Scenario generation method and system based on consumer behavior psychological model and artificial intelligence

Abstract

The present invention provides a scenario generation method and system based on a consumer behavior psychological model and artificial intelligence, which involves consumer behavior analysis, including: dividing consumer groups; using two classification models to obtain consumer motivations for the same consumer group, and obtaining consumer motivations N1 and N2; if N1 is the same as N2, it is determined that the elements in the existing consumer scenario are invalid elements, and a new consumer scenario needs to be generated; if N1 is different from N2, it is determined that the existing consumer scenario has effectively stimulated consumers, and there are effective elements, and it is necessary to adjust the existing consumer scenario on the basis of retaining the effective elements; scenario element effectiveness analysis: obtaining the relationship between multiple scenario elements and consumer behavior indicators in the existing consumer scenario, and screening out effective elements; constructing a generative network model to generate new scenarios or optimize existing scenarios. The technical solution provided by the present invention can generate more targeted and stimulating consumer scenarios to promote consumer behavior.

Description

Technical Field

[0001] The present invention relates to the technical field of consumer behavior analysis, and in particular to a scenario generation method and system based on a consumer behavior psychological model and artificial intelligence. Background Art

[0002] With the continuous development of theories and models of consumer behavior analysis, as well as the technological progress of artificial intelligence in big data collection, computing power and algorithms, coupled with the actual demand of the cultural tourism business and shopping industries to create new consumption scenarios, the technology of intelligent generation of consumption scenarios has also emerged. This technology can automatically generate new consumption scenarios that meet consumer preferences based on consumers' historical behaviors, current needs and market trends, and can also provide optimization solutions for transforming existing scenarios. These scenarios can be virtual, such as product recommendation pages on e-commerce platforms, live streaming interfaces, and digital cultural tourism experience scenarios; they can also be real, such as the display layout, decoration and facilities of offline stores. By intelligently generating consumption scenarios, operators can better attract consumers' attention, stimulate consumers' motivation, strengthen consumers' willingness, and improve purchase conversion rates. In the process of realizing the present invention, the inventor found that the design ideas of the existing consumption scenario generation solutions are mainly based on the creative design of the environment from the perspective of the operator, focusing on the exhibition layout of the display of goods, and rarely considering the purchase behavior and psychological needs analysis from the perspective of consumers. There is also a lack of artificial intelligence generation technology based on consumer behavior data collection and psychological analysis. According to the stimulus-response theory (stimulus-organism-response, SOR) of behavioral psychology, consumers are motivated to consume and show positive purchasing behavior after being stimulated by both external scenes and internal psychological factors, while existing scene creation or generation technologies do not pay attention to this aspect. In fact, consumer behavior is influenced by both personal and environmental factors. Currently, consumer scene technologies generated based on artificial intelligence are more developed with consumer personal factors such as consumption level and personality preferences as intermediaries. In addition, in the technical route based on consumer personal factors, existing technologies mostly use the capture of consumer clothing and other characteristics, and rarely use the capture of consumer action and expression data to achieve the purpose of consumer psychological analysis, resulting in the inability of existing consumer scenes to effectively stimulate consumer motivation.

[0003] Therefore, how to generate more targeted and stimulating consumption scenarios becomes a technical problem that needs to be solved. Summary of the invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art or related technology, and discloses a scene generation method and system based on a consumer behavior psychological model and artificial intelligence, which can identify and retain effective scene elements in existing consumer scenes, and provide targeted improvement solutions for existing consumer scenes to generate more targeted and stimulating consumer scenes.

