KR102596352B1 - AI-based product recommendation system based on user psychology - Google Patents

Abstract

According to one embodiment, in a product recommendation system based on user psychology through artificial intelligence, a system including a server that recommends products based on user psychology through artificial intelligence is provided.

Description

Product recommendation system based on user psychology through artificial intelligence {AI-based product recommendation system based on user psychology}

The present invention relates to technology for providing a product recommendation system based on user psychology through artificial intelligence.

In general, technology for recommending user-tailored products involves recommending other products similar to the product the user previously preferred, or calculating the similarity of multiple users and recommending products preferred by other users with high similarity.

In addition, a method of recommending customized products to the user was used based on information such as the user's gender, age, intention to purchase the product, and region of residence.

However, although these recommendation systems can recommend products by analyzing the user's preferences, there is a problem in that they cannot take into account the user's psychological state according to the user's personal situation, which can change every day.

Additionally, there was a problem in recommending newly registered products because there was no purchase data.

Korean Patent No. 10-2093087 Korean Patent Publication No. 10-2023-0030870 Korean Patent No. 10-1846422 Korean Patent No. 10-2109338

According to one embodiment, the purpose of the product recommendation system based on user psychology through artificial intelligence is to provide a system that includes a server that recommends products based on user psychology through artificial intelligence.

In addition, the server includes a processor and memory, and the processor operates a website that recommends products based on user psychology through artificial intelligence. When a user account is logged in to the website through a user terminal, An interface and a guide image for acquiring first data are provided to the user terminal, wherein the first data is data about the user's today's psychological state written through the user terminal, and the first data is provided through the user terminal. Obtaining and providing an interface and a guide image for obtaining second data to the user terminal, wherein the second data is data about the user's psychological characteristics written through the user terminal, and is provided through the user terminal. Obtaining second data, inputting the first data to a first artificial intelligence model pre-trained to predict a preferred clothing item based on data on today's psychological state to obtain a first output, and obtaining a first output, The data is input to a second artificial intelligence model that is pre-trained to predict preferred clothing items based on data on psychological characteristics to obtain a second output, and a first clothing item is provided to the user terminal based on the first output. A second clothing item is recommended based on the item and the second output, and when a clothing item is purchased through the user terminal, the first data and the purchased clothing item are input into the first artificial intelligence model for continuous learning ( Continual Learning), and a system for continuous learning by inputting the second data and the purchased clothing item into the second artificial intelligence model can be provided.

In addition, the first clothing item and the second clothing item are each a plurality of clothing items, and the processor sends the user terminal a first clothing item based on the first output and a second clothing item based on the second output. When recommending an item, a system can be provided that sorts and displays a plurality of first and second clothing items in the order of their highest evaluation scores.

In addition, the first data is created in the same interface every time through the user terminal, and includes today's good things and their degree, difficult things and their degree, meal record, exercise record, sleep record, one-line review of each record, and the user terminal. Includes the color of each record specified through, and the second data provides different activities for each day of the week in order to obtain various psychological characteristics of the user through the user terminal, but is composed of one or more of pictures, colors, and text. Expressed, the above activities can provide a system that includes one of the following: expressing one's mind with colors, recording and matching colors by life cycle, writing a letter to oneself, writing a vision, checking habits, and freely expressing oneself through drawings.

In addition, the processor inputs review data for each clothing item into a third artificial intelligence model pre-trained to analyze reviews of clothing items, obtains a third output that is review analysis data for each clothing item, and obtains a third output, which is review analysis data for each clothing item. The artificial intelligence model includes first data acquired through a plurality of user terminals, a list of clothing items purchased through each user terminal on the date of acquiring the first data, ratings of all clothing items, and the third data of each clothing item. It is pre-trained with a dataset consisting of outputs, and the second artificial intelligence model includes second data acquired through a plurality of user terminals, purchase history of each user account, ratings of all clothing items, and third data of each clothing item. It is possible to provide a system that is pre-trained with a dataset consisting of output.

In addition, the first artificial intelligence model includes a multi-layer perceptron, a recurrent neural network (RNN), a convolutional neural network (CNN), and a recurrent convolutional neural network (R-CNN). It is a multi-input model based on one or more of the convolutional neural network) algorithm, and the second artificial intelligence model is an image processing model based on one of the CNN, Faster R-CNN, and YOLO algorithms; and an ensemble model including a multi-input model based on one or more of the multi-layer perceptron, recursive neural network, CNN, and R-CNN algorithms, and the third artificial intelligence model is SVM (Support Vector Machine) and Naive Bayes. , Neural Networks, and a system that is a Sentiment Analysis model based on one of the Decision Tree algorithms can be provided.

