Cigarette case design optimization method based on consumer sensory evaluation

Abstract

The invention discloses a cigarette case design optimization method based on consumer sensory evaluation. The method comprises the following steps: obtaining consumer sensory evaluation data; performing data preprocessing and preparation, and performing standardization processing on scores of different sensory dimensions; performing data separation and recombination based on design element classification and price group labeling; statistical analysis and decision tree model construction; the analysis result is visually displayed; and carrying out comprehensive analysis and result integration on the basic data and the visual result. According to the cigarette case design optimization method based on consumer sensory evaluation, through comprehensive collection of consumer sensory evaluation data and deep analysis of preferences and demands of consumers on each design element of the cigarette case, a scientific data basis is provided for cigarette case design, and optimization of a design scheme and reduction of design risks are facilitated; by means of visualization and comprehensive analysis, market dynamics can be grasped in time, and a design strategy can be adjusted; comprehensive and systematic data support is provided for cigarette case design decision, and the market adaptability and consumer satisfaction of cigarette case design are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of cigarette box design, and more particularly to a cigarette box design optimization method based on consumer sensory evaluation. Background Art

[0002] In the field of cigarette packaging design, traditional design methods often rely heavily on the designer's personal experience and subjective judgment, lacking in-depth exploration and systematic analysis of consumer sensory experiences and market feedback. This results in cigarette packaging designs that may not accurately meet consumer preferences and needs, which in turn affects the product's market competitiveness. Consumers' sensory evaluations of cigarette packaging cover multiple dimensions, such as the simplicity of the appearance, the coordination of color matching, and the style of the font design. Different consumers have significant preferences in these areas, and it is difficult to fully capture them through simple market research. Existing research methods may only collect superficial information and cannot deeply analyze the factors that influence consumer preferences, making it difficult for cigarette packaging designs to accurately match consumers' diverse needs.

[0003] In the past, cigarette packaging designers lacked scientific data support when determining design directions and elements. Design teams often chose design solutions based on experience, without fully considering the impact of different design elements on consumers and the relationships between them. This data-free decision-making method increases the risk of design failure and may cause cigarette packaging designs to fail to attract the target consumer group, affecting product sales. With the continuous changes in the market environment and the updating of consumer aesthetic concepts, cigarette packaging designs need to be adjusted in a timely manner. However, due to the lack of an effective data collection and analysis system, companies find it difficult to quickly and accurately understand market dynamics and changing trends in consumer demand. As a result, cigarette packaging designs lag behind in adapting to market changes and are unable to launch products that meet market demand in a timely manner.

[0004] Therefore, there is an urgent need for a cigarette box design optimization method based on consumer sensory evaluation. Summary of the Invention

[0005] The purpose of the present invention is to provide a cigarette box design optimization method based on consumer sensory evaluation to solve the problems in the above-mentioned prior art. It can deeply analyze consumers' preferences and needs for various design elements of cigarette boxes, provide scientific data basis for cigarette box design, help optimize design schemes and reduce design risks.

[0006] The present invention provides a cigarette box design optimization method based on consumer sensory evaluation, which includes:

[0007] Obtain consumer sensory evaluation data;

[0008] Data preprocessing and preparation, and standardization of scores for different sensory dimensions;

[0009] Separate and reorganize data based on design element classification and price group labeling;

[0010] Conduct statistical analysis and build a decision tree model based on the reorganized data;

[0011] Visually display the statistical analysis results and the analysis results of the decision tree model;

[0012] Comprehensive analysis and result integration are performed on basic data and visualization results, wherein the basic data is data related to statistical analysis and decision tree model.

[0013] In the above-mentioned method for optimizing cigarette box design based on consumer sensory evaluation, preferably, obtaining consumer sensory evaluation data includes:

[0014] Using the pandas library in Python to read data, first try Encoding read data, if unicode Decode Error occurs, try to read it with gbk encoding; if gbk encoding reading also fails, try to read it with Code reading,

[0015] The data preprocessing and preparation, and the standardization of scores for different sensory dimensions, include:

[0016] Data integrity check;

[0017] Data consistency check;

[0018] Scoring standardization;

[0019] Data cleaning and denoising.

[0020] In the above-mentioned method for optimizing cigarette box design based on consumer sensory evaluation, preferably, the data integrity check includes: for missing numerical data columns, filling them with statistical values or interpolation methods; for missing non-numerical data columns, filling them according to the mode or business logic, or deleting the corresponding columns.

[0021] The data consistency check includes: checking whether there are inconsistencies in the data,

[0022] The score standardization process includes: standardizing the scores of different sensory dimensions using a linear transformation formula Standardization is performed, where represents the original score, Indicates the minimum value of the original score of this dimension. Indicates the maximum value of the original score of this dimension. represents the normalized score,

[0023] The data cleaning and denoising includes: identifying and processing outliers in the data by drawing box plots and calculating the interquartile range method.

[0024] In the above-mentioned cigarette box design optimization method based on consumer sensory evaluation, preferably, the data separation and reorganization based on design element classification and price group labeling includes:

[0025] Design element classification and mapping establishment;

[0026] Importance column identifier definition;

[0027] Price group mapping and data annotation;

[0028] Data reorganization and structuring processing.

[0029] In the above-mentioned cigarette box design optimization method based on consumer sensory evaluation, preferably, the classification and mapping of design elements include:

[0030] Based on data column name With keywords , according to the Levenshtein distance Calculate the edit distance between two strings and set the distance threshold ,like , then the data column is considered to belong to the corresponding design element category.