[0005] The first aspect of the present invention discloses a scene generation method based on a consumer behavior psychological model and artificial intelligence, including: dividing consumer groups: using a convolutional neural network and a K-means clustering algorithm based on the appearance data of consumers to divide consumers' individual aesthetic preferences and form multiple consumer groups; using two classification models for the same consumer group to obtain consumer motivations: constructing training data and training a first decision tree classification model based on consumers' individual aesthetic preferences and corresponding Censydiam consumer motivation labels; using the trained first decision tree classification model to classify consumers' individual aesthetic preferences into one of the Censydiam consumer motivations to obtain consumer motivation N1; using a 3D convolutional neural network to extract features based on consumers' action and expression data to obtain action and expression features; using a TF-IDF algorithm to extract features based on consumers' comment data to obtain comment features; constructing training data and training a second decision tree classification model based on action and expression features, comment features and corresponding Censydiam consumer motivation labels; using the trained second decision tree classification model to classify consumers' individual aesthetic preferences into one of the Censydiam consumer motivations One of the consumption motivations in Censydiam is used to obtain consumption motivation N2; consumption scenario stimulation analysis: compare consumption motivation N1 with consumption motivation N2. If N1 and N2 are the same, it is determined that the elements in the existing consumption scenario are invalid elements and a new consumption scenario needs to be generated. If N1 and N2 are different, it is determined that the existing consumption scenario has effectively stimulated consumers and there are effective elements. It is necessary to adjust the existing consumption scenario on the basis of retaining the effective elements; scenario element effectiveness analysis: when N1≠N2, use linear regression analysis or association rule analysis to obtain the relationship between multiple scenario elements and consumer behavior indicators in the existing consumption scenario, and screen out effective elements; construct a generative network model: collect image data, video data and text data related to the consumption scenario as training materials to train the generative adversarial network and obtain the trained generative adversarial network; generate new scenarios or optimize existing scenarios: input consumer demand information into the trained generative adversarial network to generate new consumption scenario images, or retain effective elements to form experience, and use the trained generative adversarial network to generate adjusted images of existing consumption scenarios.

[0006] In this technical solution, the result of K-means clustering is to form multiple clusters representing different aesthetic preferences, each cluster represents a consumer group, and each aesthetic preference corresponds to a consumer group. Consumer behavior indicators include collected consumer action expression data and purchase data. For example, by identifying the collected consumer expressions, positive expressions represent that the scene elements have a positive and effective stimulation on consumer motivation, and negative expressions represent that the scene elements have a negative impact. When the purchase volume and transaction price in the purchase data are used as consumer behavior indicators, the purchase volume and transaction price become higher, indicating that the scene elements have a positive impact, and vice versa, it means that the scene elements have a negative impact. About consumer demand: When N1=N2, the consumer demand information that needs to be input is the individual aesthetic preference of this consumer group after K-means clustering; when N1≠N2, the consumer demand information that needs to be input is the individual aesthetic preference corresponding to N2 (the individual aesthetic preference information of the 8 consumer motivations of the Censydiam model needs to be collected in advance).

[0007] According to the scene generation method based on consumer behavior psychological model and artificial intelligence disclosed by the present invention, preferably, it also includes: collecting consumer data: obtaining consumers' appearance data, purchase data, action expression data and comment data; wherein, appearance data includes facial features, skin features, body shape, clothing accessories, hairstyle and hair color, action expression data includes facial expressions, motion trajectory, body posture, interactive behavior, abnormal behavior, purchase data includes the type, quantity, unit price and payment amount of goods or services purchased by consumers, and comment data includes online comments, offline messages and complaint information; collecting scene element data: obtaining the spatial atmosphere data, spatial density data, environmental sound data, air quality data, light lighting data, guide system data, personnel flow data, theme style data, brand logo data, color pattern data, background music data, space layout data, decoration data, temperature and humidity data, sound and lighting data, air quality data, product display data, line channel data of the consumer's consumption scene.

[0008] According to the scene generation method based on consumer behavior psychological model and artificial intelligence disclosed by the present invention, preferably, the convolutional neural network is VGG or ResNet.

