Product evaluation method and electronic device

Abstract

The application discloses a product evaluation method and an electronic device. The method comprises the following steps: obtaining a user's respective component feedback value and overall feedback value of a target product in multiple dimensions, wherein the component feedback value of each dimension comprises a positive incentive value and a negative penalty value; according to a preset nonlinear regression model, the component feedback value is taken as an independent variable, the overall feedback value is taken as a dependent variable, and the model is substituted into calculation to determine model parameters; the model parameters comprise a positive inclination coefficient and a negative inclination coefficient respectively set for the positive incentive value and the negative penalty value; for the component feedback value, the demand type is determined according to the difference between the positive inclination coefficient and the negative inclination coefficient; the target feedback weight coefficient corresponding to the component feedback value is determined according to the positive inclination coefficient, the negative inclination coefficient and the demand type; and the overall evaluation degree of the target product is determined according to the component feedback value and the corresponding target feedback weight coefficient, so that the demand type is accurately identified and the weight is adaptively assigned, and objective evaluation of user satisfaction is realized.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more particularly to a product evaluation method and an electronic device. Background Technology

[0002] In the field of product experience optimization, companies are increasingly relying on multi-dimensional user evaluation data to measure product quality and satisfaction.

[0003] In related technologies, a linear weighted scoring method is used, which sums the user's scores for each dimension according to fixed weights to obtain the overall satisfaction score. This method assumes that the influence of all dimensions is linear and that the weights of positive and negative feedback are fixed, without considering the non-linear effects of different dimensions on overall satisfaction or the differences between positive and negative feedback. However, user feedback usually contains mixed positive and negative information, and the importance of different dimensions varies significantly. How to accurately quantify the contribution of each dimension to overall satisfaction becomes a key issue in optimizing the evaluation model. Summary of the Invention

[0004] This invention provides a product evaluation method and an electronic device to solve the problem in related technologies that fail to capture the asymmetric effects of positive and negative feedback, ignore the interaction effects between dimensions, and thus cause the evaluation results to be distorted.

[0005] According to one aspect of the present invention, a product evaluation method is provided, comprising: Obtain the user's component feedback values ​​for the target product in multiple dimensions and the overall feedback value for the target product. The component feedback values ​​for each dimension include positive incentive values ​​and negative penalty values. According to the preset nonlinear regression model, the component feedback values ​​of multiple dimensions are used as independent variables of the model, and the overall feedback value is used as the dependent variable of the model. These are substituted into the nonlinear regression model for fitting calculation to determine the model parameters in the nonlinear regression model. Each model parameter includes a positive bias coefficient and a negative bias coefficient set for the positive incentive value and the negative penalty value of each dimension, respectively. For each component feedback value of the dimension, the demand type of the dimension is determined based on the difference between the positive propensity coefficient and the negative propensity coefficient; The target feedback weight coefficient corresponding to the component feedback value is determined based on the positive tendency coefficient, the negative tendency coefficient, and the demand type. The overall evaluation score of the target product is determined based on the component feedback values ​​of each dimension and the corresponding target feedback weight coefficients, wherein the overall evaluation score is used to characterize the user's satisfaction with the target product.

[0006] According to another aspect of the present invention, a product evaluation apparatus is provided, comprising: The feedback value determination module is used to obtain the user's component feedback values ​​for the target product in multiple dimensions and the overall feedback value for the target product. The component feedback values ​​for each dimension include positive incentive values ​​and negative penalty values. The model parameter determination module is used to determine the model parameters in the nonlinear regression model by substituting the component feedback values ​​of multiple dimensions as independent variables and the overall feedback value as the dependent variable into the nonlinear regression model according to a preset nonlinear regression model. Each model parameter includes a positive bias coefficient and a negative bias coefficient set for the positive incentive value and the negative penalty value for each dimension, respectively. The demand type determination module is used to determine the demand type of each dimension based on the difference between the positive propensity coefficient and the negative propensity coefficient for each component feedback value. The feedback weight coefficient determination module is used to determine the target feedback weight coefficient corresponding to the component feedback value based on the positive tendency coefficient, the negative tendency coefficient, and the demand type. The evaluation degree determination module is used to determine the overall evaluation degree of the target product based on the component feedback values ​​of each dimension and the corresponding target feedback weight coefficients, wherein the overall evaluation degree is used to characterize the user's satisfaction with the target product.

[0007] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the product evaluation method according to any embodiment of the present invention.

[0008] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the product evaluation method described in any embodiment of the present invention.

[0009] According to another aspect of the present invention, embodiments of this disclosure also provide a computer program product, including a computer program that, when executed by a processor, implements the product evaluation method as described in any of the embodiments of this disclosure.

[0010] The technical solution of this invention involves obtaining the user's component feedback values ​​for a target product across multiple dimensions, as well as the overall feedback value for the target product. Each component feedback value includes a positive incentive value and a negative penalty value, accurately capturing the user's multi-dimensional positive and negative perceptual differences. Following a preset nonlinear regression model, the component feedback values ​​across multiple dimensions are used as independent variables, and the overall feedback value is used as the dependent variable. These are substituted into the nonlinear regression model for fitting calculation to determine the model parameters. Each model parameter includes a positive bias coefficient and a negative bias coefficient set for the positive incentive value and the negative penalty value for each dimension, quantifying the nonlinear differences in the positive and negative influences of each dimension. For each component feedback value of a dimension, the difference between the positive bias coefficient and the negative bias coefficient is used to determine the dimension. The system identifies demand types based on coefficient differences, providing a basis for product differentiation and iteration. It determines the target feedback weight coefficient corresponding to the component feedback value based on the positive tendency coefficient, the negative tendency coefficient, and the demand type. This allows for adaptive weight allocation, precise fusion of positive and negative feedback, and elimination of subjective bias. The overall evaluation degree of the target product is determined based on the component feedback value of each dimension and the corresponding target feedback weight coefficient. This overall evaluation degree characterizes the user's satisfaction with the target product. By integrating multi-dimensional components and dynamic weights, it accurately quantifies user satisfaction, addressing the issues of asymmetric influence from inability to capture positive and negative feedback, ignoring inter-dimensional interaction effects, and resulting in distorted evaluation results. Through nonlinear fitting, it automatically quantifies the contribution of positive and negative feedback, accurately identifies demand types, and adaptively assigns weights, achieving an objective and reliable evaluation of user satisfaction.

[0011] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart of a product evaluation method provided in Embodiment 1 of the present invention; Figure 2 This is a flowchart of a product evaluation method provided according to Embodiment 2 of the present invention; Figure 3This is a schematic diagram of the structure of a product evaluation device according to Embodiment 3 of the present invention; Figure 4 This is a schematic diagram of the structure of an electronic device that implements the product evaluation method of this invention. Detailed Implementation

[0014] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0015] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0016] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0017] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0018] It is understood that before using the technical solutions disclosed in the various embodiments of this disclosure, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this disclosure in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0019] For example, upon receiving a user's active request, a prompt message is sent to the user to explicitly inform them that the requested operation will require the acquisition and use of the user's personal information. This allows the user to independently choose whether to provide personal information to the software or hardware, such as the electronic device, application, server, or storage medium performing the operations of this disclosed technical solution, based on the prompt message.

[0020] As an optional but non-limiting implementation, in response to a user's active request, sending a prompt message to the user can be done via a pop-up window, where the prompt message can be presented in text format. Furthermore, the pop-up window can also include a selection control allowing the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0021] It is understood that the above notification and user authorization process are merely illustrative and do not constitute a limitation on the implementation of this disclosure. Other methods that comply with relevant laws and regulations may also be applied to the implementation of this disclosure.

[0022] It is understood that the data involved in this technical solution (including but not limited to the data itself, the acquisition or use of the data) shall comply with the requirements of relevant laws, regulations and related provisions.

[0023] Example 1 Figure 1 The present invention provides a flowchart of a product evaluation method according to Embodiment 1. This embodiment is applicable to situations where product satisfaction evaluation and experience optimization are based on multi-dimensional user feedback. The method can be executed by a product evaluation device, which can be implemented in hardware and / or software. Optionally, it can be implemented through an electronic device, such as a mobile terminal, a PC, or a server.

[0024] like Figure 1 As shown, the method may specifically include: S110. Obtain the user's component feedback values ​​for the target product in multiple dimensions and the overall feedback value for the target product, wherein the component feedback values ​​for each dimension include positive incentive values ​​and negative penalty values.

