[0077] As an optional implementation, in step S1, in order to ensure the representativeness, accuracy and diversity of the questionnaire data, the questionnaire often needs to collect a large number of samples. In the questionnaire, an invalid questionnaire refers to a questionnaire that cannot be used for data analysis due to various reasons, such as an incomplete questionnaire, consistency errors, random filling, lack of key information, malicious answers, and questionnaires that do not meet the screening criteria. Invalid questionnaires may affect the overall quality of the questionnaire data and the analysis results, so they need to be identified and eliminated before data analysis. To ensure the quality of the data and the accuracy of the analysis results. The criteria for identifying invalid questionnaires are as follows:

[0078] 1) Contradictory answers to similar questions

[0079] For example, a question in the questionnaire is "How often do you surf the Internet?"; a question in the questionnaire is "How long do you spend online each week?". The inconsistent answers in the collected questionnaires, which include both "not surfing the Internet" and "how to surf the Internet", are screened and removed from the subsequent statistics.

[0080] 2) Select fuzzy items for all questions

[0081] Taking into account the different levels of understanding of the options among the subjects, fuzzy options are set for the subjects to choose when they are in a dilemma between extreme options. Since the subjects who choose all fuzzy items in all questions containing fuzzy items are not serious, their questionnaires are excluded.

[0082] As an optional implementation, step S2 specifically includes:

[0083] Step S21, performing a first category determination on all questions in the target questionnaire to determine the first category to which each question belongs; the first category is a single-choice question or a multiple-choice question.

[0084] Step S22, based on the first category to which each question belongs, select the corresponding numerical coding method, and numerically encode the answer corresponding to each question according to the selected numerical coding method to obtain the digitized target questionnaire; when the first category to which the question belongs is a single-choice question, label encode the answer corresponding to the question; when the first category to which the question belongs is a multiple-choice question, encode the answer corresponding to the question 0-1.

[0085] Specifically, label encoding can map text data to numerical labels for ordered variables. It is simple and direct, easy to operate, and can intuitively reflect the number of categories through the number of values. This application uses label encoding to process multiple-choice questions. Since the number of options for each question is different, the numerical encoding is set according to the number of options. For example, a question has 6 options A to F, which are encoded as numbers 1 to 6 respectively. When processing numerically, numbers are only used to distinguish categories, and their own size does not represent the evaluation of the question. Among them, the expression of label encoding is:

[0086] y i =i,i∈{1,2,...,n}(1)

[0087] Among them, y i Represents the numerical label corresponding to the i-th option. For example, options A, B, C, D, E, and F are assigned labels 1, 2, 3, 4, 5, and 6 respectively; n represents the total number of options contained in a multiple-choice question.

[0088] Since it is difficult to directly label the answers to multiple-choice questions, converting them into 01 variables according to the options is a more intuitive way of processing. For a multiple-choice question with several options and indefinite items, each option is set as a variable, using 01 coding, the checked variable is recorded as 1, and the unchecked variable is recorded as 0. The expression of 0-1 coding is:

[0089]

[0090] Among them, y ij represents the binary code value of the jth option of the kth question; m is the total number of options contained in a multiple-choice question. If this option is selected, then y kj =1; otherwise y ij =0.

[0091] For example, according to the above coding principles, a numerical target survey questionnaire was obtained, which provided a basis for subsequent analysis. The numerical target survey questionnaire example is as follows:

[0092] Table 1 Numericalized target questionnaire

[0093]

[0094] As an optional implementation, step S3 specifically includes:

[0095] Step S31, preprocessing the digitized target questionnaire to obtain the preprocessed target questionnaire; after completing label encoding and 0-1 encoding, preprocessing the digitized target questionnaire is a key step to ensure data quality, improve data analysis accuracy and model performance. Preprocessing includes missing value processing, outlier detection and processing, standardization and normalization, etc.

[0096] Step S32, using Cronbach's alpha coefficient to perform reliability test on all questions in the preprocessed target questionnaire to obtain the target questionnaire after reliability test.