[0009] According to the scene generation method based on consumer behavior psychological model and artificial intelligence disclosed in the present invention, preferably, the step of dividing consumers' individual aesthetic preferences based on their appearance data using convolutional neural networks and K-means clustering algorithms specifically includes: using a pre-trained convolutional neural network to extract features from consumers' appearance images to obtain high-dimensional feature vectors; then, using the feature-extracted data as input, and applying the K-means clustering algorithm to cluster the appearance features extracted by the convolutional neural network after determining the optimal K value through the elbow rule algorithm or the silhouette coefficient algorithm, so as to classify a variety of consumer groups with different characteristics.

[0010] According to the scene generation method based on consumer behavior psychological model and artificial intelligence disclosed by the present invention, preferably, a generative adversarial network is specifically included: a generator, which generates new consumer scene images according to input consumer needs; and a discriminator, which is used to distinguish between generated images and real images.

[0011] According to the scenario generation method based on consumer behavior psychological model and artificial intelligence disclosed in the present invention, preferably, it also includes: after the new consumption scenario or the adapted consumption scenario is put into use, the effectiveness of the new scenario is evaluated based on the before-and-after comparison of consumer purchase data and customer flow data, and the generative adversarial network is fine-tuned or retrained according to the evaluation results.

[0012] According to the scenario generation method based on the consumer behavior psychological model and artificial intelligence disclosed in the present invention, preferably, the relationship between multiple scenario elements and consumer behavior indicators in the existing consumer scenario is obtained by linear regression analysis or association rule analysis, and the step of screening out effective elements specifically includes: determining the independent variable as the scenario element, determining the dependent variable as the consumer behavior indicator; assuming that the dependent variable Y is related to the independent variable X1, the independent variable X2, ..., the independent variable X n There is a linear relationship between them: that is, Y = β0+β1X1+β2X2+...+β n X n +ε, where β0 is a constant term, β1, β2, ..., β n is the regression coefficient, which reflects the degree and direction of influence of each independent variable on the dependent variable, and ε is the random error term; the optimization algorithm is used to estimate the regression coefficients β1, β2, ..., β n ; Conduct significance test on the linear regression model to determine whether the independent variable has a significant impact on the dependent variable; conduct goodness of fit test on the linear regression model to evaluate the overall fit effect and determine whether the linear regression model meets the assumptions of linear regression; determine the effective scene elements that produce positive stimulation based on the positive and negative sign and size of the regression coefficient.

[0013] In this technical solution, a positive regression coefficient indicates that the scene elements have a positive impact on consumer behavior, and a negative regression coefficient indicates that the scene elements have a negative impact on consumer behavior; a large absolute value of the regression coefficient indicates that the consumption scene elements have a significant impact on consumer behavior; a small absolute value of the regression coefficient indicates that the consumption scene elements have an insignificant impact on consumer behavior.

[0014] The second aspect of the present invention discloses a scene generation system based on a consumer behavior psychological model and artificial intelligence, comprising: a memory for storing program instructions; a processor for calling the program instructions stored in the memory to implement a scene generation method based on a consumer behavior psychological model and artificial intelligence as any of the above-mentioned technical solutions.

[0015] The beneficial effects of the present invention include at least: the present invention conducts a comprehensive analysis of consumer psychology based on multi-dimensional data, and generates more targeted and stimulating new consumption scenarios, or provides optimization solutions for existing consumption scenarios, based on consumer personal preferences and effective stimulating elements in consumption scenarios. On the one hand, it improves the psychological analysis model based on consumer behavior, providing more scientific technical support for the generation of new scenarios; on the other hand, it solves the problem of the lack of identification and retention of effective elements in existing consumption scenarios in the prior art, and provides technical support for the generation of effect optimization solutions for consumption scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A flow chart of a scene generation method based on a consumer behavior psychological model and artificial intelligence according to an embodiment of the present invention is shown.

[0017] Figure 2 A schematic block diagram of a scene generation system based on a consumer behavior psychological model and artificial intelligence according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0018] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention can also be implemented in other ways different from those described herein, and therefore, the present invention is not limited to the limitations of the specific embodiments disclosed below.