[0025] In this context, "user" can be understood as an individual or group using or experiencing the target product. As the source of feedback data, users provide evaluations of various product dimensions and overall satisfaction ratings, making them the primary actors in the evaluation process. "Target product" can be understood as the specific product to which users conduct multi-dimensional feedback evaluations and satisfaction assessments. As the core evaluation object for the entire feedback collection, model fitting, demand determination, weight calculation, and overall evaluation, it serves as the carrier for all data operations and analyses. "Dimension" can be understood as the various evaluation aspects or attributes that constitute the product evaluation. Deconstructing the product evaluation system allows for the quantification of user feedback from a segmented perspective, supporting refined sub-item evaluations and demand type determination. "Component feedback value" can be understood as the specific quantitative value of user feedback for a single dimension of the target product, carrying the user's positive or negative evaluation information within that single dimension. "Positive incentive value" can be understood as the positive quantitative score representing the user's recognition and satisfaction with that dimension within the component feedback value of a single dimension. It quantifies the degree of positive user preference for that single dimension and participates in regression model fitting and positive tendency coefficient calculation. The negative penalty value can be understood as a negative vectorized score representing user dissatisfaction or criticism of a single dimension's component feedback value. It quantifies the degree of negative user resistance in a single dimension and participates in regression model fitting and negative tendency coefficient calculation. The overall feedback value can be understood as the comprehensive score or feedback given by users to the overall satisfaction with the target product. It serves as the dependent variable (output / target value) of the model and is used to compare and fit with the component feedback values ​​of each dimension to calibrate the model's accuracy.

[0026] One alternative implementation defines an evaluation dataset. ,in Indicates the total number of users; Indicates the first The overall rating of the product by each user, with a value range of [value missing]. Integers or consecutive real numbers; For the first The multidimensional component feedback values ​​of each user, among which This indicates the total number of evaluation dimensions. Indicates the first The user in the first The component feedback values ​​are distributed across several dimensions, where negative numbers represent negative sentiment, integers represent positive sentiment, and 0 represents neutral sentiment. To eliminate differences in data units and enhance model convergence, the component feedback values ​​are further refined. Zero-mean normalization (Z-score normalization) is performed to obtain the normalized component feedback values. -1 represents extremely negative emotions. It represents extremely positive emotions.

[0027] In one optional implementation, the target product includes at least cigarettes, alcohol, and software. When the target product is cigarettes, the dimensions include, but are not limited to, aroma, taste, aftertaste comfort, appearance, and additional flavors. When the target product is alcohol, the dimensions include, but are not limited to, aroma, taste, body, smoothness on the palate, sweetness, aftertaste, spiciness, and color / appearance. When the target product is software, the dimensions include, but are not limited to, functional completeness, smoothness of operation, aesthetic appeal, ease of use, response speed, and stability. Aftertaste comfort can be understood as the degree of comfort in the mouth and throat after smoking a cigarette, representing an evaluation dimension of the residual experience after smoking, supplementing the quantitative evaluation of the entire cigarette smoking experience. Additional flavors can be understood as the additional flavor characteristics (such as mint, fruit, or flavor capsules) of cigarettes beyond the basic tobacco flavor, a unique evaluation dimension distinguishing differences in flavor profiles, used to capture differentiated experiences. By processing multi-dimensional feedback from target flattening, it achieves multi-scenario coverage of physical consumer goods and virtual software, and significantly improves the accuracy of product evaluation through multi-dimensional sensory quantification.

[0028] Based on the above scheme, optionally, the positive incentive value and the negative penalty value are determined according to the component feedback value, including at least one of the following: when the component feedback value is positive, the component feedback value and a preset fixed value are respectively determined as the positive incentive value and the negative penalty value; when the component feedback value is negative, the fixed value and the absolute value of the component feedback value are respectively determined as the positive incentive value and the negative penalty value.

[0029] Understandably, when the component feedback value is positive (i.e., greater than 0), the positive stimulus value equals the component feedback value, and the negative penalty value equals a fixed value, which can be 0. When the component feedback value is negative (i.e., less than 0), the positive stimulus value equals the fixed value, and the negative penalty value equals the absolute value of the component feedback value. Essentially, this involves splitting the potentially signed component feedback value into two non-negative channels (positive channels).

[0030] One alternative implementation, based on the nonlinear assumption of the Kano model—that the impact of positive and negative performance on satisfaction is nonlinear—introduces dummy variables to standardize the component feedback values. Decomposed into positive excitation values and negative penalty value Construct an asymmetric feature space; the specific mapping rule is as follows: for any i... Dimensions ( ):like Then let ,and ;like Then let ,and ;like Then let ,and Therefore, the original The 3D feature space is expanded to The asymmetric feature space of dimension 1, for the 1st dimension For each sample, its extended feature vector is represented as: .

[0031] By adopting this technical solution, through the separation of positive and negative feedback and the setting of fixed value baseline, the signed component feedback can be decomposed into independent positive incentive and negative penalty channels, which facilitates differentiated weighting and modeling, avoids positive and negative cancellation, and more accurately portrays the heterogeneous impact of user satisfaction and dissatisfaction.

[0032] S120. According to the preset nonlinear regression model, the component feedback values ​​of multiple dimensions are used as independent variables of the model, and the overall feedback value is used as the dependent variable of the model. These are substituted into the nonlinear regression model for fitting calculation to determine the model parameters in the nonlinear regression model. Each model parameter includes a positive bias coefficient and a negative bias coefficient set for the positive incentive value and the negative penalty value of each dimension, respectively.

[0033] The nonlinear regression model can be understood as a pre-defined nonlinear function fitting model used to establish a mapping relationship between dimensional feedback and overall feedback, thereby building a mathematical correlation framework between component feedback values ​​and overall feedback values, and realizing the nonlinear relationship modeling from component feedback to overall feedback. The fitting calculation can be understood as the process of substituting independent and dependent variable data into the model, solving for the optimal model parameters through algorithms, approximating the true mapping relationship, calibrating the nonlinear regression model, determining the positive and negative tendency coefficients corresponding to each dimension, and completing model construction. The model parameters can be understood as unknown coefficient variables to be solved in the nonlinear regression model, used to determine the model mapping rules, quantifying the degree of influence of positive and negative feedback from each dimension on the overall evaluation, and serving as the core basis for subsequent demand determination and weight calculation. The positive tendency coefficient can be understood as a model parameter configured separately for the positive incentive value of a single dimension, representing the influence weight of positive feedback, measuring the contribution intensity of the positive evaluation of that dimension to the overall product feedback, and used for demand type determination and feedback weight calculation. The negative tendency coefficient can be understood as a model parameter configured separately for the negative penalty value of a single dimension. It represents the influence weight of negative feedback, measures the intensity of the negative evaluation of this dimension on the overall product feedback, and is used for demand type determination and feedback weight solution.

[0034] Based on the above scheme, optionally, the nonlinear regression model is determined according to the following formula: ; in, This represents the overall feedback value. This represents the regression intercept term. This indicates the number of dimensions in the component feedback value. Indicates the first The positive tendency coefficients of each dimension, Indicates the first Positive incentive values ​​in each dimension, Indicates the first The negative tendency coefficient of each dimension Indicates the first Negative penalty values ​​in each dimension This indicates the error term.

[0035] The regression intercept term can be understood as the baseline constant term inherent in the nonlinear regression model when all independent variables are zero. It compensates for the baseline bias beyond all dimensional feedback variables, improving the overall model fitting accuracy. The number of dimensions can be understood as the number of dimensions involved in the evaluation (e.g., aroma, taste, aftertaste comfort, appearance, and additional flavor, a total of 5), limiting the scope of the summation operation and determining how many dimensions of positive and negative feedback variables the model includes. The error term can be understood as the residual deviation between the true overall feedback value and the overall feedback value fitted by the model, accommodating random disturbances and accidental factors not captured by the model, and measuring the model fitting bias. Follows a mean of 0 and a variance of normal distribution An alternative implementation uses Ordinary Least Squares (OLS) to estimate the model parameters and defines a target loss function. Minimize the sum of squared residuals: ; Indicates the number of users. Indicates the first Individual users; through the target loss function Find the partial derivatives and set them to zero to obtain the parameter vector. The optimal estimate is obtained; after obtaining the parameter estimates, a t-test is performed on the regression coefficients of each dimension, and insignificant variables with a p-value (the significance probability of the t-test) greater than the preset significance level (e.g., 0.05) are removed, retaining only the significant ones. and Used for Kano attribute classification and weight calculation in subsequent step S3; final output and These are key quantitative indicators used to determine basic, exciting, or expected needs.

[0036] By adopting this technical solution, the independent weighting of positive incentives and negative penalties is separated, and the heterogeneous contribution of positive and negative influences of each dimension to the overall feedback is accurately quantified, thereby improving the model fitting accuracy and the explanatory power of user satisfaction, and supporting refined product optimization.

[0037] S130. For each of the component feedback values ​​of the dimension, determine the demand type of the dimension based on the difference between the positive tendency coefficient and the negative tendency coefficient.