[0097] Specifically, in order to ensure the accuracy and scientificity of the questionnaire data analysis, this application converts the questionnaire into a numerical scale through coding, and then uses reliability test and validity test to conduct a comprehensive verification of the pre-processed target questionnaire. Reliability test is used to evaluate the consistency and stability of questionnaire data, while validity test is used to verify the accuracy and rationality of the questionnaire measurement content. These two verification methods can effectively identify and eliminate potential data problems and ensure that subsequent data analysis is based on reliable and valid data. The importance of reliability test lies in ensuring the consistency and stability of questionnaire data. High reliability means that the questionnaire can obtain consistent results at different time points or under different conditions, thereby ensuring the reliability of the measurement results. If the reliability of the questionnaire is low, it may mean that there are problems with the questionnaire design or that the respondents' answers are unstable, which will affect the accuracy of data analysis and the credibility of the conclusions. Therefore, through reliability test, the quality of questionnaire data can be effectively evaluated and improved. This application uses Cronbach's Alpha to perform reliability test, and Cronbach's Alpha is used to measure the consistency between multiple questions. The calculation formula of Cronbach's Alpha is:

[0098]

[0099] Where α is Cronbach's α coefficient; N is the number of questions; is the average covariance between items; It is the average value of the variance between items. The average covariance between items reflects the consistency between items. If the covariance between items is high, it means that the items are highly consistent. This means that when respondents score high on one item, they usually score high on another item (or the same if they score low). This situation usually indicates that the content or concepts measured by the two items are related. The average value of the item variance reflects the variability of the response to each item, which refers to the difference in scores of a item among all respondents. If the item variance is large, it means that there are large differences in the respondents' answers to this item; a small variance means that the respondents' answers to this item are more consistent. Among them, the value of Cronbach's alpha coefficient is between 0 and 1, and the higher the value, the better the reliability. Generally speaking, an alpha value exceeding 0.7 is considered to be acceptable reliability.

[0100] Step S33, use the formula X=ΛF+∈ to perform validity test on all questions in the target questionnaire after reliability test to obtain the tested target questionnaire; wherein X is the observed variable vector, Λ is the factor loading matrix, F is the common factor vector, and ∈ is the error vector.

[0101] Specifically, the purpose of validity testing is to ensure that the questionnaire can accurately measure the expected characteristics and concepts. High efficiency means that the questionnaire is not only superficially reasonable, but also can truly reflect the actual situation of the measured object. If the validity of the questionnaire is low, it means that the measurement results may be biased and cannot truly reflect the characteristics and opinions of the respondents, which will directly affect the accuracy of the survey questionnaire data analysis and the scientific nature of the research conclusions. Through validity testing, the rationality of the questionnaire design and the accuracy of the measurement results can be effectively verified, thereby improving the credibility and scientific nature of the questionnaire data analysis. This application adopts structural validity testing and verifies the rationality of the questionnaire structure through factor analysis methods. Factor analysis is used to identify and verify latent variables (factors) in questionnaires and evaluate the relationship between each question and these latent variables. Through factor analysis, it can be confirmed whether the questions in the questionnaire reasonably reflect the underlying structure and whether these structures conform to the expected theoretical model.

[0102] Specifically, X is the observed variable vector. In the analysis of questionnaires, the observed variables are usually the scores of each question item (or question item) in the questionnaire. For example, if there are 10 questions in the questionnaire, each question has several options, and the respondents' answers can be quantified into scores. Assuming that the questionnaire has 3 questions and 5 respondents, the observed variable vector may be a 5×3 matrix, indicating the scores of each respondent on these 3 questions.

[0103] Λ is the factor loading matrix, which indicates the correlation between each question and the latent factor. It reflects the extent to which a question can explain a latent factor. If there are 3 questions and 2 latent factors (that is, the number of factors is 2), then the factor loading matrix Λ is a 3×2 matrix.

[0104] F is a common factor vector. Common factors refer to latent variables behind the questionnaire data. Latent variables are not directly measured, but inferred through the common variability of multiple questions. If two latent factors in the questionnaire are identified through factor analysis, and there are 5 respondents, then F is a 5×2 matrix. Each column represents a factor, and each row represents a respondent's score on this factor.

[0105] ∈ is the error vector, which represents the part of the observed score that cannot be explained by the common factors. It includes measurement errors and topic-specific variations that are unrelated to the factors. In a 5×3 observed variable matrix, the error vector ∈ is also a 5×3 matrix. Each element represents the part of a particular respondent's score on a topic that cannot be explained by the latent factor.

[0106] As an optional implementation, step S31 specifically includes:

[0107] Step S311, using the mean filling method to process the missing values ​​of the answers to all questions in the digitized target questionnaire, to obtain the target questionnaire after the missing value processing.

[0108] Specifically, in the questionnaire, respondents may skip certain questions or omit them during the filling process, resulting in missing values ​​in the data set. In order to ensure the integrity and consistency of the data, the present invention uses the mean filling method to process the missing values. Specifically, the mean of the feature is used to fill the missing values. This method is simple and effective and is suitable for situations where the proportion of missing values ​​is small and the data distribution is relatively symmetrical. The formula for mean filling is as follows:

[0109]

[0110] Among them, x b represents the value of the feature after filling, x g represents all non-missing values, and q represents the number of non-missing values.