[0019] Glossary:

[0020] Censydiam consumer motivation model: analyzes consumers' psychological motivations and behavioral strategies when facing needs or desires through two dimensions - "self-adaptation" and "release-repression". The horizontal dimension of "self-adaptation" focuses on the attitude of individuals in dealing with social relationships, while the vertical dimension of "release-repression" reflects the psychological tendency of individuals in fighting against inferiority complexes. Based on the two dimensions, the types of consumer motivations are further subdivided, mainly including highlighting personal values, finding group belonging, releasing inner desires, returning to inner rationality, vitality / exploration, integration / communication, comfort / safety, and personality / uniqueness.

[0021] Convolutional Neural Networks (CNNs): Deep neural networks are mainly used to process data with a grid structure, such as image data. CNNs can automatically extract high-level feature representations from raw images by simulating the way neurons in the human brain's visual cortex process visual information, thereby performing tasks such as image recognition, classification, and detection.

[0022] K-means clustering algorithm: an algorithm that divides the points in a data set into K clusters, where each cluster is represented by the mean (center point) of all the points within it.

[0023] Decision tree classification algorithm: Classification is performed by building a model similar to a tree structure. This tree structure consists of nodes and directed edges, where each internal node represents a test condition for a feature or attribute, each branch represents the output of a test result, and each leaf node represents a category.

[0024] Linear regression analysis: used to study the linear relationship between one or more independent variables (explanatory variables) and a dependent variable (response variable). Through methods such as the least squares method, the optimal linear model parameters can be found to determine the degree of influence of the independent variables on the dependent variable.

[0025] Generative Adversarial Network (GAN): It consists of two parts: the generator and the discriminator. The generator is responsible for generating fake data that is as close to the real data as possible, while the discriminator is responsible for distinguishing whether the input data is real or generated by the generator. Through adversarial training between the two networks, GAN can learn the real distribution of data and generate high-quality data samples.

[0026] TF-IDF (Term Frequency-Inverse Document Frequency): It consists of two parts: TF (term frequency) and IDF (inverse document frequency). TF is used to measure the frequency of a word in a document, and IDF is used to measure the prevalence or uniqueness of a word in the entire document collection. The calculation method of TF-IDF is to multiply TF and IDF to get a comprehensive word weight. This weight reflects the importance of the word in the document. At the same time, TF-IDF can convert text data into feature vector representation, which is convenient for machine learning algorithms to process.

[0027] like Figure 1 As shown, according to an embodiment of the present invention, the artificial intelligence generation principle for consumption scenarios lacks in-depth attention to consumer psychology and consumption scenarios, and there is a problem of focusing on "consumer portraits" but neglecting "deep description of consumer psychology and matching of consumption scenarios". The present invention discloses a scenario generation method based on a consumer behavior psychological model and artificial intelligence, the method comprising: step S101, data collection: using professional instruments and related technologies to realize the collection of consumer data, comment data and scenario data; step S102, consumer psychology analysis: based on the collected consumer data and comment data, using the Censydiam consumer motivation model to divide consumer motivations N1 and N2, and compare N1 and N2; step S103, scenario element stimulation analysis: in the case of N1≠N2, based on the collected scenario data, regression analysis is used to obtain effective elements and invalid elements that affect consumer behavior; step S104, scenario generation and adjustment: when the scenario does not stimulate consumer behavior (N1=N2), a targeted scenario is generated based on consumer psychology analysis; when the scenario affects consumer behavior (N1≠N2), effective elements are retained and invalid elements are adjusted based on consumer psychology to achieve scenario optimization.