[0031] The definition of importance column identifiers includes: determining identifiers based on a deep understanding of questionnaire design and data structure to accurately identify columns in the data that are related to the importance scores of each design element,

[0032] The price group mapping and data labeling includes: extracting price column information from the data based on a pre-set price group mapping relationship. First, by traversing the data column names, searching for a column containing a price keyword as a price column; then, processing each data in the price column, extracting the first character using a string truncation operation, converting the first character into a specific price group name based on the price group mapping relationship, and adding a price group column to the data set for labeling;

[0033] The data reorganization and structuring processing includes: reorganizing the data based on design element classification and price group labeling, specifically including: grouping data columns of the same category based on design element category, aggregating data columns of the same category together, and stratifying them according to price groups; at the same time, associating and integrating importance score data with corresponding design element data to construct a structured data table.

[0034] In the above-mentioned method for optimizing cigarette box design based on consumer sensory evaluation, preferably, the statistical analysis and decision tree model construction based on the reorganized data include:

[0035] Group data preparation;

[0036] Calculation of mean scores for design element categories;

[0037] Calculation of the mean importance of design elements;

[0038] Preparation for decision tree model construction;

[0039] Decision tree model construction and evaluation.

[0040] In the above-mentioned method for optimizing cigarette box design based on consumer sensory evaluation, preferably, the grouping data preparation includes:

[0041] For each grouped data, first check the integrity and validity of the data to see if there are too many missing values or outliers that affect the analysis. If so, process them according to the data preprocessing and preparation methods.

[0042] The calculation of the mean value of the design element category score includes: assuming that a certain design element category is , the column set contained in this category is , the sample size is , use the following formula to calculate the mean score of this category :

[0043]

[0044] in, Indicates the Liedi The score of the samples,

[0045] The calculation of the mean importance of the design elements includes: assuming that the importance of a certain design element is , the following formula is used to calculate the mean importance of design factors:

[0046]

[0047] in, For the The score of the samples in the importance column,

[0048] The decision tree model construction preparation includes:

[0049] Target variable selection: Select the design category corresponding to the highest value from the mean score of each design factor category as the target variable of the decision tree analysis, denoted as , in the target category In the above example, we further select the specific design features with the highest scores and record them as ;

[0050] Classification target definition: based on design features The average rating of As the threshold, define the "high score" target variable , determined using the following formula Value:

[0051]

[0052] in, Indicates that the sample has design characteristics Ratings on

[0053] Feature selection: using the target category The target features in As features, the scores of other design elements other than , for each category , taking the columns it contains as features,

[0054] The decision tree model construction and evaluation includes:

[0055] Data preprocessing: Before building the decision tree model, the selected feature data is preprocessed. For missing values, multiple imputation methods are used for data processing;

[0056] Model training: Using CART algorithm, using Python The CART algorithm performs feature selection and tree construction based on the Gini index. The Gini index is used to measure the impurity of the data set. The formula of the Gini index is:

[0057]

[0058] in, represents a dataset, represents the number of categories in the dataset, Indicates the The proportion of class samples in the data set. When building a decision tree, select the feature with the largest Gini index drop as the split node to gradually build a classification decision tree. Set the maximum depth of the decision tree to , the minimum number of sample leaf nodes is , the random seed is , fit the decision tree model through training data,

[0059] Model evaluation: Use the test data set to evaluate the trained decision tree model, using accuracy as the evaluation indicator.

[0060] In the above-mentioned method for optimizing cigarette box design based on consumer sensory evaluation, preferably, the visual display of statistical analysis results and analysis results of the decision tree model includes:

[0061] Decision tree visualization;

[0062] Heat map drawing;

[0063] Importance analysis chart drawing.

[0064] In the above-mentioned cigarette box design optimization method based on consumer sensory evaluation, preferably, the decision tree visualization includes:

[0065] When judging a node, the design features The average rating of As the threshold, the samples are divided into different branches, and the node division effect is measured by information gain. The calculation formula of information gain is:

[0066]

[0067] in, Indicates the sample set contained in the current node, Indicates the attribute used for partitioning, Representation attributes The number of values of express Properties The upper value is A subset of samples, Represents a sample set The information entropy is calculated as , Represents a sample set Middle The proportion of class samples;

[0068] When drawing a decision tree, a professional drawing library is used to convert the structural information of the decision tree into a visual graph: the size of the node is adjusted according to the number of samples. The larger the number of samples, the larger the node; the thickness of the branch represents the proportion of samples on the branch. The larger the proportion, the thicker the branch.

[0069] The heat map drawing includes:

[0070] Pivot the data and construct a data matrix with price group as row index, style as column index, and the most popular design category as cell content. Suppose there is Price groups, Style, constructed matrix for The matrix of Indicates the Price group and Target statistics under different styles;

[0071] Normalize the data in the matrix. The normalized data range is between;

[0072] Use the drawing library to draw a heat map. In the heat map, the depth of the color represents the size of the data;

[0073] The importance analysis chart drawing includes: drawing a bar chart with design category as the horizontal axis and importance score as the vertical axis, and grouping and displaying the data according to price groups, adding error bars on the bar chart, and setting the length of the error bars to a preset multiple of the standard error to indicate the degree of dispersion of the data, and using the Pearson correlation coefficient to calculate the correlation coefficient between the importance of each design category.

[0074] In the above-mentioned method for optimizing cigarette box design based on consumer sensory evaluation, preferably, the comprehensive analysis and result integration of basic data and visualization results include:

[0075] Integrate basic data;

[0076] Integrate relevant data for visualization results;

[0077] Construct comprehensive analysis indicators;

[0078] Establishing a framework for integrating results;

[0079] Results presentation and storage,

[0080] The integration of basic data includes: summarizing the key data obtained through statistical analysis and decision tree model construction steps, the key data including the mean score of each design element category under different price groups and sensory dimension groups , Mean importance of design elements , and the relevant parameters of the decision tree model,

[0081] The integrated visualization results related data includes: extracting the data based on the heat map and importance analysis chart, and extracting the mean and standard error of the importance score of each design category from the importance analysis chart , used to measure the stability and significance of importance scores,

[0082] The construction of comprehensive analysis indicators includes: calculating the comprehensive impact coefficient by the following formula , the calculation formula is:

[0083]

[0084] in, represents the weight coefficient, and , Indicates the normalization of the mean score of the design factor category. Indicates the normalization of the mean importance of design factors. represents the accuracy of the decision tree model,

[0085] The establishment of the result integration framework includes: constructing a three-dimensional data integration framework based on price group, sensory dimension and design element category as classification dimensions,

[0086] The result presentation and storage include: creating a data frame using Python's pandas library, presenting the integrated data in a table, and saving the data frame in CSV format.