[0111] Step S312, using a box plot method to perform outlier detection on the answers to all questions in the target questionnaire after missing value processing, to obtain the target questionnaire after outlier detection.

[0112] The formula IQR=Q3-Q1 is used to calculate the interquartile range of all answers in the target questionnaire after missing value processing; wherein, Q1 is the first quartile, the position of the 25th percentile in the data set; Q3 is the third quartile, the position of the 75th percentile in the data set; IQR is the interquartile range; based on the quartiles and the interquartile range, it is determined whether each answer in the target questionnaire after missing value processing meets the preset range, and the answers that do not meet the preset range are eliminated, that is, if x1<Q1-1.5×IQR or x1>Q3+1.5×IQR, then the value is an abnormal value, and x1 represents the data in the target questionnaire after missing value processing.

[0113] Specifically, outliers refer to values ​​that are significantly different from other data points, which may be caused by respondents' mistakes, data entry errors, or extreme situations. In questionnaire data, outliers can significantly affect the results of statistical analysis and model training, so they need to be detected and processed. This application uses the box plot method for outlier detection.

[0114] Assume the dataset is: [10,12,14,15,18,21,22,24,30,35].

[0115] Since the questions are questionnaires, for specific data-type questions, the data are concrete, and for range or text-type questionnaire questions, they are coded and re-assigned to obtain data sets.

[0116] 1) Calculate Q1 and Q3:

[0117] For the following data set: [10, 12, 14, 15, 18, 21, 22, 24, 30, 35], the 25th percentile is 14, so Q1=14; the third quartile is 24, so Q3=24.

[0118] 2) Calculate IQR:

[0119] IQR=Q3-Q1=24-14=10.

[0120] 3) Calculate the lower and upper limits:

[0121] LowerBound=14-1.5×10=14-15=-1

[0122] UpperBound=24+1.5×10=24+15=39

[0123] 4) Detecting outliers:

[0124] Since there are no values ​​less than -1 or greater than 39 in the dataset, there are no outliers in this dataset.

[0125] Step S313, standardize the answers to all questions in the target questionnaire after outlier detection to obtain a standardized target questionnaire.

[0126] Specifically, different questions in the questionnaire may involve different dimensions and scales, such as age, income, satisfaction score, etc. Standardization is to adjust the data to a distribution with a mean of 0 and a standard deviation of 1, so that data with different features are comparable. Standardization eliminates the dimensional differences between different features, allowing each feature to be compared on the same scale, which helps to accelerate the convergence of optimization algorithms such as gradient descent and improve model performance. The formula for standardization is as follows:

[0127]

[0128] Among them, x′ represents the standardized value, x2 represents the original value of the data in the target questionnaire after outlier detection, μ represents the mean, and σ represents the standard deviation.

[0129] Step S314, performing normalization processing on the standardized target survey questionnaire to obtain a pre-processed target survey questionnaire.

[0130] Specifically, normalization is to scale the data to the range of [0,1] so that data of different features have the same scale. Different features of questionnaire data may have different ranges, for example, age ranges from 0-100, while satisfaction scores range from 1-5. Normalization helps improve the training effect and prediction performance of the model. The formula of the normalization method is as follows:

[0131]

[0132] Among them, z p represents the normalized target survey questionnaire data, z represents the standardized target survey questionnaire data, min(z) and max(z) represent the minimum and maximum values ​​of the corresponding target survey questionnaire data, respectively.

[0133] As an optional implementation, step S4 specifically includes:

[0134] Step S41, performing a second category determination on all questions in the verified target questionnaire to determine the second category to which the question belongs; the second category is a regression problem or a classification problem.

[0135] Step S42, when it is determined that the second category to which the question belongs is a regression problem, the Pearson correlation coefficient is used to calculate the correlation between the question score and the target variable, and the questions whose correlation meets the first preset threshold are selected; the target variable is the sample value of the target questionnaire output result.

[0136] Step S43, when it is determined that the second category to which the question belongs is a classification question, the chi-square test is used to calculate the chi-square statistic between the question and the target variable, and the questions whose chi-square statistic meets the second preset threshold are selected to complete feature selection and obtain the target questionnaire after feature selection.

[0137] Specifically, feature selection of the tested target questionnaire is a key step to ensure the effectiveness of data analysis and modeling. This application uses feature selection to extract and select features with high correlation with the target variable, remove redundant and irrelevant features, thereby enhancing the generalization ability and predictive performance of the model. Among them, questionnaire data usually contains a large number of features, and only a part of these features are closely related to the target variable. Through feature selection, effective features can be extracted and generated, and redundant and irrelevant features can be removed, thereby improving the predictive performance of the model. Accurate feature selection enables the model to better capture patterns and regularities in the data and improve the accuracy of the analysis results.