[0038] The demand types can be understood as user preference categories for a given dimension, determined by the difference between positive and negative tendency coefficients for the same dimension. These demand types include, but are not limited to, basic demands, exciting demands, and expected demands, providing a classification basis for accurately configuring feedback weights. Basic demands can be understood as fundamental, essential needs that users believe the product must possess and should be met by default. Failure to meet this dimension results in strong negative perception, while achieving it brings no significant positive surprise; it is used to identify the product's essential bottom-line functional dimensions. Exciting demands can be understood as surprising demands where the product's extra features can bring unexpected positive feelings, even without explicit user expectations. Achieving this dimension results in extremely high positive gains, while its absence brings no significant negative dissatisfaction; it is used to uncover product-enhancing features. Expected demands can be understood as regular demands where users have clear expectations; the better the product performs, the higher the satisfaction, and the worse the product performs, the lower the satisfaction. This dimension has a balanced positive and negative feedback impact and serves as an intermediate evaluation dimension for routine product optimization and core user concerns.

[0039] S140. Determine the target feedback weight coefficient corresponding to the component feedback value based on the positive tendency coefficient, the negative tendency coefficient, and the demand type.

[0040] The target feedback weight coefficient can be understood as the weighted coefficient of the feedback value of each dimension component, which is determined by combining the differences in positive and negative tendency coefficients of the dimensions and the type of demand. Differentiated evaluation weights are assigned to different dimensions to distinguish the importance of each dimension to the overall product satisfaction.

[0041] Based on the above scheme, optionally, determining the target feedback weight coefficient corresponding to the component feedback value according to the positive tendency coefficient, the negative tendency coefficient, and the demand type includes: determining a benchmark weight according to the positive tendency coefficient and the negative tendency coefficient; determining an initial feedback weight coefficient according to the benchmark weight and a preset reinforcement factor corresponding to the demand type; determining the sum of the initial feedback weight coefficients corresponding to multiple component feedback values; and determining the target feedback weight coefficient corresponding to the component feedback value according to the sum and the initial feedback weight coefficient.

[0042] The baseline weight can be understood as the basic original weight value calculated from the positive and negative propensity coefficients of the same dimension. It serves as the original weight base without considering demand type correction, providing a calculation basis for the subsequent introduction of reinforcement factors. The preset reinforcement factor can be understood as a fixed correction coefficient set in advance for different demand types, amplifying or contracting the weights for basic, expected, and exciting demands to reflect the differences in demand types. The initial feedback weight coefficient can be understood as a temporary weight value that has not been normalized after being corrected by the baseline weight combined with the preset reinforcement factor for the corresponding demand type. It carries the inherent weight of the dimension and the effect of demand type correction, providing an intermediate variable for the subsequent normalization solution of the target weight. The sum can be understood as the sum of the initial feedback weight coefficients corresponding to all evaluation dimensions, serving as the normalization denominator to convert each initial weight into the final target weight in proportion form.

[0043] In one optional implementation, the first... Baseline weights for each attribute dimension The benchmark weight The calculation formula is as follows: ; in, and The obtained numbers are respectively the first and second numbers. The positive and negative tendency coefficients of each dimension are analyzed. The positive and negative tendency coefficients are combined by the root mean square method to eliminate the influence of the sign direction of the coefficients and characterize the potential influence of the dimension under normal conditions.

[0044] Based on the positive tendency coefficient of output and negative tendency coefficient Based on the difference determination results, establish a classification index set for attribute dimensions; define the complete attribute set as... ,Will Divided into three mutually exclusive subsets: basic requirements This includes all requirements that are determined to be basic. Individual attribute dimensions; set of exciting needs This includes all those identified as excitement-type needs. Individual attribute dimensions; expected demand set This includes all the first ones that are determined to be expected needs. Each attribute dimension; and satisfies The intersection of any two subsets is an empty set.

[0045] By introducing psychological parameters based on prospect theory, non-linear reinforcement factors (including at least punishment, reward, and ignore factors) are set for different types of needs to simulate the differences in users' sensitivity to "loss" and "gain"; a punishment reinforcement factor is defined. This is used to amplify the weight of basic needs when they are not met, and its value range is set to... This characterizes the user's zero tolerance for the failure of basic functions; it defines the reward enhancement factor. This is used to amplify the weight of exciting needs when they are met more than expected; the value range is set to... Characterizes users' high sensitivity to unexpected surprises; defines the ignoring factor. This is used to reduce the weight of excitatory needs when they are not met; its value is set to [value missing]. This represents the user's tolerance for the absence of unnecessary functions.

[0046] Combining the aforementioned classification index set, benchmark weights, and enhancement factors, for the first... The first user's Each dimension constructs a real-time system dependent on the component feedback values. Initial feedback weighting coefficients for positive and negative polarities The initial feedback weighting coefficient The mathematical expression is a piecewise function: ; Subsequently, the initial feedback weighting coefficients were adjusted. Perform normalization to obtain the target feedback weight coefficients. : ; Among them, the denominator For the first Users in all The sum of weights across all dimensions; the target feedback weight coefficient Able to reflect the first in real time The instantaneous contribution rate of a user's emotional experience at a specific point in time and for a specific dimension to the overall evaluation score is used to achieve a technological leap from static average weight to dynamic personalized weight.

[0047] This technical solution generates a baseline weight based on positive and negative tendency coefficients, and obtains the initial weight by combining the demand type matching reinforcement factor. After normalization, the target feedback weight coefficient is obtained. This allows for the reasonable allocation of weights according to different dimensions of demand attributes, thereby improving the accuracy and rationality of product evaluation results.

[0048] S150. Determine the overall evaluation score of the target product based on the component feedback values ​​of each dimension and the corresponding target feedback weight coefficients, wherein the overall evaluation score is used to characterize the user's satisfaction with the target product.

[0049] The overall evaluation score can be understood as a comprehensive product evaluation value calculated by integrating the feedback values ​​of each dimension and the corresponding target feedback weight coefficients. This value quantifies the overall user satisfaction with the target product and serves as the final quantitative indicator for product evaluation and improvement. The level of satisfaction can be understood as the user's subjective recognition and satisfaction level based on their experience and overall feeling of the product across various dimensions. This is the ultimate goal of product evaluation, and the overall evaluation score is its quantitative expression.

[0050] Based on the above scheme, optionally, determining the overall evaluation degree of the target product according to the component feedback values ​​of each dimension and the corresponding target feedback weight coefficients includes: standardizing the component feedback values ​​of each dimension to zero mean, and determining the overall evaluation degree by weighted summation of the standardized results and the corresponding target feedback weight coefficients.

[0051] The zero-mean standardization can be understood as a standardization method that makes the mean of the data 0 and the standard deviation 1, unifying the numerical scale of feedback values ​​from different dimensions and avoiding interference from differences in units with the weighted evaluation results.

[0052] An alternative implementation method is the standardized component feedback value. and the normalized asymmetric dynamic weights in step S3 Calculate the first Individual users Instantaneous overall assessment degree generated at any moment The calculation formula is as follows: ; in, These are the component feedback values ​​after zero-mean standardization, and their values ​​are typically in the range of... between; The target feedback weight coefficient satisfies ;therefore It represents the overall product performance perceived by the user based on their specific need type (Kano model) and the intensity of their emotions.

[0053] By adopting this technical solution, the feedback values ​​of each dimension are standardized to zero mean to eliminate the interference of differences in the dimensions of different dimensions. Then, the overall evaluation degree is calculated by weighted summation of the target feedback weight coefficients, so that the multi-dimensional evaluation data are comparable and the overall product satisfaction evaluation results are objective and accurate.

[0054] Based on the above scheme, optionally, after determining the overall evaluation score of the target product according to the component feedback values ​​of each dimension and the corresponding target feedback weight coefficient, the method further includes: obtaining the reference evaluation scores of the user at multiple second moments within a preset time period; determining the timeliness evaluation score according to the overall evaluation score and the multiple reference evaluation scores; and determining the comprehensive evaluation score according to the timeliness evaluation score and a preset index mapping algorithm, wherein the second moment is earlier than the first moment when the component feedback values ​​are collected.

[0055] The preset time period can be understood as a pre-defined time interval that limits the time range of the reference evaluation degree, controls the length of historical data, and avoids outdated data interfering with the current judgment. The second moment can be understood as each historical time node earlier than the feedback collection moment within the preset time period, serving as the sampling point for historical evaluation data and providing a time-series comparison sample. The reference evaluation degree can be understood as the product evaluation degree obtained by the user at the second moment (historical moment), serving as a historical benchmark / comparison value, used to calculate the timeliness evaluation degree together with the current overall evaluation degree. The timeliness evaluation degree can be understood as an evaluation value with time trend attributes obtained by integrating the current overall evaluation degree and multiple historical reference evaluation degrees, taking into account both the current evaluation and historical trends, reflecting the time-series fluctuation characteristics of product evaluation. The indicator mapping algorithm can be understood as a pre-defined rule used to map the timeliness evaluation degree to a more intuitive or unified comprehensive indicator, or to transform it before integrating it with other dimensions. The comprehensive evaluation degree can be understood as the final product evaluation indicator that, after time-series fusion and algorithm mapping, simultaneously includes the current evaluation, historical trends, and timeliness characteristics, serving as the final output evaluation result, comprehensively representing the long-term and immediate user satisfaction level of the product.