[0087] The present invention provides a cigarette box design optimization method based on consumer sensory evaluation. By comprehensively collecting consumer sensory evaluation data and conducting in-depth data preprocessing, separation and reorganization, and statistical analysis, it can accurately analyze consumers' preferences and needs for various cigarette box design elements. A decision tree model is constructed based on a large amount of data, and the average importance and score of each design element are calculated, providing a scientific data basis for cigarette box design decisions. Based on this quantitative data, design elements can be reasonably selected, design solutions can be optimized, design risks can be reduced, and the design success rate and product market competitiveness can be improved. With the help of visual display and comprehensive analysis results, the differences in consumer preferences for cigarette box designs in different price groups and styles can be quickly understood, and market dynamics and changing trends in consumer demand can be timely grasped. Based on this information, companies can quickly adjust cigarette box design strategies, launch products that meet market demand, and enhance product adaptability and competitiveness in the market. Basic data and data related to visualization results are integrated to construct comprehensive analysis indicators and establish a three-dimensional data integration framework, providing comprehensive and systematic data support for cigarette box design decisions. This data can not only be used for current design decisions, but can also be stored in CSV format, facilitating subsequent report writing, data sharing, and data mining, providing a strong guarantee for companies' long-term cigarette box design optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0088] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be further described below with reference to the accompanying drawings, in which:

[0089] Figure 1 This is a flow chart of an embodiment of the cigarette box design optimization method based on consumer sensory evaluation provided by the present invention. DETAILED DESCRIPTION

[0090] Various exemplary embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. The description of the exemplary embodiments is merely illustrative and is in no way intended to limit the present disclosure, its application, or use. The present disclosure can be implemented in many different forms and is not limited to the embodiments described herein. These embodiments are provided to make the present disclosure thorough and complete and to fully convey the scope of the present disclosure to those skilled in the art. It should be noted that unless otherwise specifically stated, the relative arrangement of parts and steps, the composition of materials, numerical expressions, and numerical values set forth in these embodiments should be interpreted as being merely exemplary and not as limiting.

[0091] The terms "first," "second," and similar terms used in this disclosure do not indicate any order, quantity, or importance, but are simply used to distinguish different parts. Terms such as "include" or "comprising" mean that the elements preceding the term include the elements listed after the term, and do not exclude the possibility of also including other elements. Terms such as "upper," "lower," and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0092] In the present disclosure, when a specific component is described as being located between a first component and a second component, there may or may not be an intervening component between the specific component and the first component or the second component. When a specific component is described as being connected to another component, the specific component may be directly connected to the other component without an intervening component, or may not be directly connected to the other component but have an intervening component.

[0093] All terms (including technical or scientific terms) used in this disclosure have the same meaning as those understood by one of ordinary skill in the art to which this disclosure belongs, unless otherwise specifically defined. It should also be understood that terms defined in, for example, general dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an idealized or highly formal sense, unless explicitly defined herein.

[0094] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, they should be considered part of the specification.

[0095] like Figure 1 As shown, the cigarette box design optimization method based on consumer sensory evaluation provided in this embodiment includes the following steps in actual implementation:

[0096] Step S1: Obtain consumer sensory evaluation data.

[0097] In one embodiment of the present invention, raw questionnaire data stored in CSV format is obtained from a collection channel. This data, collected from a consumer sensory evaluation survey on cigarette package design elements, contains a wealth of information. Specifically, the data is read using the pandas library in Python. First, an attempt is made to read the data using UTF-8 encoding. If a UnicodeDecodeError occurs, the data is read using GBK encoding. If GBK encoding also fails, the data is read using UTF-8-SIG encoding. Because data may exist in different encoding formats, the present invention employs a strategy of attempting multiple encodings to ensure accurate data reading.

[0098] Step S2: Data preprocessing and preparation, and standardization of scores for different sensory dimensions.

[0099] In one embodiment of the cigarette box design optimization method based on consumer sensory evaluation of the present invention, step S2 may specifically include:

[0100] Step S21: Data integrity check.

[0101] In one embodiment of the present invention, after reading the data, a comprehensive data integrity check is performed. The number of missing values in each column is counted. If the proportion of missing values in a column exceeds a preset threshold (e.g., 20%), a response is determined based on the specific situation. Specifically, missing numerical data columns are filled using statistical values or interpolation. Missing non-numeric data columns are filled based on the mode or business logic, or the corresponding columns are deleted.

[0102] For example, in some embodiments of the present invention, for numerical data columns with few missing values, statistical values such as the mean and median can be used to fill the missing values. If there are many missing values, consider deleting the column or interpolating the missing values based on information from other related columns. For non-numeric data columns, if there are few missing values, fill the missing values based on the mode or business logic. If there are too many missing values, carefully consider whether to delete the column.

[0103] For example, for the "Purchase Price" column, if there are a small number of missing values, we can analyze the correlation between the age, gender and other information of other respondents and the purchase price, and use regression models and other methods to predict the missing values and fill them in.

[0104] Step S22: Data consistency check.