[0138] Specifically, this application uses the following two methods for regression problems (i.e. continuous problems, such as ratings, overall satisfaction) and classification problems (i.e. categorical problems, such as multiple-choice questions, yes / no questions, etc.):

[0139] 1) Pearson correlation coefficient

[0140] The Pearson correlation coefficient is used to measure the linear correlation between continuous features and target variables. For example, by calculating the correlation coefficient between each questionnaire question score and the target variable (such as satisfaction score), questions that are highly correlated with the target variable can be screened out. The formula is as follows:

[0141]

[0142] Among them, r represents the Pearson correlation coefficient, U i and V i Represent the data feature value and target variable value respectively, and Represent the mean of data features and target variables respectively, and p represents the number of samples. Through this method, questions that have a significant impact on the target variable are identified, thereby providing more effective data features for the model.

[0143] The Pearson correlation coefficient method is applicable to regression problems. For example, if the survey questionnaire contains multiple ratings expressed in numerical form (such as ratings from 1 to 10), the Pearson correlation coefficient can be used to evaluate the linear correlation between these ratings and the target variable (such as overall satisfaction).

[0144] 2) Chi-square test

[0145] The chi-square test is used to measure the independence between categorical features and the target variable and is applicable to classification problems. By calculating the chi-square statistic between each categorical variable (such as the options corresponding to the multiple-choice questions in the questionnaire) and the target variable, the categorical features that have a significant impact on the target variable can be screened out. The formula is as follows:

[0146]

[0147] Among them, χ 2 represents the chi-square statistic, O i represents the observation frequency, E i In this way, we can identify the multiple-choice options that have a significant impact on the target variable, thereby optimizing the model input.

[0148] The chi-square test is applicable to classification problems. By calculating the independence of categorical features and the target variable (usually also categorical, such as satisfied / unsatisfied), categorical features that have a significant impact on the target variable can be identified, which helps to remove option features that are not related to the target variable.

[0149] After completing data preprocessing and feature selection, the next stage is the data integration stage. This application has selected an adaptable modeling method that can simplify the process and ensure consistency and efficiency in the analysis of survey data from different types of questionnaires. Considering the diversity and complexity of survey data from different questionnaires, the random forest model, as a modeling method with strong adaptability and stable performance, was selected for the final modeling of this application. Data integration is to combine the processed feature data into a structured data set for subsequent model training and evaluation.

[0150] As an optional implementation, step S5 is specifically as follows:

[0151] 1) Feature combination: All feature data after feature selection are combined into a complete data matrix by column. Assuming there are m samples and n selected features, the dimension of the integrated data matrix X is m×n.

[0152] 2) Target variable: Align the corresponding target variable y with the feature data matrix X. The target variable can be a categorical label (such as satisfaction level) or a continuous variable (such as a rating).

[0153] 3) Data segmentation: The integrated data set is divided into a training set and a test set. The training set is used for model training, and the test set is used to evaluate model performance. The segmentation ratio is 80% training set and 20% test set.

[0154] As an optional implementation, in step S6, the training process of the questionnaire result prediction model specifically includes:

[0155] Step S61, input the training samples into the random forest model to obtain the sample prediction value of the questionnaire results.

[0156] Step S62, constructing a loss function according to the sample prediction value of the questionnaire result and the sample value of the questionnaire result, and adjusting the model parameters of the random forest model according to the loss function to obtain a questionnaire result prediction model.

[0157] Specifically, this application uses a random forest model for modeling. Random forest is an integrated learning method that improves the generalization ability and stability of the model by building multiple decision trees. The following are the detailed application steps of the random forest model:

[0158] 1. Feature input: Each question or option in the questionnaire forms a feature. For example, the answer scores of multiple questions, the selected options, etc. can all be used as data features, including continuous features and categorical features. Among them, continuous features: for example, questions with a score of 1-5; categorical features: for example, the options in the multiple-choice questions (which can be converted into numerical form through coding).

[0159] 2. Output results:

[0160] Classification output: Finally, a category label is output, indicating the classification result of a certain sample.

[0161] Regression output: Output a continuous value representing the predicted score or value for a sample.

[0162] 3. Data sampling: Perform multiple sampling with replacement on the training data to generate multiple subsets. Each subset has the same size as the original training set but contains repeated samples.