[0028] In this embodiment, professional monitoring equipment such as cameras and thermal imagers are used to collect consumer data (appearance and action expression data) and scene data, and relevant comment data is obtained by means of public opinion monitoring software and manual collection; then, based on the Censydiam consumer motivation model, consumer appearance data is used to divide consumer motivation N1, and consumer action expression data and relevant comment data are used to divide consumer motivation N2, and N1 and N2 are compared to draw conclusions on the effectiveness of consumer scene stimulation; then, a linear regression algorithm is used to obtain scene elements that affect consumer behavior in the consumer environment, and effective scene elements are retained and formed into experience, and invalid scene elements are optimized and transformed using artificial intelligence algorithms. This method can identify and retain effective elements in existing consumer scenes, provide targeted improvement plans for the optimization of existing consumer scenes, generate more targeted and stimulating consumer scenes, and provide feasible selection plans for the generation of new consumer scenes.

[0029] According to the scene generation method based on the consumer behavior psychological model and artificial intelligence disclosed in the above embodiment, preferably, step S101 specifically includes: (1) consumer data collection: using surveillance cameras, motion capture equipment, payment settlement counter recording equipment, etc., respectively collect consumer appearance data, motion expression data and purchase data. The appearance data includes facial features, skin features, body shape, clothing accessories, hairstyle and hair color, etc., the motion expression data includes facial expressions, movement trajectories, body postures, interactive behaviors, abnormal behaviors, etc., and the purchase data includes the type, quantity, unit price, payment amount, etc. of the goods or services purchased by the consumer. Collect data such as consumer online comments, offline messages and government department complaints by using public opinion monitoring system software, web crawler tools or manual collection methods. (2) Consumption scene data collection: Use surveillance cameras, thermal imagers, intelligent temperature and humidity monitoring equipment, color sensors, microphones and other equipment to collect the main scene elements in external open scenes, internal and external transition scenes, and internal consumption scenes. The external open scene elements include space atmosphere, space density, ambient sound, air quality, lighting, guide system, etc. The internal and external transition scene elements include personnel flow, theme style, brand logo, lighting, color pattern, background music, etc. The internal consumption scene elements include space layout, decoration, temperature and humidity, sound and lighting, air quality, background music, color pattern, product display, line channel, etc.

[0030] According to the scenario generation method based on the consumer behavior psychological model and artificial intelligence disclosed in the above embodiment, preferably, step S102 specifically includes:

[0031] S102-1: Based on the appearance data of consumers, convolutional neural network (CNN) and K-means clustering algorithm are used to divide consumers' individual aesthetic preferences. The specific method is as follows: First, use pre-trained convolutional neural network (CNN) models such as VGG and ResNet to extract features from the appearance image data of consumers collected by S101 to obtain high-dimensional feature vectors; then, use the extracted data as input, determine the optimal K value through the elbow rule, silhouette coefficient and other methods, and then use the K-means clustering algorithm to cluster the appearance data extracted by the CNN model to obtain consumer groups with different characteristics (such as dressing style, brand preference, etc.).

[0032] S102-2: Based on the theory of Censydiam consumer motivation model and the individual aesthetic preferences of consumers, the consumer motivation N1 is obtained by using the decision tree classification algorithm. The specific method is as follows: First, a data set of known consumer motivations (including consumer characteristics and the corresponding 8 censydiam consumer motivation labels) is collected as a training set and a test set; then, the decision tree model is trained using the training set to form classification rules, and the performance of the decision tree model is evaluated and optimized using the test set; finally, the data after K-means clustering in step S102-1 is input into the trained decision tree model, and the group is divided into consumer motivation N1 among the 8 consumer motivations of censydiam.

[0033] S102-3: Process the data using 3D convolutional neural network (3D CNN) based on the consumer action and expression data. The specific method is as follows: First, extract the action and expression data of consumers in different consumer groups after K-means clustering in step S102-1; then, use the pre-trained 3D convolutional neural network (3D CNN) model to extract features from the consumer action and expression data to obtain high-dimensional feature vectors, such as purchase categories, browsing time, dwell time, rebellious behavior, herd behavior, impulse buying behavior, and comments that the price is too high.