[0105] Specifically, check for inconsistencies in the data, such as inconsistent data types and inconsistent data formats, to ensure that the data types of all data columns meet expectations. For example, standardize the data type of the "Age" column to numeric, and convert any age data in string format. For the "Purchase Price" column, if there are different representations (such as "80-120 yuan" and "80~120 yuan"), convert them to a standard format to ensure data consistency and comparability.

[0106] Step S23: score standardization processing.

[0107] Specifically, the scores of different sensory dimensions are standardized using the linear transformation formula Standardization is performed, where represents the original score, Indicates the minimum value of the original score of this dimension. Indicates the maximum value of the original score of this dimension. Represents the normalized score. After the above processing, all data are measured on the same scale, which facilitates subsequent statistical analysis and model training.

[0108] Taking the dimension of “simplicity” as an example, in the original score, 1 means “very complex” and 5 means “very simple”, but the scores of different questionnaires may have scale differences.

[0109] Step S24: data cleaning and denoising.

[0110] Specifically, the outliers in the data are identified and processed by drawing box plots and calculating the interquartile range (IQR) method to determine the range of outliers in the data.

[0111] In one embodiment of the present invention, for continuous data, if a sample value is less than or greater than (in is the lower quartile, is the upper quartile, ), it is considered an outlier. For outliers, you can handle them according to the actual situation, such as deleting the outliers, replacing them with reasonable boundary values (such as or ) etc. At the same time, check whether there are duplicate records in the data. If so, delete the duplicate samples to ensure the accuracy and uniqueness of the data.

[0112] Step S3: Separate and reorganize data based on design element classification and price group labeling.

[0113] After completing step S2, the data obtained is cleaned, standardized, and complete. At this time, in order to further explore the impact of different design elements on consumer sensory evaluation, the present invention classifies the design elements. In one embodiment of the cigarette box design optimization method based on consumer sensory evaluation of the present invention, step S3 may specifically include:

[0114] Step S31: Classification and mapping of design elements.

[0115] Specifically, based on the data column name With keywords , according to the Levenshtein distance Calculate the edit distance between two strings and set the distance threshold ,like , then the data column is considered to belong to the corresponding design element category.

[0116] The Levenshtein distance is the minimum number of operations required to transform one string into another by inserting, deleting, or replacing characters. This mapping process provides the foundation for subsequent grouping and analysis by design element category.

[0117] In one embodiment of the present invention, through extensive market research and professional analysis, six main categories of design elements are identified: opening and closing method, illustrations, fonts, colors, composition, and shading. For each category, a mapping relationship with relevant keywords is established, such as design_categories={'opening and closing method':['opening and closing method'], 'illustrations':['illustrations', 'brand 1ogo', 'fine brushwork'], 'fonts':['fonts'], 'colors':['background color', 'splicing color'],"composition':「'composition'], 'shading':「'shading']}. The establishment of this mapping relationship is based on the principle of text matching. The present invention uses a string similarity algorithm (such as the Levenshtein distance algorithm) to determine the degree of matching between data column names and keywords.

[0118] Step S32: Definition of importance column identifiers.

[0119] Specifically, based on an in-depth understanding of the questionnaire design and data structure, identifiers are determined to accurately identify columns in the data that are related to the importance scores of each design element.

[0120] By defining importance column identifiers for design categories, we can accurately measure the importance of each design element. In subsequent analysis, we can extract the corresponding importance score data by identifying whether the data column names contain these identifiers. For example, importance column identifiers might be: importance_identifiers = ['%1.Opening and Closing Style', '%2.Illustration', '%3.Font', '%4.Color', '%5.Composition'].

[0121] Step S33: Price group mapping and data labeling.

[0122] In one embodiment of the present invention, price column information is extracted from the data based on a pre-set price group mapping relationship. For example, the price group mapping relationship is as follows:

[0123]

[0124] Specifically, first, we traverse the data column names to find the column containing price keywords (such as price or price) as the price column; then, we process each data in the price column, extract the first character using the string truncation operation, and then convert the first character into a specific price group name based on the price group mapping relationship, and add a new price group column to the dataset for labeling.

[0125] For example, if a data item in the price column is "A100 yuan," the first character "A" is extracted and, based on the mapping relationship, the price group of the sample is labeled "High Price (80-120 yuan)." This process provides a preliminary classification of the data by price group, providing a key classification basis for subsequent in-depth analysis of different price groups.

[0126] Step S34: Data reorganization and structural processing.

[0127] Specifically, based on design element classification and price group labeling, the data was reorganized. This involved grouping data columns of the same category together based on design element category, and stratifying them by price group. Furthermore, the importance score data was associated and integrated with the corresponding design element data to construct a structured data table. In this process, the principles of pivot tables were applied, with row and column conversions and data aggregation operations employed to transform the original questionnaire data into a structure more amenable to analysis.

[0128] Step S4: Perform statistical analysis and build a decision tree model based on the reorganized data.

[0129] After completing step S3, the data has been classified and labeled according to the design element category and price group. On this basis, further preparatory work is carried out for the grouped data of each price group and sensory dimension. In one embodiment of the cigarette box design optimization method based on consumer sensory evaluation of the present invention, step S4 may specifically include:

[0130] Step S41: Prepare group data.

[0131] Specifically, for each grouped data, first check the integrity and validity of the data to ensure the reliability of the subsequent analysis results, and check whether there are too many missing values or outliers that affect the analysis. If so, process it according to the data preprocessing and preparation method (step S2).

[0132] Step S42: Calculate the mean of the design element category scores.

[0133] By calculating the mean score of each design element category, we can understand the evaluation level of different design element categories in the minds of consumers. Specifically, let a design element category be , the column set contained in this category is , the sample size is , use the following formula to calculate the mean score of this category :

[0134]

[0135] in, Indicates the Liedi Ratings of samples.

[0136] This formula calculates the average score for a design feature category by doubly summing the scores for all columns in that category and dividing the sum by the product of the sample size and the number of columns.