[0163] 4. Decision tree training: Train a decision tree for each subset. Each tree randomly selects features to split at each split, thereby increasing the diversity of the model. The training process is as follows:

[0164] 4.1 Randomly select feature subsets: When each node is split, a subset (such as a combination of variables after one-hot encoding) is randomly selected from the features in the questionnaire data, and then the best feature is selected from this subset for splitting. This randomness increases the diversity of the model, allowing different trees to produce different split structures when processing the same data.

[0165] 4.2 Evaluating Split Points

[0166] Depending on the task, different formulas are used to evaluate the split point:

[0167] 4.2.1 Classification problem: using the Gini index for measurement.

[0168]

[0169] Among them, Gini represents the Gini index; D represents the data set of the current node, C is the number of categories, and h l is the proportion of samples in the lth class in the data set.

[0170] Gini index after node splitting:

[0171]

[0172] Among them, D1 and D2 are two subsets after splitting according to a certain feature, |D| is the total number of samples in the original data set, and |D1| and |D2| are the number of samples in the subsets respectively.

[0173] 4.2.2 Regression problem: Use mean square error for measurement.

[0174]

[0175] Where n is the number of samples, s is a is the true value of the ath sample, is the predicted value (usually the sample mean of the split node).

[0176] 4.3 Select the best split point:

[0177] For classification problems, select the features and splitting points that minimize the Gini index after splitting; for regression problems, select the features and splitting points that minimize the mean square error after splitting.

[0178] 4.4 Splitting node: According to the selected optimal splitting point, the data of the current node is divided into two sub-nodes, corresponding to the two groups of data subsets under the feature splitting condition.

[0179] 4.5 Recursive subtree construction: For each generated child node, repeat the above splitting process until the stopping condition (such as the minimum number of samples, maximum depth or purity, etc.) is met. The tree construction process is recursive, and each time a node meets the stopping condition, the node becomes a leaf node.

[0180] 4.6 Tree completion:

[0181] When all nodes cannot be split any further, the tree training is complete. The output of the leaf node is based on the sample data it contains:

[0182] Classification problem: The leaf node output is the most common category in the current node sample.

[0183] Regression problem: The output of a leaf node is the mean of the current node sample.

[0184] 5. Prediction Integration:

[0185] When predicting, the prediction results of all decision trees are integrated. For classification problems, the majority voting method is used; for regression problems, the average method is used.

[0186] Classification Problem: Predicting Outcomes It is the majority vote of the prediction results of each decision tree T.

[0187]

[0188] Regression Problem: Predicting an Outcome For each decision tree T d The average of the prediction results.

[0189]

[0190] Where f is the number of decision trees, T d (x) is the prediction result of the dth decision tree for the data feature x.

[0191] The questionnaire output result prediction model can be used for both classification and regression problems, but its specific use depends on the goal and data type of the questionnaire. Normally, a mature questionnaire will set different types of questions, and the questions that need to be investigated will also consider multiple angles.

[0192] Classification problem: The goal of the questionnaire question is to predict the category results, such as "Is the user satisfied?", "Will the customer buy again?", etc.; Regression problem: The goal of the questionnaire question is to predict continuous values, such as predicting customer satisfaction scores, user retention time, etc. Random forests can also integrate the predicted values ​​of multiple decision trees to get the average value, thereby realizing the prediction of continuous target variables. Therefore, the questionnaire output result prediction model does not simply classify a type of question, but analyzes the goal of the questionnaire question. The following are a few examples to illustrate the specific application steps of the questionnaire output result prediction model in classification and regression problems.

[0193] 1. Regression problem:

[0194] 1) How much time do you spend online per week?

[0195] A. Less than 7 hours B. 7-14 hours C. 14-20 hours D. More than 20 hours E. No Internet access

[0196] 2) What is your attitude towards the trustworthiness of AI learning tools in answering questions?

[0197] A. Very high B. Average C. Don't believe

[0198] 3) What is your overall satisfaction rating for the online education platform?

[0199] 1. Very dissatisfied 2. Dissatisfied 3. Average 4. Satisfied 5. Very satisfied

[0200] Output:

[0201] Predicted overall satisfaction score: For each respondent, the questionnaire output prediction model outputs a continuous value between 1 and 5, indicating the predicted satisfaction score for the online education platform.

[0202] Data characteristics:

[0203] 1) Internet access duration (after encoding conversion):

[0204] A:1B:2C:3D:4E:0

[0205] 2) Attitude towards the trustworthiness of AI tools (after coding conversion):

[0206] A:1B:2C:0

[0207] 3) Other possible features

[0208] Gender: 0 (male), 1 (female)

[0209] Grade: 1(Freshman), 2(Sophomore), 3(Junior), 4(Senior)

[0210] Assume the following data:

[0211] Table 2 Data table

[0212]

[0213] The output of the questionnaire output prediction model is as follows:

[0214] For a new respondent, the data characteristics are: Internet time 14-20 hours (coded as 3), average credibility of tool A (coded as 2), gender is male (coded as 0), and grade is sophomore (coded as 2).