[0056] This technical solution introduces historical reference assessments earlier than the current time to construct a timeliness assessment to capture the changing trends and stability of user satisfaction. A comprehensive assessment score is then generated through indicator mapping, taking into account both immediate evaluation and time evolution characteristics, thereby enhancing the dynamic interpretability and decision-making foresight of the assessment results.

[0057] Based on the above scheme, optionally, determining the timeliness assessment degree based on the overall assessment degree and multiple reference assessment degrees includes: for each reference assessment degree, determining the time difference between the first moment and the second moment, and determining a reference time influence coefficient of the reference assessment degree based on the time difference; determining an initial assessment degree based on multiple reference assessment degrees, their corresponding reference time influence coefficients, and the overall assessment degree; determining a comprehensive time influence coefficient based on the sum of multiple reference time influence coefficients; and determining the timeliness assessment degree based on the ratio of the initial assessment degree to the comprehensive time influence coefficient.

[0058] The time difference can be understood as the time interval difference between a first moment and a single second moment, measuring the distance of historical assessment data from the current moment, and serving as the input basis for calculating the reference time influence coefficient. The reference time influence coefficient can be understood as a time decay / influence weight coefficient configured for each reference assessment degree based on the size of the time difference, reflecting that the more distant the historical assessment, the lower the influence, and the more recent the influence, achieving time-series weighting of historical data. The initial assessment degree can be understood as an intermediate assessment value calculated by fusing the overall assessment degree, each reference assessment degree, and their corresponding reference time influence coefficients, summing the contributions of the current period and all historical time-series assessments, providing the original numerator for solving the timeliness assessment degree. The comprehensive time influence coefficient can be understood as the sum of all reference time influence coefficients, serving as the normalized denominator, used to balance the total time-series weights, and achieving the conversion and correction of the initial assessment degree.

[0059] One optional implementation involves obtaining the current system computation time point (first moment). and the Release time of each evaluation data point ; Calculate the time lag of the evaluation data The units are uniformly converted to "days"; an effective sliding time window is set. (For example, 90 days or 180 days), perform data time-domain truncation filtering: if If the data is invalid, the evaluation data is considered invalid historical data and will not be included in subsequent calculations to reduce the system's computational load and eliminate interference from outdated information; if If so, retain this data and include it in the attenuation calculation process. Based on the forgetting curve principle, an exponential time decay function is constructed. This is used to quantify the dissipation of information value over time; to address the limitations of traditional attenuation coefficients. For technical problems that are difficult to determine based on experience, the "information half-life" parameter is introduced. That is, setting when time has elapsed After that, the weight of the evaluation data decays to half of its initial value; based on the definition of half-life, the decay coefficient is derived. (That is, the reference time influence coefficient): ; Construct the final time decay function expression: ; in, The value is dynamically set based on the product lifecycle characteristics: for fast-moving consumer goods or high-frequency iterative software, the value is set... To enhance sensitivity to recent feedback; for durable consumer goods, set... To maintain the stability of a long-term reputation.

[0060] This technical solution dynamically assigns differentiated weights to historical assessments based on time differences, combines them with current assessments for weighted aggregation, and then normalizes them using a comprehensive time impact coefficient to obtain the timeliness assessment degree. This accurately quantifies the changing trend and stability of satisfaction over time, avoiding historical data lag or excessive interference.

[0061] Based on the above scheme, optionally, after determining the comprehensive evaluation degree according to the timeliness evaluation degree and the preset index mapping algorithm, the method further includes: determining the average time influence coefficient according to the comprehensive time influence coefficient and the number of the second time moments; and determining the user's feedback inducing factors for the target product according to the comprehensive evaluation degree and the average time influence coefficient.

[0062] The average time influence coefficient can be understood as the average time influence weight value at a single point in time obtained by equally dividing the comprehensive time influence coefficient according to the number of second moments. It represents the average influence of each historical moment on the current evaluation and is a key parameter for identifying feedback triggering factors. The feedback triggering factors can be understood as the intrinsic causes and inducements that lead users to generate the current comprehensive evaluation result and are influenced by time-related characteristics. By exploring the deep-seated driving reasons behind user evaluations, we can provide a basis for product optimization and user preference analysis.

[0063] An alternative implementation utilizes the time decay function. As a global aggregation weight, it represents the overall evaluation degree of a single sample instantaneously across all valid time windows. A weighted average is performed to eliminate the impact of sample size fluctuations and generate the original composite index. (Timeliness assessment): ; in, For time window Total number of valid samples; denominator It serves as a normalization function, representing the sum of time weights for all valid samples (i.e., the comprehensive time influence coefficient), preventing the overall index value from drifting due to a low number of recent comments.

[0064] The comprehensive evaluation score is determined based on the timeliness evaluation score and the preset index mapping algorithm according to the following formula: ; Alternatively, the extreme value mapping method can be used: ; in, Indicates the overall evaluation level. Indicates the baseline score. Indicates the scaling factor. Indicates the timeliness assessment level. This indicates the cutoff threshold. When a product experiences a recent quality crisis (a surge in negative reviews that are very recent), due to... lead to And the pain points have been addressed. Factor amplification, It will show a sharp downward trend, thus triggering an early warning.

[0065] This technical solution calculates the average time impact coefficient by combining the comprehensive time impact coefficient with the number of historical moments. It then uses the comprehensive evaluation degree to accurately deduce the factors that trigger user feedback. This not only quantifies the degree of time's impact on the evaluation but also uncovers the underlying causes behind the evaluation, providing a basis for product optimization decisions.

[0066] The technical solution of this invention involves obtaining the user's component feedback values ​​for a target product across multiple dimensions, as well as the overall feedback value for the target product. Each component feedback value includes a positive incentive value and a negative penalty value, accurately capturing the user's multi-dimensional positive and negative perceptual differences. Following a preset nonlinear regression model, the component feedback values ​​across multiple dimensions are used as independent variables, and the overall feedback value is used as the dependent variable. These are substituted into the nonlinear regression model for fitting calculation to determine the model parameters. Each model parameter includes a positive bias coefficient and a negative bias coefficient set for the positive incentive value and the negative penalty value for each dimension, quantifying the nonlinear differences in the positive and negative influences of each dimension. For each component feedback value of a dimension, the difference between the positive bias coefficient and the negative bias coefficient is used to determine the dimension. The system identifies demand types based on coefficient differences, providing a basis for product differentiation and iteration. It determines the target feedback weight coefficient corresponding to the component feedback value based on the positive tendency coefficient, the negative tendency coefficient, and the demand type. This allows for adaptive weight allocation, precise fusion of positive and negative feedback, and elimination of subjective bias. The overall evaluation degree of the target product is determined based on the component feedback value of each dimension and the corresponding target feedback weight coefficient. This overall evaluation degree characterizes the user's satisfaction with the target product. By integrating multi-dimensional components and dynamic weights, it accurately quantifies user satisfaction, addressing the issues of asymmetric influence from inability to capture positive and negative feedback, ignoring inter-dimensional interaction effects, and resulting in distorted evaluation results. Through nonlinear fitting, it automatically quantifies the contribution of positive and negative feedback, accurately identifies demand types, and adaptively assigns weights, achieving an objective and reliable evaluation of user satisfaction.

[0067] Example 2 Figure 2This is a flowchart of a product evaluation method provided in Embodiment 2 of the present invention. This embodiment is a further refinement of determining the demand type of the dimension based on the difference between the positive propensity coefficient and the negative propensity coefficient, building upon the previous embodiments. Optionally, the demand type includes at least basic demand, exciting demand, and expected demand. Determining the demand type of the dimension based on the difference between the positive propensity coefficient and the negative propensity coefficient includes: determining the absolute value of the difference between the positive propensity coefficient and the negative propensity coefficient; and determining the demand type of the dimension based on the absolute value of the difference and the magnitude of the absolute values ​​of the positive propensity coefficient and the negative propensity coefficient. Detailed implementation can be found in the description of this embodiment. Technical features that are the same as or similar to those in the previous embodiments will not be repeated here.

[0068] like Figure 2 As shown, the method may specifically include: S210. Obtain the user's component feedback values ​​for the target product in multiple dimensions and the overall feedback value for the target product, wherein the component feedback values ​​for each dimension include positive incentive values ​​and negative penalty values.

[0069] S220. According to the preset nonlinear regression model, the component feedback values ​​of multiple dimensions are used as independent variables of the model, and the overall feedback value is used as the dependent variable of the model. These are substituted into the nonlinear regression model for fitting calculation to determine the model parameters in the nonlinear regression model. Each model parameter includes a positive bias coefficient and a negative bias coefficient set for the positive incentive value and the negative penalty value of each dimension, respectively.