[0137] For example, for the "Illustration" category, if it contains three columns of data: "Illustration," "Brand Logo," and "Detailed Painting," we calculate the sum of the ratings for each column, add the sums of the three columns together, and finally divide the sum by the product of the sample size and 3 (the number of columns) to obtain the mean rating for the "Illustration" category. This mean comprehensively reflects consumers' overall evaluation of the "illustration" design element category.

[0138] Step S43: Calculate the mean importance of design elements.

[0139] By calculating the average importance of each design element, we can measure consumers’ awareness of the importance of each design element. Specifically, let the importance of a design element be , the following formula is used to calculate the mean importance of design factors:

[0140]

[0141] in, For the The score of each sample in this importance column.

[0142] The mean importance of the design factor is obtained by summing the scores of all samples in the importance column and dividing it by the number of samples.

[0143] For example, for the importance column of the "font" design factor, add up the scores of all samples in the column and divide it by the total number of samples to get the average importance of the "font" design factor, so as to understand consumers' average evaluation of the importance of "font".

[0144] Step S44: Prepare for building a decision tree model.

[0145] In one embodiment of the cigarette box design optimization method based on consumer sensory evaluation of the present invention, step S44 may specifically include:

[0146] Step S441, target variable selection: select the design category corresponding to the highest value from the mean score of each design factor category as the target variable of the decision tree analysis, denoted as , in the target category In the above example, we further select the specific design features with the highest scores and record them as .

[0147] The present invention selects target variables based on the assumption that consumers pay more attention to the design category with the highest score and the rules by which this category is affected by other design elements are more valuable for research.

[0148] Step S442, classification target definition: based on design features The average rating of As the threshold, define the "high score" target variable , determined using the following formula Value:

[0149]

[0150] in, Indicates that the sample has design characteristics Rating on .

[0151] In this way, the samples are divided into two categories: "high score" and "non-high score", which builds a clear classification target for the decision tree model.

[0152] Step S443, feature selection: using the target category The target features in As features, the scores of other design elements other than , for each category , taking the columns it contains as features.

[0153] This feature selection strategy is based on the idea of "excluding target features and studying the impact of other design factors on the high score of the target category", which helps to accurately identify the key design factors that affect the high score of the target category.

[0154] Step S45: Decision tree model construction and evaluation.

[0155] In one embodiment of the cigarette box design optimization method based on consumer sensory evaluation of the present invention, step S45 may specifically include:

[0156] Step S451, data preprocessing: Before building the decision tree model, the selected feature data is preprocessed, and for missing values, multiple imputation method is used for data processing.

[0157] Among them, the multiple imputation method is based on Monte Carlo simulation, which generates multiple imputation values through multiple simulations to take into account the uncertainty of missing values. For the distribution differences of feature data, standardization methods are used to deal with them, such as Z-score standardization, with the following formula:

[0158]

[0159] in, represents the original feature data, represents the mean of the feature data, Represents the standard deviation. Standardization allows different feature data to have the same scale, improving model training results.

[0160] Step S452, model training: Using the CART algorithm, a decision tree model is constructed using the Python scikit-learn library. The CART (Classification and Regression Trees) algorithm performs feature selection and tree construction based on the Gini Index. The Gini Index is used to measure the impurity of a dataset. The formula for the Gini Index is:

[0161]

[0162] in, represents a dataset, represents the number of categories in the dataset, Indicates the The proportion of class samples in the data set. When building a decision tree, select the feature with the largest Gini index drop as the split node to gradually build a classification decision tree. Set the maximum depth of the decision tree to , the minimum number of sample leaf nodes is , the random seed is , fit the decision tree model through the training data.

[0163] The decision tree model is used to identify design categories and features that consumers pay close attention to, providing a clear direction for cigarette box design, making the design more in line with consumers' actual expectations, and thus improving the market appeal of the product.

[0164] Step S453: Model evaluation: Use the test data set to evaluate the trained decision tree model, using accuracy as the evaluation indicator.

[0165] The calculation formula for accuracy is: .

[0166] By calculating the accuracy rate, we can evaluate the accuracy of the model in classifying the test data, judge the performance of the model, and determine whether the model can effectively reveal the impact of different design factors on the high score of the target category.

[0167] Step S5: Visually display the statistical analysis results and the analysis results of the decision tree model.

[0168] In one embodiment of the cigarette box design optimization method based on consumer sensory evaluation of the present invention, step S5 may specifically include:

[0169] Step S51: Decision tree visualization.

[0170] After the decision tree model is constructed in step S4, the internal structure and decision logic of the decision tree model are relatively complex and difficult to understand intuitively. By adopting visualization technology, the classification process of the decision tree and the role of each design element in it can be clearly presented. Decision tree visualization is mainly based on graph theory, which displays the nodes, branches and leaf nodes of the decision tree in a graphical manner. In one embodiment of the cigarette box design optimization method based on consumer sensory evaluation of the present invention, step S51 may specifically include:

[0171] Step S511: When determining a node, use the design feature The average rating of As the threshold, the samples are divided into different branches, and the node division effect is measured by information gain. The calculation formula of information gain is:

[0172]

[0173] in, Indicates the sample set contained in the current node, represents the attributes used for partitioning (i.e., design elements), Representation attributes The number of values of express Properties The upper value is A subset of samples, Represents a sample set The information entropy is calculated as , Represents a sample set Middle The greater the information gain, the greater the "purity improvement" obtained by using this attribute for division, that is, the more critical the role of this design factor in the decision tree.

[0174] Each node in the decision tree represents a judgment condition for a design element, and its judgment is based on the relevant statistics (such as the mean) of each design element calculated in step S4. For example, when judging the "luxury level of font design" node, the mean of the font design luxury level scores calculated in step S4 is used as the threshold to divide the samples into different branches.