[0215] The output of the questionnaire output prediction model is: the satisfaction score is 3.5.

[0216] 2. Classification problem:

[0217] Suppose the goal of the survey is to predict students' views on the possibility that artificial intelligence learning tools will replace teachers in future education based on some of their characteristics. The following questions can be used as data features:

[0218] 1. Gender 2. Major 3. Grade 4. Personality 5. Time online 6. Have you used learning software tools? 7. Do you want to obtain learning resources from universities across the country? 8. Have teachers recommended learning software? 9. Do you want to use artificial intelligence learning tools to help you complete your homework?

[0219] Assume the following data:

[0220] Table 3 Data table

[0221]

[0222] The output of the questionnaire output prediction model is:

[0223] For a new student, the characteristics are: male gender (coded as 0), major in science and engineering (coded as 1), sophomore (coded as 1), firm personality (coded as 3), 7-14 hours of Internet time (coded as 1), never used learning software (coded as 0), do not want to obtain learning resources (coded as 0), no idea of ​​helping to complete homework (coded as 0). The output of the questionnaire output prediction model is: the view on artificial intelligence learning tools replacing teachers in future education is "unclear".

[0224] Here is another classification problem:

[0225] Assuming the goal is to predict students' Internet usage purposes based on some of their characteristics, the target variables according to the questionnaire can be: learning, looking up information (1) browsing news (2) entertainment games (3) chatting and making friends (4) resource downloading (5);

[0226] For a new student, the characteristics are: gender male (coded as 0), major science and engineering (coded as 1), grade sophomore (coded as 1), firm personality (coded as 3), online time 7-14 hours (coded as 1), never used learning software (coded as 0), do not want to obtain learning resources (coded as 0), the teacher has not recommended learning software (coded as 0). The output of the questionnaire output prediction model is: the purpose of going online is "entertainment games" (3).

[0227] As an optional implementation, after step S6, the method further includes: analyzing and visualizing the predicted value of the survey results of the questionnaire. This step aims to interpret the prediction results of the model and present the analysis results to the user in an intuitive form.

[0228] 1) Basic frequency analysis: Count the frequency of each option or answer to understand the distribution of data and the overall tendency of respondents. Frequency distribution tables and histograms can be used to intuitively display data distribution.

[0229] 2) Cross-analysis: Identify different responses to a question in a questionnaire from different groups, discover hidden patterns and trends, verify research hypotheses, and provide strong support for decision making.

[0230] 3) Result visualization: Visualize the analysis results so that users can more intuitively understand the performance of the data and model. Visual charts such as feature importance graphs, confusion matrices, predicted value vs. actual value comparison graphs, ROC curves, and AUC values ​​are used to display the questionnaire and the questionnaire questions reflected by the established model. The following are some visualization diagrams.

[0231] Basic information categories:

[0232] The questionnaire focuses on the educational background of the subjects, covering four subject categories: literature, history, science and engineering, management and art. The results of the questionnaire are visualized as follows: Figure 3 As shown, it can be seen that science and engineering students account for 62%, and business and management students account for 20%, ranking second. This shows that the current distribution of questionnaires is both targeted and universal.

[0233] Prospect Category:

[0234] To explore the aspects that college students are most concerned about when it comes to learning software, the results of the questionnaire are visualized as follows: Figure 4 As shown, the subjects are most concerned about the learning effect, with 84% of the subjects choosing this option; 73% of the subjects are concerned about whether the learning resources are abundant; the number of samples concerned about the learning expenses is the smallest, but it is also close to 50%, indicating that each option of this question involves the issues that college students are most concerned about.

[0235] Cross-analysis:

[0236] The four dimensions (or other dimensions) of major, grade, online time and online method in the personal basic information category are described. The diagram of the questionnaire results is shown below: Figure 5 shown.

[0237] Classification problem, concrete label:

[0238] The survey questionnaire result data or labels are visualized, and the obtained survey questionnaire result visualization diagram is as follows Figure 6 shown.

[0239] Beneficial effects of this application:

[0240] 1) Combining data preprocessing with feature selection optimizes the data quality and the construction of data features, effectively improving the predictive performance of the model.

[0241] 2) A questionnaire output prediction model that comprehensively considers multiple variables was constructed. The model can fully extract and utilize the complex relationships and features in the questionnaire data, especially when dealing with high-dimensional and multi-category data. The variables that have the greatest impact on the analysis results can be identified, further improving the interpretability of the model.