[0034] S102-4: Process the data based on the consumer-related comment data using text vectorization. The specific method is as follows: First, extract the relevant comment data of consumers in different consumer groups after K-means clustering in step S102-1; then, perform text preprocessing on the consumer comment data, including cleaning, word segmentation, and removal of stop words; finally, use the TF-IDF method to extract the TF-IDF values ​​of words related to consumer actions, purchasing behavior, price, satisfaction, etc. in the text data of consumer-related comments and construct feature vectors to facilitate machine learning processing.

[0035] S102-5: Based on the processed consumer action and expression data and related comment data according to the Censydiam consumer motivation model theory, the consumer's consumption motivation N2 is obtained using the decision tree classification algorithm. The specific method is: first, collect a data set of known consumer consumption motivations (including consumer behavior, expression, evaluation and the corresponding 8 censydiam consumer motivation labels, that is, map the consumer's action and expression characteristics to the 8 dimensions of consumer motivations of the Censydiam model, such as the degree of positivity of facial expressions reflects the consumer's interest and desire, the frequency of behavioral actions reflects the consumer's willingness to buy, and the type and price of purchased items reflect the consumer's consumption ability and preference, etc.) as a training set and a test set; then, use the training set to train the decision tree model to form classification rules, and use the test set to evaluate and optimize the performance of the decision tree model; finally, input the data after K-means clustering in step S102-1 into the trained decision tree model, and divide the group into consumer motivation N2 among the 8 consumer motivations of censydiam.

[0036] In this embodiment, the underlying logic of consumer psychology analysis is to analyze consumer groups divided by the K-means clustering algorithm. The clustered consumer groups are divided based on appearance data, and the groups have the same aesthetic preferences and consumption motivations N1; S102-3 is to analyze the action expression data and related comment data of consumers within the clustered consumer groups to explore whether the consumption scene elements have an impact on the consumption motivation of this group, and finally derive the consumption motivation N2.

[0037] According to the scene generation method based on the consumer behavior psychological model and artificial intelligence disclosed in the above embodiment, preferably, step S103 specifically includes: comparing consumer motivations N1 and N2. If N1 and N2 are the same, it means that the existing scene does not produce effective stimulation to consumers, and the existing scene elements are invalid elements; if N1 and N2 are different, it means that the existing scene has produced effective stimulation to consumers and there are effective elements. When consumer motivations N1 and N2 are different, linear regression is used to analyze the relationship between the elements corresponding to the three different consumer scenarios and the consumer actions in the scenarios, and it is found out which elements in the scene will affect consumer behavior. The specific method used is:

[0038] First, determine that the independent variable is the consumption scenario element and the dependent variable is the consumer behavior indicator. Assume that the dependent variable (Y) is related to the independent variable (X1, X2, ..., X n ), which can be expressed as: Y=β0+β1X1+β2X2+...+β n X n +ε, where β0 is a constant term, β1, β2, ..., β nis the regression coefficient, ε is the random error term, and the least squares method (OLS) or other optimization algorithms are used to estimate each regression coefficient. The coefficient β reflects the degree and direction of the influence of each variable on the dependent variable. Next, the model is subjected to a significance test (such as a t-test) to determine whether each independent variable has a significant effect on the dependent variable, and a goodness-of-fit test (such as an R² test) to evaluate the overall fit of the model and check whether the model meets the basic assumptions of linear regression (such as normality, independence, homoscedasticity, etc. of the residuals); finally, according to the positive and negative and size of the regression coefficient, the effective scene elements that produce positive stimulation, the invalid scene elements that produce negative stimulation, and the invalid scene elements that do not produce stimulation are obtained.