[0175] Step S512: When drawing the decision tree, use a professional drawing library (such as Graphviz in Python combined with the export_graphviz function of scikit-learn) to convert the structural information of the decision tree into a visual graph: adjust the size of the node according to the number of samples. The larger the number of samples, the larger the node; the thickness of the branch represents the proportion of samples on the branch. The larger the proportion, the thicker the branch.

[0176] In this way, the importance and impact path of different design factors in the decision-making process can be intuitively observed.

[0177] Step S52: drawing a heat map.

[0178] The heat map is used to display the distribution of the most popular design categories in each price group under different styles. The present invention draws the heat map based on the principles of data perspective and matrix visualization. In the analysis results of step S4, relevant data for each design category under different price groups and different styles has been obtained. In one embodiment of the cigarette box design optimization method based on consumer sensory evaluation of the present invention, step S52 may specifically include:

[0179] Step S521: Pivot the data and construct a data matrix with price group as row index, style as column index, and the most popular design category as cell content (which can be the frequency of occurrence or mean score of the design category under the price group and style). Assume that there are Price groups, Style, constructed matrix for The matrix of Indicates the Price group and Target statistics in different styles.

[0180] Step S522: normalize the data in the matrix. The normalized data range is between.

[0181] Normalization can enhance the readability of heat maps and facilitate differentiation using color gradients in heat maps. Normalization is performed using the following formula:

[0182]

[0183] in, Representation matrix The minimum value in Representation matrix The maximum value in .

[0184] Step S523: Draw a heat map using a drawing library (such as the heatmap function of the seabor library in Python). In the heat map, the depth of the color represents the size of the data.

[0185] Using different color mapping schemes (such as YIGnBu), you can visually compare consumer preferences for design categories across price groups and styles. For example, darker areas indicate a more popular design category within that price group and style. Heat maps can quickly reveal similarities and differences in preferences across price groups and styles.

[0186] Step S53: Draw an importance analysis chart.

[0187] In step S4, the mean importance of each design category in different price groups has been calculated. By plotting an importance analysis chart, the differences in importance evaluations of each design category in different price groups can be demonstrated. Specifically, a bar chart is drawn with design category as the horizontal axis and importance score as the vertical axis. The design categories are grouped and displayed according to price group (different colored bars represent different price groups). Error bars are added to the bar chart. The length of the error bars is set to a preset multiple (such as 1 or 2 times the standard error) to indicate the degree of data dispersion. The Pearson correlation coefficient is used to calculate the correlation coefficient between the importance of each design category.

[0188] By adding error bars, we can more accurately reflect the degree of difference in importance of each design category. Error bars can also more intuitively determine whether the difference between the means of importance of different design categories is statistically significant, that is, whether the difference is significant. The calculation of error bars is based on the concept of standard error in statistics. The formula for calculating standard error is as follows:

[0189]

[0190] in, represents the standard deviation of the sample, Indicates the sample size.

[0191] The correlation coefficient can be used to further analyze the relationship between the importance of each design category. The correlation coefficient uses the Pearson Correlation Coefficient, which is calculated as follows:

[0192]

[0193] in, and Represents two different design categories in The importance score of samples. The value range of the correlation coefficient is The closer the correlation coefficient is to 1, the stronger the positive correlation between the importance of the two design categories is. The closer the correlation coefficient is to -1, the stronger the negative correlation is. The closer the correlation coefficient is to 0, the weaker the correlation is. By calculating the correlation coefficient and marking it in the chart or through additional analysis, we can gain a deeper understanding of the intrinsic relationship between the importance of each design category.

[0194] Step S6: Comprehensively analyze and integrate the basic data and the visualization results, wherein the basic data is data related to statistical analysis and decision tree model.

[0195] In one embodiment of the cigarette box design optimization method based on consumer sensory evaluation of the present invention, step S6 may specifically include:

[0196] Step S61: Integrate basic data.

[0197] After obtaining the visualization analysis results in step S5, the various analysis results have been presented in an intuitive form. However, in order to further explore the value of the data, facilitate comprehensive comparison and further research, it is necessary to integrate the scattered results. Specifically, the key data obtained through the statistical analysis and decision tree model construction step (step S4) are summarized. The key data includes the mean score of each design element category under different price groups and sensory dimension groups. , Mean importance of design elements , and related parameters of the decision tree model, such as accuracy These data are a quantitative reflection of the relationship between consumer sensory evaluation and design elements, providing the core basis for comprehensive analysis.

[0198] Step S62: Integrate data related to the visualization results.

[0199] Specifically, the data on which the heat map and importance analysis chart are based are extracted by combining the information in the visualization analysis results obtained in step S5. For example, from the heat map data matrix Get the most popular design categories for each price group and style, as well as the corresponding normalized data , used to analyze different price groups and style preference patterns. In one embodiment of the present invention, the mean and standard error of the importance score of each design category are extracted from the importance analysis chart. , which is used to measure the stability and significance of the importance scores.

[0200] Step S63: Construct comprehensive analysis indicators.

[0201] By constructing comprehensive analysis indicators, we can more comprehensively evaluate the impact of design elements on consumer sensory evaluation. Specifically, the comprehensive impact coefficient is calculated using the following formula: Comprehensive Impact Coefficient (CI), calculated as:

[0202]

[0203] in, represents the weight coefficient, and , Indicates the normalization of the mean score of the design factor category. Indicates the normalization of the mean importance of design factors. Represents the accuracy of the decision tree model.

[0204] Among them, the weight coefficient can be determined by multi-criteria decision-making methods such as Analytic Hierarchy Process (AHP), and it can be set based on the research purpose and actual situation, taking into account the relative importance of each factor. , which can make different indicators comparable. The comprehensive influence coefficient can comprehensively reflect the overall influence of design elements in consumer evaluation. The higher the value, the better the overall performance of the design element in attracting consumers.