[0070] S230. For each of the component feedback values ​​of the said dimension, determine the absolute value of the difference between the positive tendency coefficient and the negative tendency coefficient.

[0071] The absolute value of the difference can be understood as the non-negative value obtained by taking the difference between the positive and negative tendency coefficients and removing the positive and negative signs. It quantifies the degree of deviation between the two coefficients and serves as the first criterion for classifying different demand types.

[0072] S240. Based on the absolute value of the difference and the magnitude of the absolute values ​​of the positive and negative propensity coefficients, determine the demand type of the dimension, wherein the demand type includes at least basic demand, excitement demand, and expectation demand.

[0073] Based on the above scheme, optionally, determining the demand type of the dimension according to the absolute value of the difference and the absolute values ​​of the positive and negative propensity coefficients includes at least one of the following: when the absolute value of the difference is not greater than a preset threshold, the demand type is determined to be an expected demand; when the absolute value of the difference is greater than a preset threshold and the absolute value of the positive propensity coefficient is less than the absolute value of the negative propensity coefficient, the demand type is determined to be a basic demand; when the absolute value of the difference is greater than a preset threshold and the absolute value of the positive propensity coefficient is greater than the absolute value of the negative propensity coefficient, the demand type is determined to be an exciting demand.

[0074] The preset threshold can be understood as a fixed critical value set in advance to determine the boundary between the absolute values ​​of the differences, and whether the difference in the dividing coefficient is significant, serving as a dividing standard to distinguish between expected needs and basic and exciting needs.

[0075] An alternative implementation, basic requirement Ω base : (The absolute value of the negative propensity coefficient is greater than the absolute value of the positive propensity coefficient); Excitatory demand Ω excite : (The absolute value of the positive propensity coefficient is greater than the absolute value of the negative propensity coefficient); Expectant demand Ω perform : and Approximate, the absolute value of the difference is less than or equal to the threshold δ (e.g., δ=0.1).

[0076] This technical solution categorizes demand into expected, basic, and exciting types by absolute difference and coefficient magnitude, accurately identifying the heterogeneous impact patterns of each dimension on user satisfaction. This provides a quantitative basis for differentiated weight allocation and product optimization strategies, enhancing the explanatory power of the evaluation and the pertinence of decision-making.

[0077] S250. Determine the target feedback weight coefficient corresponding to the component feedback value based on the positive tendency coefficient, the negative tendency coefficient, and the demand type.

[0078] S260. Determine the overall evaluation score of the target product based on the component feedback values ​​of each dimension and the corresponding target feedback weight coefficients, wherein the overall evaluation score is used to characterize the user's satisfaction with the target product.

[0079] The technical solution of this invention quantifies the absolute value of the difference between positive and negative tendency coefficients and combines the magnitude of their absolute values ​​to classify each dimension into basic, exciting, or expected needs. Without relying on human experience or questionnaire classification, it can accurately identify the heterogeneous effect patterns of different dimensions on user satisfaction, providing an objective and quantifiable basis for subsequent differentiated weight allocation, strengthening factor selection, and product optimization strategies, and significantly improving the interpretability and decision-making pertinence of the evaluation model.

[0080] Example 3 This embodiment of the invention is an optional embodiment of the above embodiments. This embodiment takes the target product as cigarettes as an example, and the dimensions are aroma sufficiency, smoothness of taste, comfort of aftertaste, cigarette appearance and craftsmanship and unique flavor of flavor capsules as examples. The overall feedback value is the overall satisfaction score as an example, and the component feedback value is the sensory attribute emotional score as an example.

[0081] The data comes from a tobacco company's digital consumer feedback platform. The system first extracts data from the database. A set of valid evaluation data for the target product.

[0082] Define the evaluation dataset . in, For the first Consumers gave their overall satisfaction ratings for this cigarette product. Data collection used a 5-point scale (1 = very dissatisfied, 5 = very satisfied). In this embodiment, to improve the fitting accuracy of the regression model, [the following parameters were used]. Treat it as a continuous variable.

[0083] For the first The multi-dimensional sensory attributes and emotional scores of consumers. This example selects the factors that have the greatest impact on cigarette quality. The key attribute dimensions are as follows: Aroma richness; Smoothness of texture; The aftertaste is pleasant; The appearance and craftsmanship of cigarettes; The unique flavor of the popping boba.

[0084] In the original questionnaire, the above attributes were scored using a 7-point Likert scale. To standardize the scale and establish a coordinate system with "neutral expectation" as the origin, the system performed zero-mean Z-score normalization on the raw scores to obtain standardized sentiment scores. .when When the value is 0, it indicates that the performance of this dimension is in line with the industry average (neutral); when When, it indicates that the performance in this dimension is better than expected (positive sentiment); when When this occurs, it indicates that the performance of this dimension is below expectations (negative sentiment).

[0085] Traditional linear regression assumes a linear relationship between the independent and dependent variables, meaning that the positive results from good performance are symmetrical to the negative results from poor performance. However, this clearly does not hold true in the cigarette consumption experience. For example, if the "cigarette appearance" is damaged (…), Users will be extremely disgusted if the appearance is perfect ( Users simply take it for granted, which does not significantly improve overall satisfaction. .

[0086] The psychological characteristics of this Kano model are transformed into a computer-computable mathematical model. The system introduces dummy variable technology to construct an asymmetric feature space.

[0087] For any i The first sample Each attribute dimension has the following mapping logic: Positive incentive component (i.e., positive incentive value) Construction: ; That is, the variable only has a value when users give a positive review to this dimension; otherwise, it is 0. This represents the degree to which something "exceeded expectations".

[0088] Negative penalty component (i.e., negative penalty value) Construction: ; That is, the variable only has a value (absolute value) when a user gives a negative review to this dimension; otherwise, it is 0. This represents the degree to which something is "below expectations" (i.e., the depth of the pain point).

[0089] Assume the first Consumer No. )right" The smoothness of the texture was rated as extremely high. ), and for " The cigarette's appearance received a negative review. Internally, this data is then transformed into: for the taste dimension: , Regarding the appearance dimension: , Through this step, the original The feature space of dimension is expanded to 3D asymmetric eigenvectors .

[0090] Based on the processed dataset, the following multivariate regression model is constructed to fit the consumer psychology curve: In this equation, each coefficient has a clear physical meaning and can directly guide the formulation improvement and process optimization of cigarette products: (Intercept): Represents the condition when all attributes are at a neutral level (i.e., all...) When a brand is considered to have a certain level of consumer satisfaction with a particular cigarette, this reflects the brand's pricing power or basic reputation. (Positive sentiment coefficient, also known as positive tendency coefficient): Characterizes the... The ability to "add icing on the cake" in various dimensions. For example, if "bursting boba flavor" The large number indicates that as long as the popping beads are made well, the satisfaction rate will increase significantly. (Negative sentiment coefficient, also known as negative tendency coefficient): Characterizes the... The slope of the dimension "decline in satisfaction due to negative reviews". It should be noted that in the least squares calculation, because... This represents the absolute value (positive number) of a negative score, and negative performance will lead to... The original value decreased, therefore the calculated value was... It is usually a negative value. In the subsequent determination logic of this embodiment, we focus on its absolute value. That is, the extent of the impact.

[0091] The parameters of the above equations are solved using ordinary least squares (OLS). The target loss function is then constructed. The aim is to minimize the difference between predicted and actual ratings. Sum of squared residuals between: ; Solving the parameter vector using matrix operations After the solution is completed, a t-test is automatically performed. Assume a significance level is set. If a certain attribute dimension (e.g., "packaging color") and The p-values ​​were all greater than 0.05, indicating that this attribute had no statistically significant impact on overall satisfaction (belonging to indifferent demand). This variable will be automatically removed from the subsequent model to ensure the model's simplicity and robustness.

[0092] according to and The type of requirement is determined as follows: Basic requirement Ω base : (The absolute value of the negative propensity coefficient is greater than the absolute value of the positive propensity coefficient); Excitatory demand Ω excite : (The absolute value of the positive propensity coefficient is greater than the absolute value of the negative propensity coefficient); Expectant demand Ω perform : and Approximate, the absolute value of the difference is less than or equal to the threshold δ (e.g., δ=0.1).

[0093] Based on the above calculations, regression coefficients for this cigarette product across five dimensions were output. The following are typical results from a specific calculation and an explanation of the judgment logic: Case A: Dimension (Cigarette appearance and manufacturing process) Calculation results: (Not significant) (Significant and with large absolute value).

[0094] Logical judgment: (0.85 >> 0.05).

[0095] Conclusion: Mark "cigarette appearance and manufacturing process" as a basic requirement.

[0096] Business Interpretation: This is entirely in line with the laws of the tobacco industry. If the appearance of the cigarette is substandard (such as having a hollow end or being damaged), consumers will be extremely dissatisfied (with this factor carrying significant weight); however, if the appearance is acceptable, it is taken for granted, and no matter how well it is made, it will not significantly increase satisfaction.