[0039] According to the scene generation method based on the consumer behavior psychological model and artificial intelligence disclosed in the above embodiment, preferably, step S104 specifically includes: based on the consumer's consumption motivations N1 and N2, using the generative adversarial network (GAN) in the generation model to achieve scene generation or adaptation optimization, retaining and forming experience for effective scene elements that produce positive stimulation to consumers, and optimizing the design of invalid scene elements that produce negative stimulation to consumers or no stimulation. The specific method used is:

[0040] First, data collection and preprocessing are carried out. Images, videos, text descriptions and other data related to consumption scenarios covering different styles, elements and scenarios are collected through web crawlers, public data sets, coordination and cooperation, etc. The collected data are preprocessed by cleaning, denoising and other preprocessing work to ensure the quality and availability of the data. According to the requirements of the GAN model, the data is converted into a format suitable for model training; then, the GAN model is constructed and trained. The design of the GAN model includes two parts: the generator and the discriminator. The generator is responsible for generating new consumption scene images according to the input consumer needs, while the discriminator is responsible for distinguishing the generated images from the real images. The GAN model is trained with the preprocessed data. During the training process, the parameters of the generator and the discriminator are continuously adjusted to optimize the generation effect and discrimination ability of the model. Finally, the consumption scene is generated or optimized, and the trained GAN model is used to generate new consumption scene images according to consumer needs. After the new consumption scene generated or optimized is actually applied, the effectiveness of the new scene can be evaluated based on the before-and-after comparison of consumer purchase data and passenger flow data. Later, the GAN model can be fine-tuned or retrained according to the evaluation results to optimize the generation effect and play a greater role in actual application.

[0041] like Figure 2As shown, according to another embodiment of the present invention, a scene generation system 200 based on a consumer behavior psychological model and artificial intelligence is also disclosed, including: a memory 201, used to store program instructions; a processor 202, used to call the program instructions stored in the memory to implement the scene generation method based on the consumer behavior psychological model and artificial intelligence as in the above-mentioned embodiment.

[0042] All or part of the steps in the various methods of the above embodiments can be completed by controlling the relevant hardware through a program, and the program can be stored in a readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable programmable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other readable medium that can be used to carry or store data.

[0043] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A scenario generation method based on consumer behavior psychological model and artificial intelligence, characterized in that: include: Divide consumer groups: Based on consumers’ appearance data, use convolutional neural networks and K-means clustering algorithms to divide consumers’ individual aesthetic preferences and form multiple consumer groups; Two classification models are used to obtain consumption motivations for the same consumer group: training data is constructed based on the consumer's individual aesthetic preferences and the corresponding Censydiam consumption motivation labels, and the first decision tree classification model is trained; the trained first decision tree classification model is used to classify the consumer's individual aesthetic preferences into one of the Censydiam consumption motivations, and the consumption motivation N1 is obtained; based on the consumer's action and expression data, a 3D convolutional neural network is used to extract features to obtain action and expression features; Based on the consumer review data, the TF-IDF algorithm is used to extract features and obtain review features; Construct training data based on the action expression features, comment features and corresponding Censydiam consumer motivation labels and train the second decision tree classification model; use the trained second decision tree classification model to classify the consumer's individual aesthetic preference into one of the Censydiam consumer motivations, and obtain the consumer motivation N2; Analysis of the stimulus of consumption scenarios: Compare consumption motivation N1 with consumption motivation N2. If N1 and N2 are the same, it is determined that the elements in the existing consumption scenario are invalid elements and a new consumption scenario needs to be generated. If N1 and N2 are different, it is determined that the existing consumption scenario has effectively stimulated consumers and there are effective elements. It is necessary to adjust the existing consumption scenario on the basis of retaining the effective elements. Scenario element validity analysis: When N1≠N2, use linear regression analysis or association rule analysis to obtain the relationship between multiple scenario elements and consumer behavior indicators in existing consumption scenarios, and screen out effective elements; Construct a generative network model: collect image data, video data, and text data related to consumption scenarios as training materials to train the generative adversarial network and obtain the trained generative adversarial network; Generate new scenarios or optimize existing scenarios: input consumer demand information into the trained generative adversarial network to generate new consumer scenario images, or retain the effective elements to form experience, and use the trained generative adversarial network to generate adapted images of existing consumer scenarios.