[0242] 3) Use box plots to detect abnormal data and fill in missing values ​​to ensure data integrity and consistency. Normalization eliminates the dimensional differences between different features and optimizes the training effect of the model.

[0243] 4) The questionnaire output result prediction model trained based on the random forest model can accurately capture the patterns and trends in the questionnaire data, and shows good prediction performance and adaptability in different types of questionnaires.

[0244] 5) Through visual analysis results, the predictive ability and feature importance of the model are intuitively displayed, providing a scientific basis for questionnaire design and result interpretation.

[0245] Based on the same inventive concept, the present application also provides a questionnaire result determination system, which is based on the questionnaire result determination method of the questionnaire, and the questionnaire result determination system includes: a target questionnaire determination unit 1, a digitized target questionnaire determination unit 2, a verified target questionnaire determination unit 3, a feature-selected target questionnaire determination unit 4, a data feature determination unit 5 and a questionnaire output result determination unit 6.

[0246] The target questionnaire determination unit 1 is used to obtain the filled questionnaires and eliminate invalid questionnaires from the filled questionnaires to obtain the target questionnaires; invalid questionnaires include questionnaires with contradictory answers to similar questions and questionnaires with ambiguous items selected for all questions.

[0247] The numerically quantified target questionnaire determination unit 2 is used to numerically encode the answers to all questions in the target questionnaire to obtain the numerically quantified target questionnaire.

[0248] The verified target questionnaire determination unit 3 is used to perform reliability test and validity test on the digitized target questionnaire to obtain the verified target questionnaire.

[0249] The target questionnaire determination unit 4 after feature selection is used to perform feature selection on all questions in the tested target questionnaire by using the Pearson correlation coefficient and the chi-square test to obtain the target questionnaire after feature selection.

[0250] The data feature determination unit 5 is used to perform data integration on the target questionnaire after feature selection to obtain data features of the target questionnaire.

[0251] The questionnaire output result determination unit 6 is used to input the data features of the target questionnaire into the questionnaire survey result prediction model to obtain the predicted value of the questionnaire survey result; the questionnaire survey result prediction model is obtained by training the random forest model with training samples; the training samples include the sample values ​​of the data features of the questionnaire and the sample values ​​of the questionnaire survey result.

[0252] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0253] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0254] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0255] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, 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.

[0256] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for determining a survey result of a questionnaire, characterized in that: The method for determining the survey results of the questionnaire includes: Obtain the completed questionnaires and eliminate invalid questionnaires from the completed questionnaires to obtain the target questionnaire; invalid questionnaires include questionnaires with contradictory answers to similar questions and questionnaires with ambiguous items selected for all questions; Numerically encode the answers to all questions in the target questionnaire to obtain the numerically encoded target questionnaire; Conduct reliability and validity tests on the digitized target questionnaire to obtain the tested target questionnaire; Using the Pearson correlation coefficient and chi-square test, feature selection is performed on all questions in the target questionnaire after the test to obtain the target questionnaire after feature selection; Perform data integration on the target questionnaire after feature selection to obtain the data features of the target questionnaire; The data features of the target questionnaire are input into the questionnaire survey result prediction model to obtain the predicted value of the questionnaire survey result; the questionnaire survey result prediction model is obtained by training the random forest model with training samples; the training samples include the sample values ​​of the data features of the questionnaire and the sample values ​​of the questionnaire survey result.

2. The method for determining the survey result of a questionnaire according to claim 1, characterized in that: The answers to all questions in the target questionnaire are numerically coded to obtain the numerically coded target questionnaire, which specifically includes: Performing a first category determination on all questions in the target questionnaire to determine the first category to which each question belongs; the first category is a single-choice question or a multiple-choice question; Based on the first category to which each question belongs, the corresponding numerical coding method is selected, and the answers corresponding to each question are numerically encoded according to the selected numerical coding method to obtain the numerically encoded target questionnaire; when the first category to which the question belongs is a single-choice question, the answer corresponding to the question is label-encoded; when the first category to which the question belongs is a multiple-choice question, the answer corresponding to the question is encoded 0-1.

3. The method for determining the survey result of a questionnaire according to claim 2, characterized in that: The expression of label encoding is: y i =i,i∈{1,2,...,n}; Among them, y i Indicates the numerical label corresponding to the i-th option; n indicates the total number of options contained in a single-choice question; The expression for 0-1 encoding is: Among them, y ij Represents the binary code value of the jth option of the kth question; m is the total number of options contained in a multiple-choice question.