[0097] Case B: Dimensions (The unique flavor of popping boba) Calculation results: , .

[0098] Logical judgment: (0.65 >> 0.10).

[0099] Conclusion: "Breakfast flavor" is labeled as an excitatory need.

[0100] Business Analysis: The flavor of the capsule is a point of innovation. If the flavor is unique, consumers will be pleasantly surprised (with significant weight); if the flavor is average or even lacks a capsule, consumers will not give a bad review because it is not a necessary feature of traditional cigarettes.

[0101] Case C: Dimension (Aroma intensity) Calculation results: , .

[0102] Logical judgment: (The two are similar).

[0103] Conclusion: Marked as expected demand.

[0104] Business Analysis: The stronger the aroma, the higher the satisfaction; the weaker the aroma, the lower the satisfaction. The two show a linear relationship and form the core foundation of product competitiveness.

[0105] The steps described in this embodiment address the pain point that traditional linear weighting methods (such as directly averaging or expert scoring weights) cannot reflect the true psychological structure of consumers. Specifically, in the subsequent (weighting calculation), the "basic needs" identified in this step will be assigned a negative score range (...). Extremely high penalty weight ( The factors are assigned normal weights in the positive range. This means that if a cigarette brand recently receives a batch of negative reviews for "short-selling" cigarettes, even if the number of negative reviews is not large, the calculated satisfaction index (CSI) will drop sharply because it belongs to basic demand and is located in the negative range, thus quickly triggering a quality warning.

[0106] We completed the Kano attribute classification of cigarettes across various sensory dimensions (basic, stimulating, and expectant). This step aims to construct a dynamic weighting function based on these classification results, capable of changing in real-time with the user's emotional polarity, thereby simulating the different sensitivities to "pain points" and "pleasure points" in real-world consumer psychology.

[0107] To quantify the importance of each dimension under "normal" conditions, the calculated regression coefficients were used. and The benchmark weights are calculated using the root mean square method. .

[0108] by" Taking "aroma sufficiency" as an example: Assuming step S2 calculates... , The benchmark weights are calculated as follows: ; by" Taking "cigarette appearance and manufacturing process" as an example (basic type): assuming calculations... (Very small) (Very large). Therefore, its benchmark weight is calculated as follows: ; Note: Although the positive coefficient of this dimension is small, the root mean square algorithm still gives it a high baseline weight because the negative coefficient is large, which accurately reflects its importance as a "basic threshold".

[0109] Traverse all Generate a baseline weight vector based on each dimension. .

[0110] A reinforcement factor is defined. To achieve dynamic "transformation" of the weights, a psychological moderating factor based on Prospect Theory is introduced. In this embodiment, the specific configuration parameters for the cigarette product are as follows: Punishment Enhancement Factor The rationale behind this setting is that cigarettes are high-frequency, fast-moving consumer goods, and consumers have extremely low tolerance for quality defects (such as poor smoking experience or excessive off-flavors). Setting a 2.5 times penalty means that the negative impact of a serious defect is equivalent to the positive impact of 2.5 equally good features.

[0111] Reward Enhancement Factor The criteria for setting this up include giving a 1.5x weighted reward to innovative features such as popping beads, in order to encourage products to exceed expectations in these dimensions and boost overall satisfaction.

[0112] Ignore factors Rationale: For excitement-driven needs (such as popping boba), if the product isn't well-made (e.g., the popping boba is difficult to pop or the flavor is weak), consumers will generally just feel "it's not fun," rather than being as angry as they would be if they encountered a quality issue. Therefore, its weight is reduced to 20% to avoid unnecessary negative impacts on the overall index.

[0113] For the For each user, the system assigns scores based on their specific ratings for each dimension. The positive or negative value, combined with the set to which it belongs, is used for real-time table lookup calculation. .

[0114] Take the following two scenarios as examples: Scenario A: A user who encountered a quality incident (sample No. 108) was very angry because he encountered a "blank cigarette" (appearance defect), but he thought the flavor of the menthol capsule was okay.

[0115] Dimension (Cigarette appearance, basic type) ): User rating: (Negative score, serious dissatisfaction).

[0116] Triggering logic: belongs to and Basic pain point punishment.

[0117] Weight calculation: .

[0118] Result: The weight of this dimension instantly increased by 3.5 times, becoming the core factor dominating overall satisfaction.

[0119] Dimension (Breakfast flavor, stimulating) ): User rating: (Positive score, I think it's okay).

[0120] Triggering logic: belongs to and Exciting highlight rewards.

[0121] Weight calculation: .

[0122] Scenario B: A user who is indifferent to innovation (sample 209) thinks the cigarette looks normal, but feels that the flavor capsule has no taste and is even a bit strange.

[0123] Dimension (Cigarette appearance, basic type) ): User rating: (Positive score, feels normal).

[0124] Triggering logic: belongs to and Basic standards are the norm.

[0125] Weight calculation: .

[0126] Effect: Returns to normal weighting, does not take up weight in the score.

[0127] Dimension (Breakfast flavor, stimulating) ): User rating: (Negative score, I don't think it's good).

[0128] Triggering logic: belongs to and Excitatory deficiency is ignored.

[0129] Weight calculation: .

[0130] Result: The system "ignored" this negative review. Because poor performance is common for excitement-driven demands, it shouldn't significantly lower the CSI index. This avoids the unjustified case of "a sharp drop in score due to innovation" found in traditional weighted methods.

[0131] Based on the above logic, the following piecewise function model (code-based logic) is constructed: ; .

[0132] Sum of weights for different users Dramatic fluctuations can occur (for example, the total weight of user 108 might be as high as 4.0, while that of user 209 might only be 1.5). To ensure the mathematical rigor of subsequent index calculations, normalization must be performed.

[0133] Taking user 108 as an example, assuming the dynamic weight calculation result vector of its 5 dimensions is as follows: (Data for illustrative purposes only). Total weights and sums. .

[0134] The final standard weight for the "cigarette appearance" dimension. for: ; This means that in user 108's review, the "cigarette appearance" dimension alone contributed nearly 50% of the influence, and any negative score regarding appearance would be directly passed on to the final overall evaluation.

[0135] In traditional methods, one negative review for an empty cigarette can be diluted by 99 positive reviews. However, in this method, because... The presence of this factor amplifies the weight of this negative review several times over, which can quickly lower the CSI curve and allow management to detect process fluctuations on the production line immediately.

[0136] In traditional methods, if some users are not used to a newly introduced popping bead, it can lead to a drop in ratings and dampen the enthusiasm of developers. This method, however, addresses this issue by... The factor (ignore factor) automatically filters out minor negative feedback regarding the innovation, only addressing issues when the innovation truly resonates with users (triggering) High scores are only counted when there are rewards, thus encouraging companies to iterate their products.

[0137] In actual cigarette production and sales monitoring, data is highly time-sensitive. The system first extracts time parameters from the data entering the calculation queue.

[0138] Set the current calculation time point The date is January 20, 2026. The database is being read to retrieve the date stored therein. The publication timestamp of each evaluation data item .

[0139] For each data point, calculate its number of days from now: .

[0140] Considering that the inventory turnover cycle of cigarette products is typically around 3 months, this embodiment sets an effective sliding time window. sky.

[0141] If a comment was posted on September 1, 2025 ( The system classifies this data as "failed historical data." While this data is helpful for analyzing long-term trends, it will be discarded when calculating the real-time "Dynamic CSI" to reduce computational load and eliminate interference from outdated information.

[0142] Only when Only then does the data enter the subsequent attenuation calculation process.

[0143] For cigarettes, a fast-moving consumer good, this embodiment sets an information half-life. sky.

[0144] Physical implications: This means that after 14 days of a review's release, its influence on the current CSI index will decrease to half of what it was on the day of release; after 28 days, it will decrease to a quarter. This setting aligns with the rapid iteration of consumer word-of-mouth in the internet age.

[0145] Automatic calculation of attenuation coefficient : ; For the The evaluation data, its time weighting function Determined as: ; Scenario demonstration: Data A (released today) ): (Full weight, greatest influence).

[0146] Data B (released two weeks ago) ): (Semi-weighted).

[0147] Data C (released two months ago) ): (Only a meager 5% influence is retained to maintain the smoothness of the exponent and prevent the curve from breaking).

[0148] Continuing with user 108 (the angry user) as an example: The "empty cigarette" defect (basic type defect) was encountered.

[0149] The normalized dynamic weights for the "cigarette appearance" dimension have been calculated. (Very high weight).

[0150] Its appearance dimension standardized score is (Seriously bad review).

[0151] Assume that the scores for the other four dimensions are all 0 (neutral).

[0152] PSI calculation: ; Note: Although the original score is only -1.8, the user's overall PSI is significantly lowered because this dimension represents basic needs and its weight is amplified.