[0205] Step S64: Establish a result integration framework.

[0206] Specifically, a three-dimensional data integration framework was constructed using price group, sensory dimension, and design element category as classification dimensions. Within this framework, the integrated data was systematically organized. For example, for each price group (e.g., "high price (80-120 yuan)"), each sensory dimension (e.g., "luxury"), and each design element category (e.g., "illustration"), the corresponding design element category score average was recorded. , Mean importance of design elements , decision tree model accuracy , and the calculated comprehensive impact coefficient At the same time, relevant data in the heat map and importance analysis charts (such as the most popular design categories, standard errors of importance scores, etc.) are also included in the corresponding positions under this framework, forming a complete data cube structure to facilitate multi-dimensional cross-analysis.

[0207] Step S65: Present and store the results.

[0208] Specifically, a data frame was created using Python's pandas library, the integrated data was presented in a table, and the data frame was saved in CSV format. The data frame columns included price group, sensory dimension, design element category, mean rating for each design element category, mean importance of each design element, decision tree model accuracy, and comprehensive impact coefficient. It also included supplementary information such as heat maps and importance analysis charts (such as the most popular design category and standard error of importance ratings). This structured table format clearly displayed the relationships between the various analysis results, facilitating data retrieval and in-depth analysis. Finally, the data frame was saved in CSV format for subsequent report writing, data sharing, and data mining, providing comprehensive and systematic data support for cigarette package design decisions.

[0209] The cigarette package design optimization method based on consumer sensory evaluation provided by the embodiments of the present invention can accurately analyze consumers' preferences and demands for various cigarette package design elements by comprehensively collecting consumer sensory evaluation data and performing in-depth data preprocessing, separation and reorganization, and statistical analysis. A decision tree model is constructed based on a large amount of data, and the mean importance and mean score of each design element are calculated, providing a scientific data basis for cigarette package design decisions. Based on this quantitative data, design elements can be rationally selected, design solutions can be optimized, design risks can be reduced, and the design success rate and product market competitiveness can be improved. With the help of visual display and comprehensive analysis results, the differences in consumer preferences for cigarette package designs in different price groups and styles can be quickly understood, and market dynamics and changing trends in consumer demand can be timely grasped. Based on this information, companies can quickly adjust cigarette package design strategies, launch products that meet market demand, and enhance their adaptability and competitiveness in the market. Basic data and data related to visualization results are integrated to construct comprehensive analysis indicators and establish a three-dimensional data integration framework, providing comprehensive and systematic data support for cigarette package design decisions. This data can not only be used for current design decisions, but can also be stored in CSV format, facilitating subsequent report writing, data sharing, and data mining, providing a strong guarantee for companies' long-term cigarette package design optimization.

[0210] Thus far, various embodiments of the present disclosure have been described in detail. To avoid obscuring the concept of the present disclosure, some details known in the art have not been described. Based on the above description, those skilled in the art can fully understand how to implement the technical solutions disclosed herein.

[0211] Although some specific embodiments of the present disclosure have been described in detail through examples, those skilled in the art will understand that the above examples are for illustration only and are not intended to limit the scope of the present disclosure. Those skilled in the art will understand that the above embodiments may be modified or some technical features may be replaced with equivalents without departing from the scope and spirit of the present disclosure. The scope of the present disclosure is defined by the appended claims.

Claims

1. A cigarette box design optimization method based on consumer sensory evaluation, characterized in that: include: Obtain consumer sensory evaluation data; Data preprocessing and preparation, and standardization of scores for different sensory dimensions; Separate and reorganize data based on design element classification and price group labeling; Conduct statistical analysis and build a decision tree model based on the reorganized data; Visually display the statistical analysis results and the analysis results of the decision tree model; Comprehensive analysis and result integration are performed on basic data and visualization results, wherein the basic data is data related to statistical analysis and decision tree model.

2. The cigarette box design optimization method based on consumer sensory evaluation according to claim 1, characterized in that: The obtaining of consumer sensory evaluation data includes: Using the pandas library in Python to read data, first try Encoding read data, if unicodeDecode Error occurs, try to read it with gbk encoding; if gbk encoding reading also fails, try to Code reading, The data preprocessing and preparation, and the standardization of scores for different sensory dimensions, include: Data integrity check; Data consistency check; Scoring standardization; Data cleaning and denoising.

3. The cigarette box design optimization method based on consumer sensory evaluation according to claim 2, characterized in that: The data integrity check includes: for missing numerical data columns, filling them with statistical values or interpolation methods; for missing non-numerical data columns, filling them according to the mode or business logic, or deleting the corresponding columns. The data consistency check includes: checking whether there are inconsistencies in the data, The score standardization process includes: standardizing the scores of different sensory dimensions using a linear transformation formula Standardization is performed, where represents the original score, Indicates the minimum value of the original score of this dimension. Indicates the maximum value of the original score of this dimension. represents the normalized score, The data cleaning and denoising includes: identifying and processing outliers in the data by drawing box plots and calculating the interquartile range method.

4. The cigarette box design optimization method based on consumer sensory evaluation according to claim 1, characterized in that: The data separation and reorganization based on the design element classification and price group labeling includes: Design element classification and mapping establishment; Importance column identifier definition; Price group mapping and data annotation; Data reorganization and structuring processing.

5. The cigarette box design optimization method based on consumer sensory evaluation according to claim 4, characterized in that: The design element classification and mapping establishment includes: Based on data column name With keywords , according to the Levenshtein distance Calculate the edit distance between two strings and set the distance threshold ,like , then the data column is considered to belong to the corresponding design element category. The definition of importance column identifiers includes: determining identifiers based on a deep understanding of questionnaire design and data structure to accurately identify columns in the data that are related to the importance scores of each design element, The price group mapping and data labeling includes: extracting price column information from the data based on a pre-set price group mapping relationship. First, by traversing the data column names, searching for a column containing a price keyword as a price column; then, processing each data in the price column, extracting the first character using a string truncation operation, converting the first character into a specific price group name based on the price group mapping relationship, and adding a price group column to the data set for labeling; The data reorganization and structuring processing includes: reorganizing the data based on design element classification and price group labeling, specifically including: grouping data columns of the same category based on design element category, aggregating data columns of the same category together, and stratifying them according to price groups; at the same time, associating and integrating importance score data with corresponding design element data to construct a structured data table.