2. The scenario generation method based on consumer behavior psychological model and artificial intelligence according to claim 1 is characterized in that: Also includes: Collecting consumer data: Obtaining consumers' appearance data, purchase data, action and expression data, and comment data; wherein the appearance data includes facial features, skin features, body shape, clothing accessories, hairstyle and hair color; the action and expression data includes facial expressions, movement trajectories, body postures, interactive behaviors, and abnormal behaviors; the purchase data includes the type, quantity, unit price, and payment amount of goods or services purchased by consumers; and the comment data includes online comments, offline messages, and complaint information; Collect scene element data: obtain the spatial atmosphere data, spatial density data, environmental sound data, air quality data, lighting data, guide system data, personnel flow data, theme style data, brand logo data, color pattern data, background music data, space layout data, decoration data, temperature and humidity data, sound and lighting data, air quality data, product display data, and line channel data of the consumption scene where the consumers are located.

3. The scenario generation method based on consumer behavior psychological model and artificial intelligence according to claim 1 is characterized in that: The convolutional neural network is VGG or ResNet.

4. The scenario generation method based on consumer behavior psychological model and artificial intelligence according to claim 1 is characterized in that: The step of using a convolutional neural network and a K-means clustering algorithm to divide consumers' individual aesthetic preferences based on their appearance data specifically includes: Use a pre-trained convolutional neural network to extract features from consumers' appearance images to obtain high-dimensional feature vectors. Then, use the extracted data as input, determine the optimal K value through the elbow rule algorithm or the silhouette coefficient algorithm, and then use the K-means clustering algorithm to cluster the appearance features extracted by the convolutional neural network to classify various consumer groups with different characteristics.

5. The scenario generation method based on consumer behavior psychological model and artificial intelligence according to claim 1 is characterized in that: The generative adversarial network specifically includes: A generator, wherein the generator generates a new consumption scene image according to input consumer demand; The discriminator is used to distinguish generated images from real images.

6. The scenario generation method based on consumer behavior psychological model and artificial intelligence according to claim 1 is characterized in that: Also includes: After the new consumption scenario or adjusted consumption scenario is put into use, the effectiveness of the new scenario is evaluated based on the before-and-after comparison of consumer purchase data and customer flow data, and the generative adversarial network is fine-tuned or retrained based on the evaluation results.

7. The scenario generation method based on consumer behavior psychological model and artificial intelligence according to claim 1 is characterized in that: The steps of using linear regression analysis or association rule analysis to obtain the relationship between multiple scene elements and consumer behavior indicators in existing consumption scenes and screen out effective elements include: Determine the independent variables as scenario elements and the dependent variables as consumer behavior indicators; Assume that the dependent variable Y is related to the independent variables X1, X2, ..., X n There is a linear relationship between them: that is, Y = β0+β1X1+β2X2+...+β n X n +ε, where β0 is a constant term, β1, β2, ..., β n is the regression coefficient, which reflects the degree and direction of influence of each independent variable on the dependent variable, and ε is the random error term; Use an optimization algorithm to estimate the regression coefficients β1, β2, ..., β n ; Conduct significance test on linear regression model to determine whether the independent variable has significant impact on the dependent variable; Performing a goodness of fit test on the linear regression model to evaluate the overall fitting effect and determine whether the linear regression model meets the assumptions of linear regression; According to the sign and size of the regression coefficient, the effective scene elements that produce positive stimulation are determined.

8. A scenario generation system based on consumer behavior psychological model and artificial intelligence, characterized in that: include: A memory for storing program instructions; A processor, used to call the program instructions stored in the memory to implement the scene generation method based on the consumer behavior psychological model and artificial intelligence as described in any one of claims 1 to 7.

Images