[0153] User 209 (the user who is unaware): Dissatisfaction with "bursting beads" (excitatory deficiency).

[0154] Calculate the weights of this dimension that are ignored. Shrink, with extremely low weight.

[0155] PSI calculation: ; Note: Despite the negative reviews, the PSI is only slightly negative and close to neutral due to the lack of excitement-type demand.

[0156] Through this step, the system compresses complex multi-dimensional evaluations into a single-dimensional Personal Satisfaction Index (PSI).

[0157] Global time-domain weighted aggregation and exponential standardization. Within the time window... One valid sample Perform weighted aggregation. The key point here is that the newer the data, the greater its "weight" in the average.

[0158] Original Index (That is, timeliness assessment) calculation. Assume there are only the two users mentioned above in the current window (for simplicity, only a value of 2 is used): User 108 (PSI=-0.887) commented yesterday. Time weight .

[0159] User 209 (PSI=-0.05) commented 60 days ago. Time weight .

[0160] ; The average time influence coefficient (traditional average) is: ; The calculated result (-0.8451) is significantly lower than the traditional average (-0.4685). This indicates that the system successfully identified serious dissatisfaction (user 108) as recent, while minor dissatisfaction (user 209) is older. Therefore, the current CSI index must be dominated by recent serious dissatisfaction, showing a sharp downward trend. This is precisely the function that companies dream of for quality crisis early warning—extremely sensitive to recent bad news and gradually forgetting about distant bad news.

[0161] Linear mapping of the exponent and hierarchical output. Original exponent. These are statistical values ​​based on the Z-score (typically between -3 and +3), which are not easily understood by management. The system uses a linear mapping to convert them into a standardized scale of 0-100.

[0162] Set mapping parameters: Benchmark score (This represents the basic score, which is neither a point nor a point of failure).

[0163] Scaling factor (Control the degree of dispersion of the scores).

[0164] Upper and lower limit truncation: .

[0165] ; Substitute the above calculation results: ; Final output and alert trigger. The system dashboard displays the current real-time satisfaction score for the target product as 71.56. Compared to the historical average (assuming 85), this represents a drop of over 10 points. System logic judgment: Determination of the decline: (Threshold).

[0166] Attribution analysis: Retrospective data revealed that the main driver of the decline was the "cigarette appearance" dimension (with amplified weight) and time weight (recent occurrence).

[0167] The "Dynamic Product Satisfaction Index Based on Asymmetric Influence Weights" constructed in this embodiment fully realizes the closed loop from micro-data to macro-decision: Timeliness Revolution: Utilizing Half-Life The parameters solve the "hindsight" problem of traditional monthly and quarterly reports. The CSI index becomes a real-time "electrocardiogram" rather than a post-mortem "autopsy report".

[0168] Balancing noise immunity and sensitivity: For occasional noise (such as a negative review from 60 days ago), due to time decay... It is so small that it will not interfere with the current exponential curve.

[0169] For sudden crises (such as yesterday's mass negative reviews), due to time weighting... And asymmetric weights quilt When magnified, the index will react dramatically in an instant.

[0170] Parameter configurability. Companies can adjust the half-life for different product lines. For example, for newly launched cigarette products, one could... Adjusting to 3 days means that you only need to focus on the reputation of the most recent week, thus achieving a "high-frequency" tactical adjustment during the new product promotion period.

[0171] The technical solution of this invention, by constructing an asymmetric regression and Kano demand classification model, accurately quantifies the heterogeneity of the positive and negative impacts of cigarettes in various dimensions. Combined with a dynamic weighting mechanism (α penalty, γ reward, η neglect), it achieves extreme sensitivity to pain points and moderate reward for pleasure points. By introducing a time half-life decay function, the index reflects recent changes in reputation in real time, avoiding the dilution of crisis signals by outdated data. After normalization and linear mapping, an intuitive CSI score is output, successfully solving the pain points of traditional linear weighting being unable to distinguish psychological expectations and difficult to capture sudden quality problems. It achieves a closed loop from micro-level single-customer evaluation to macro-level quality early warning, significantly improving the response speed and decision-making accuracy of tobacco companies to process fluctuations and market feedback.

[0172] Example 4 Figure 3 This is a schematic diagram of a product evaluation device provided in Embodiment 4 of the present invention. Figure 3 As shown, the device includes: a feedback value determination module 310, a model parameter determination module 320, a demand type determination module 330, a feedback weight coefficient determination module 340, and an evaluation degree determination module 350. Among them, The feedback value determination module 310 is used to acquire the component feedback values ​​of the user for the target product in multiple dimensions and the overall feedback value for the target product. Each component feedback value includes a positive incentive value and a negative penalty value. The model parameter determination module 320 is used to determine the model parameters in the nonlinear regression model by substituting the component feedback values ​​of multiple dimensions as independent variables and the overall feedback value as the dependent variable into the model, according to a preset nonlinear regression model. Each model parameter includes a positive skewness coefficient set for each dimension's positive incentive value and negative penalty value. The system comprises: a positive bias coefficient and a negative bias coefficient; a demand type determination module 330, configured to determine the demand type of each dimension based on the difference between the positive bias coefficient and the negative bias coefficient for each component feedback value; a feedback weight coefficient determination module 340, configured to determine the target feedback weight coefficient corresponding to each component feedback value based on the positive bias coefficient, the negative bias coefficient, and the demand type; and an evaluation degree determination module 350, configured to determine the overall evaluation degree of the target product based on the component feedback value and the corresponding target feedback weight coefficient for each dimension, wherein the overall evaluation degree characterizes the user's satisfaction with the target product. The technical solution of this invention involves obtaining the user's component feedback values ​​for a target product across multiple dimensions and the overall feedback value for the target product through a feedback value determination module. Each component feedback value includes a positive incentive value and a negative penalty value, accurately capturing the user's multi-dimensional positive and negative perceptual differences. A model parameter determination module uses a preset nonlinear regression model, substituting the component feedback values ​​of multiple dimensions as independent variables and the overall feedback value as the dependent variable, to perform fitting calculations and determine the model parameters in the nonlinear regression model. Each model parameter includes a positive bias coefficient and a negative bias coefficient set for the positive incentive value and the negative penalty value for each dimension, quantifying the nonlinear differences in the positive and negative influences of each dimension. A demand type determination module determines the demand type based on the difference between the positive bias coefficient and the negative bias coefficient for each dimension's component feedback value. The system defines the demand types for each dimension, enabling precise classification of demand types based on coefficient differences, providing a basis for product differentiation and iteration. A feedback weight coefficient determination module determines the target feedback weight coefficient corresponding to each component feedback value based on the positive tendency coefficient, the negative tendency coefficient, and the demand type. This allows for adaptive weight allocation, precise fusion of positive and negative feedback, and elimination of subjective bias. An evaluation degree determination module determines the overall evaluation degree of the target product based on the component feedback values ​​of each dimension and the corresponding target feedback weight coefficients. The overall evaluation degree characterizes the user's satisfaction with the target product. By integrating multi-dimensional components and dynamic weights, it accurately quantifies user satisfaction, addressing the issues of asymmetric influence from inability to capture positive and negative feedback and ignoring inter-dimensional interaction effects in related technologies, leading to distorted evaluation results. Through nonlinear fitting, it automatically quantifies the contribution of positive and negative feedback, accurately identifies demand types, and adaptively assigns weights, achieving an objective and reliable evaluation of user satisfaction.

[0173] Optionally, the nonlinear regression model is determined according to the following formula: ; in, This represents the overall feedback value. This represents the regression intercept term. This indicates the number of dimensions in the component feedback value. Indicates the first The positive tendency coefficients of each dimension, Indicates the first Positive incentive values ​​in each dimension, Indicates the first The negative tendency coefficient of each dimension Indicates the first Negative penalty values ​​in each dimension This indicates the error term.

[0174] Optionally, the feedback value determination module includes a decomposition submodule. The decomposition submodule is configured to, when the component feedback value is positive, determine the component feedback value and a preset fixed value as the positive incentive value and the negative penalty value, respectively; and when the component feedback value is negative, determine the fixed value and the absolute value of the component feedback value as the positive incentive value and the negative penalty value, respectively.

[0175] Optionally, the demand types include at least basic demands, exciting demands, and expected demands; the demand type determination module includes an absolute value determination submodule and a demand type determination submodule. The absolute value determination submodule is used to determine the absolute value of the difference between the positive propensity coefficient and the negative propensity coefficient; the demand type determination submodule is used to determine the demand type of the dimension based on the absolute value of the difference and the magnitude of the absolute values ​​of the positive propensity coefficient and the negative propensity coefficient.

[0176] Optionally, the demand type determination submodule is specifically used to: determine the demand type as an expected demand when the absolute value of the difference is not greater than a preset threshold; determine the demand type as a basic demand when the absolute value of the difference is greater than the preset threshold and the absolute value of the positive tendency coefficient is less than the absolute value of the negative tendency coefficient; and determine the demand type as an exciting demand when the absolute value of the difference is greater than the preset threshold and the absolute value of the positive tendency coefficient is greater than the absolute value of the negative tendency coefficient.