6. The cigarette box design optimization method based on consumer sensory evaluation according to claim 1, characterized in that: The statistical analysis and decision tree model construction based on the reorganized data include: Group data preparation; Calculation of mean scores for design element categories; Calculation of the mean importance of design elements; Preparation for decision tree model construction; Decision tree model construction and evaluation.

7. The cigarette box design optimization method based on consumer sensory evaluation according to claim 6, characterized in that: The packet data preparation includes: For each grouped data, first check the integrity and validity of the data to see if there are too many missing values or outliers that affect the analysis. If so, process them according to the data preprocessing and preparation methods. The calculation of the mean value of the design element category score includes: assuming that a certain design element category is , the column set contained in this category is , the sample size is , use the following formula to calculate the mean score of this category : in, Indicates the Liedi The score of the samples, The calculation of the mean importance of the design elements includes: assuming that the importance of a certain design element is , the following formula is used to calculate the mean importance of design factors: in, For the The score of the samples in the importance column, The decision tree model construction preparation includes: Target variable selection: Select the design category corresponding to the highest value from the mean score of each design factor category as the target variable of the decision tree analysis, denoted as , in the target category In the above example, we further select the specific design features with the highest scores and record them as ; Classification target definition: based on design features The average rating of As the threshold, define the "high score" target variable , determined using the following formula Value: in, Indicates that the sample has design characteristics Ratings on Feature selection: using the target category The target features in As features, the scores of other design elements other than , for each category , taking the columns it contains as features, The decision tree model construction and evaluation includes: Data preprocessing: Before building the decision tree model, the selected feature data is preprocessed. For missing values, multiple imputation methods are used for data processing; Model training: Using CART algorithm, using Python The CART algorithm performs feature selection and tree construction based on the Gini index. The Gini index is used to measure the impurity of the data set. The formula of the Gini index is: in, represents a dataset, represents the number of categories in the dataset, Indicates the The proportion of class samples in the data set. When building a decision tree, select the feature with the largest Gini index drop as the split node to gradually build a classification decision tree. Set the maximum depth of the decision tree to , the minimum number of sample leaf nodes is , the random seed is , fit the decision tree model through training data, Model evaluation: Use the test data set to evaluate the trained decision tree model, using accuracy as the evaluation indicator.

8. The cigarette box design optimization method based on consumer sensory evaluation according to claim 1, characterized in that: The visualization of statistical analysis results and decision tree model analysis results includes: Decision tree visualization; Heat map drawing; Importance analysis chart drawing.

9. The cigarette box design optimization method based on consumer sensory evaluation according to claim 8, characterized in that: The decision tree visualization includes: When judging a node, the design features The average rating of As the threshold, the samples are divided into different branches, and the node division effect is measured by information gain. The calculation formula of information gain is: in, Indicates the sample set contained in the current node, Indicates the attribute used for partitioning, Representation attributes The number of values of express Properties The upper value is A subset of samples, Represents a sample set The information entropy is calculated as , Represents a sample set Middle The proportion of class samples; When drawing a decision tree, a professional drawing library is used to convert the structural information of the decision tree into a visual graph: the size of the node is adjusted according to the number of samples. The larger the number of samples, the larger the node; the thickness of the branch represents the proportion of samples on the branch. The larger the proportion, the thicker the branch. The heat map drawing includes: Pivot the data and construct a data matrix with price group as row index, style as column index, and the most popular design category as cell content. Suppose there is Price groups, Style, constructed matrix for The matrix of Indicates the Price group and Target statistics under different styles; Normalize the data in the matrix. The normalized data range is between; Use the drawing library to draw a heat map. In the heat map, the depth of the color represents the size of the data; The importance analysis chart drawing includes: drawing a bar chart with design category as the horizontal axis and importance score as the vertical axis, and grouping and displaying the data according to price groups, adding error bars on the bar chart, and setting the length of the error bars to a preset multiple of the standard error to indicate the degree of dispersion of the data, and using the Pearson correlation coefficient to calculate the correlation coefficient between the importance of each design category.

10. The cigarette box design optimization method based on consumer sensory evaluation according to claim 1, characterized in that: The comprehensive analysis and result integration of basic data and visualization results include: Integrate basic data; Integrate relevant data for visualization results; Construct comprehensive analysis indicators; Establishing a framework for integrating results; Results presentation and storage, The integration of basic data includes: summarizing the key data obtained through statistical analysis and decision tree model construction steps, the key data including the mean score of each design element category under different price groups and sensory dimension groups , Mean importance of design elements , and the relevant parameters of the decision tree model, The integrated visualization results related data includes: extracting the data based on the heat map and importance analysis chart, and extracting the mean and standard error of the importance score of each design category from the importance analysis chart , used to measure the stability and significance of importance scores, The construction of comprehensive analysis indicators includes: calculating the comprehensive impact coefficient by the following formula , the calculation formula is: in, represents the weight coefficient, and , Indicates the normalization of the mean score of the design factor category. Indicates the normalization of the mean importance of design factors. represents the accuracy of the decision tree model, The establishment of the result integration framework includes: constructing a three-dimensional data integration framework based on price group, sensory dimension and design element category as classification dimensions, The result presentation and storage include: creating a data frame using Python's pandas library, presenting the integrated data in a table, and saving the data frame in CSV format.

Images