4. The method for determining the survey result of a questionnaire according to claim 1, characterized in that: The reliability and validity tests are carried out on the numerical target survey questionnaire to obtain the tested target survey questionnaire, which specifically includes: Preprocessing the digitized target questionnaire to obtain a preprocessed target questionnaire; The reliability of all questions in the preprocessed target questionnaire was tested using Cronbach's α coefficient to obtain the target questionnaire after reliability test. The formula X=ΛF+∈ is used to perform validity test on all questions in the target questionnaire after reliability test to obtain the target questionnaire after test; where X is the observed variable vector, Λ is the factor loading matrix, F is the common factor vector, and ∈ is the error vector.

5. The method for determining the survey result of a questionnaire according to claim 4, characterized in that: The calculation formula of Cronbach's alpha coefficient is: Where α is Cronbach's α coefficient; N is the number of questions; is the average covariance between items; is the average of the variances between items.

6. The method for determining the survey result of a questionnaire according to claim 4, characterized in that: The digitized target questionnaire is preprocessed to obtain the preprocessed target questionnaire, which specifically includes: The mean filling method is used to process the missing values ​​of the answers to all questions in the digitized target questionnaire to obtain the target questionnaire after missing value processing; The box plot method is used to detect outliers in the answers to all questions in the target questionnaire after missing value processing, and the target questionnaire after outlier detection is obtained; Standardize the answers to all questions in the target questionnaire after outlier detection to obtain a standardized target questionnaire; The standardized target questionnaire is normalized to obtain the preprocessed target questionnaire.

7. The method for determining the survey result of a questionnaire according to claim 6, characterized in that: The box plot method is used to detect outliers in all answers of the target questionnaire after missing value processing, including: The formula IQR = Q3-Q1 is used to calculate the interquartile range of all answers in the target questionnaire after missing value processing; where Q1 is the first quartile; Q3 is the third quartile; IQR is the interquartile range; Based on the quartiles and the interquartile range, it is determined whether each answer in the target questionnaire after the missing value processing meets the preset range, and the answers that do not meet the preset range are eliminated.

8. The method for determining the survey result of a questionnaire according to claim 1, characterized in that: Using the Pearson correlation coefficient and chi-square test, feature selection is performed on all questions in the target questionnaire after the test to obtain the target questionnaire after feature selection, which specifically includes: Performing a second category determination on all questions in the tested target questionnaire to determine the second category to which the question belongs; the second category is a regression problem or a classification problem; When it is determined that the second category to which the question belongs is a regression problem, the correlation between the question score and the target variable is calculated using the Pearson correlation coefficient, and the question whose correlation meets the first preset threshold is selected; the target variable is a sample value of the output result of the target questionnaire; When it is determined that the second category to which the question belongs is a classification problem, the chi-square test is used to calculate the chi-square statistic between the question and the target variable, and the questions whose chi-square statistic meets the second preset threshold are selected to complete the feature selection and obtain the target questionnaire after feature selection.

9. The method for determining the survey result of a questionnaire according to claim 1, characterized in that: The training process of the questionnaire result prediction model specifically includes: Input the training samples into the random forest model to obtain sample prediction values ​​of the questionnaire results; A loss function is constructed according to the sample prediction value of the questionnaire results and the sample value of the questionnaire results, and the model parameters of the random forest model are adjusted according to the loss function to obtain a questionnaire result prediction model.

10. A questionnaire result determination system, characterized in that: The questionnaire result determination system is based on the questionnaire result determination method according to any one of claims 1 to 9, and the questionnaire result determination system comprises: The target questionnaire determination unit is used to obtain the filled questionnaires and eliminate invalid questionnaires from the filled questionnaires to obtain the target questionnaires; invalid questionnaires include questionnaires with contradictory answers to similar questions and questionnaires with ambiguous items selected for all questions; A unit for determining a target questionnaire after digitization is used to digitize the answers to all questions in the target questionnaire to obtain a target questionnaire after digitization; The unit for determining the target survey questionnaire after verification is used to conduct reliability and validity tests on the digitized target survey questionnaire to obtain the target survey questionnaire after verification; The target questionnaire determination unit after feature selection is used to perform feature selection on all questions in the target questionnaire after the test by using the Pearson correlation coefficient and the chi-square test to obtain the target questionnaire after feature selection; A data feature determination unit is used to integrate the data of the target survey questionnaire after feature selection to obtain the data features of the target survey questionnaire; The questionnaire output result determination unit is used to input the data features of the target questionnaire into the questionnaire survey result prediction model to obtain the predicted value of the questionnaire survey result; the questionnaire survey result prediction model is obtained by training the random forest model with training samples; the training samples include the sample values ​​of the data features of the questionnaire and the sample values ​​of the questionnaire survey result.

Images