[0177] Optionally, the feedback weight coefficient determination module is specifically used to determine a baseline weight based on the positive tendency coefficient and the negative tendency coefficient; determine an initial feedback weight coefficient based on the baseline weight and a preset reinforcement factor corresponding to the demand type; determine the sum of the initial feedback weight coefficients corresponding to multiple component feedback values; and determine the target feedback weight coefficient corresponding to the component feedback value based on the sum and the initial feedback weight coefficient.

[0178] Optionally, the evaluation degree determination module is used to standardize the component feedback values ​​of each dimension to zero mean, and determine the overall evaluation degree by weighted summation of the standardized result and the corresponding target feedback weight coefficient.

[0179] Optionally, the product evaluation device further includes a comprehensive evaluation degree determination module. This module is used to, after determining the overall evaluation degree of the target product based on the component feedback values ​​of each dimension and the corresponding target feedback weight coefficients, obtain reference evaluation degrees from the user at multiple second moments within a preset time period, determine a timeliness evaluation degree based on the overall evaluation degree and the multiple reference evaluation degrees, and determine a comprehensive evaluation degree based on the timeliness evaluation degree and a preset index mapping algorithm, wherein the second moments are earlier than the first moment when the component feedback values ​​were collected.

[0180] Optionally, the comprehensive evaluation degree determination module includes: a timeliness evaluation degree determination submodule. The timeliness evaluation degree determination submodule is configured to, for each reference evaluation degree, determine the time difference between the first time point and the second time point, and determine a reference time influence coefficient for the reference evaluation degree based on the time difference; determine an initial evaluation degree based on multiple reference evaluation degrees, their corresponding reference time influence coefficients, and the overall evaluation degree; determine a comprehensive time influence coefficient based on the sum of multiple reference time influence coefficients; and determine the timeliness evaluation degree based on the ratio of the initial evaluation degree to the comprehensive time influence coefficient.

[0181] Optionally, the product evaluation device further includes a feedback factor determination module. This module is used to determine, after determining the comprehensive evaluation degree based on the timeliness evaluation degree and a preset index mapping algorithm, an average time influence coefficient based on the comprehensive time influence coefficient and the number of the second time points; and to determine the user's feedback triggering factors for the target product based on the comprehensive evaluation degree and the average time influence coefficient.

[0182] Optionally, the target product includes at least cigarettes, and the dimensions include at least aroma, taste, aftertaste comfort, appearance, and additional flavor.

[0183] The product evaluation device provided in the embodiments of the present invention can execute the product evaluation method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the method.

[0184] Example 5 Figure 4A schematic diagram of an electronic device 10, which can be used to implement embodiments of the present invention, is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0185] like Figure 4 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0186] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0187] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as a product evaluation method.

[0188] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication unit 19, or installed from storage unit 18, or installed from ROM 12. When the computer program is executed by processor 11, it performs the functions defined in the methods of the embodiments of the present invention.

[0189] In some embodiments, a product evaluation method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of a product evaluation method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform a product evaluation method by any other suitable means (e.g., by means of firmware).

[0190] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0191] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0192] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0193] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0194] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0195] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0196] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0197] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A product evaluation method, characterized in that, include: Obtain the user's component feedback values ​​for the target product in multiple dimensions and the overall feedback value for the target product. The component feedback values ​​for each dimension include positive incentive values ​​and negative penalty values. According to the preset nonlinear regression model, the component feedback values ​​of multiple dimensions are used as independent variables of the model, and the overall feedback value is used as the dependent variable of the model. These are substituted into the nonlinear regression model for fitting calculation to determine the model parameters in the nonlinear regression model. Each model parameter includes a positive bias coefficient and a negative bias coefficient set for the positive incentive value and the negative penalty value of each dimension, respectively. For each component feedback value of the dimension, the demand type of the dimension is determined based on the difference between the positive propensity coefficient and the negative propensity coefficient; The target feedback weight coefficient corresponding to the component feedback value is determined based on the positive tendency coefficient, the negative tendency coefficient, and the demand type. The overall evaluation score of the target product is determined based on the component feedback values ​​of each dimension and the corresponding target feedback weight coefficients, wherein the overall evaluation score is used to characterize the user's satisfaction with the target product.

2. The method according to claim 1, characterized in that, The nonlinear regression model is determined according to the following formula: ； in, This represents the overall feedback value. This represents the regression intercept term. This indicates the number of dimensions in the component feedback value. Indicates the first The positive tendency coefficients of each dimension, Indicates the first Positive incentive values ​​in each dimension, Indicates the first The negative tendency coefficient of each dimension Indicates the first Negative penalty values ​​in each dimension This indicates the error term.

3. The method according to claim 1, characterized in that, The positive excitation value and negative penalty value are determined based on the component feedback value, including at least one of the following: When the component feedback value is positive, the component feedback value and the preset fixed value are respectively determined as the positive incentive value and the negative penalty value; When the component feedback value is negative, the fixed value and the absolute value of the component feedback value are respectively determined as the positive incentive value and the negative penalty value.

4. The method according to claim 1, characterized in that, The demand types include at least basic demand, exciting demand, and expected demand; determining the demand type of the dimension based on the difference between the positive propensity coefficient and the negative propensity coefficient includes: Determine the absolute value of the difference between the positive propensity coefficient and the negative propensity coefficient; The demand type of the dimension is determined based on the absolute value of the difference, and the magnitudes of the absolute values ​​of the positive and negative propensity coefficients.

5. The method according to claim 4, characterized in that, Determining the demand type of the dimension based on the absolute value of the difference, and the magnitudes of the absolute values ​​of the positive and negative propensity coefficients, includes at least one of the following: If the absolute value of the difference is not greater than a preset threshold, the demand type is determined to be an expected demand. If the absolute value of the difference is greater than a preset threshold and the absolute value of the positive tendency coefficient is less than the absolute value of the negative tendency coefficient, the demand type is determined to be a basic demand. If the absolute value of the difference is greater than a preset threshold and the absolute value of the positive tendency coefficient is greater than the absolute value of the negative tendency coefficient, the demand type is determined to be an excitatory demand.

6. The method according to claim 1, characterized in that, The step of determining the target feedback weight coefficient corresponding to the component feedback value based on the positive propensity coefficient, the negative propensity coefficient, and the demand type includes: The baseline weight is determined based on the positive and negative tendency coefficients, and the initial feedback weight coefficient is determined based on the baseline weight and the preset reinforcement factor corresponding to the demand type. The sum of the initial feedback weight coefficients corresponding to the multiple component feedback values ​​is determined, and the target feedback weight coefficient corresponding to the component feedback value is determined based on the sum and the initial feedback weight coefficients.

7. The method according to claim 1, characterized in that, The step of determining the overall evaluation score of the target product based on the component feedback values ​​of each dimension and the corresponding target feedback weight coefficients includes: The component feedback values ​​of each dimension are standardized to zero mean, and the weighted sum of the standardized results and the corresponding target feedback weight coefficients is determined as the overall evaluation degree.

8. The method according to claim 1, characterized in that, After determining the overall evaluation score of the target product based on the component feedback values ​​of each dimension and the corresponding target feedback weight coefficients, the method further includes: The system obtains reference evaluation scores for the user at multiple second moments within a preset time period, determines a timeliness evaluation score based on the overall evaluation score and the multiple reference evaluation scores, and determines a comprehensive evaluation score based on the timeliness evaluation score and a preset index mapping algorithm, wherein the second moment is earlier than the first moment when the component feedback value is collected.

9. The method according to claim 8, characterized in that, The step of determining the timeliness assessment score based on the overall assessment score and multiple reference assessment scores includes: For each of the reference evaluation degrees, the time difference between the first time point and the second time point is determined, and the reference time influence coefficient of the reference evaluation degree is determined based on the time difference; The initial evaluation degree is determined based on the multiple reference evaluation degrees, their corresponding reference time influence coefficients, and the overall evaluation degree; The comprehensive time influence coefficient is determined by summing the multiple reference time influence coefficients. The timeliness assessment degree is determined based on the ratio of the initial assessment degree to the comprehensive time influence coefficient.

10. The method according to claim 9, characterized in that, After determining the comprehensive evaluation score based on the timeliness evaluation score and the preset index mapping algorithm, the method further includes: The average time influence coefficient is determined based on the comprehensive time influence coefficient and the number of the second time points; The factors inducing user feedback on the target product are determined based on the comprehensive evaluation degree and the average time influence coefficient.

11. The method according to claim 1, characterized in that, The target product includes at least cigarettes, and the dimensions include at least aroma, taste, aftertaste comfort, appearance, and additional flavor.

12. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the product evaluation method according to any one of claims 1-11.

Images