In addition, the dataset for pre-training the first artificial intelligence model includes a list of clothing items viewed through each user terminal on the date the first data was acquired, and each user terminal on the date the first data was acquired. It further includes a list of clothing items contained in the shopping cart, and the dataset for pre-training the second artificial intelligence model is clothing items with the top 5 views among the list of clothing items viewed through the user terminal, and the user It is possible to provide a system that further includes the top five clothing items that have been in the shopping cart for a long time among the list of clothing items in the shopping cart through the terminal.

In addition, the processor, before recommending a first clothing item based on the first output to the user terminal and a second clothing item based on the second output, recommends the first clothing item to the third artificial intelligence model. and input review data of the second clothing item to obtain review analysis data of the first clothing item and the second clothing item, and the current rating of the first clothing item is 2.0 or less, or the review analysis data has a negative sentiment. In this case, the rating of the clothing item with negative sentiment is set to 0, the first clothing item is re-predicted based on the first artificial intelligence model, and the current rating of the second clothing item is 2.0 or less, or review analysis data If is a negative emotion, the rating of the clothing item with negative emotion can be designated as 0, and a system for re-predicting the second clothing item based on the second artificial intelligence model can be provided.

In addition, when the processor recommends to the user terminal a first clothing item based on the first output and a second clothing item based on the second output, the processor recommends a plurality of the first clothing items to the user terminal. A photo of the model wearing the item and a plurality of photos of the model wearing the second clothing item are provided, and the plurality of photos of the model wearing the first clothing item and the second clothing item are automatically switched in the order in which the model moves. A display system can be provided.

In addition, when the processor recommends to the user terminal a first clothing item based on the first output and a second clothing item based on the second output, the processor recommends to the user terminal a person wearing the first clothing item. A system that provides an image of a model moving and an image of a model wearing the second clothing item moving can be provided.

Additionally, the processor may provide a system that transmits a writing notification message to the user terminal when 30 hours have elapsed since the first and second data were last acquired through the user terminal.

In addition, when the processor acquires first data and second data every day for 28 days through the user terminal, the processor sorts all the first data and second data in order of creation date, produces a digital booklet, and stores it in the user terminal. A system for providing a discount coupon for a first clothing item and a second clothing item to the user terminal may be provided.

In addition, if the processor acquires the first data and second data obtained every day through the user terminal between midnight and 9 a.m. of the next day on which the first data and second data must be acquired, the first data and second data are acquired every day. A system can be provided that determines that first data and second data were acquired on the date they should be acquired.

In addition, after obtaining first data through the user terminal, the processor provides the user terminal with today's color emotion analysis result based on the color included in the first data, wherein the color emotion analysis result is is acquired using a color emotion analysis API, the color emotion analysis API is Brandfetch Color Psychology API or ColorTag API, and the processor pre-assigns keywords included in the color emotion analysis results to each of the entire clothing items. A system can be provided that matches a label and displays clothing items with a high match on the user terminal.

In addition, the process calculates an evaluation score of each of the plurality of first and second clothing items based on [Equation 1],

[Equation 1]

The SCORE is the evaluation score of the clothing item, x_rt is the rating of the clothing item, x_vw is the number of views of the clothing item, x_sl is the sales volume of the clothing item, and x_rv is the number of reviews of the clothing item. can do.

By providing a product recommendation system based on user psychology through artificial intelligence according to the present invention, clothing items can be recommended based on the user's personal psychology that changes every day, and new clothing items with no previous purchase history can also be recommended. , It is easy to continuously obtain data about the user's psychological state and psychological characteristics, and each time the user purchases a clothing item, the performance can be updated by learning an artificial intelligence model, and the user can write about the psychological state and psychological characteristics. You can achieve psychological stability or psychotherapy effects by going out, you can provide users with photos or videos of dynamic clothing items worn, you can instill in users a sense of purpose for recording their own psychology, and you can receive multiple recommended clothing items. This has the effect of sorting and providing items in order of highest evaluation.

The object of the present invention is not limited to the object mentioned above, and other objects not mentioned can be clearly understood from the description below.

According to one embodiment, in a product recommendation system based on user psychology through artificial intelligence, a system including a server that recommends products based on user psychology through artificial intelligence is provided.

In addition, the server includes a processor and memory, and the processor operates a website that recommends products based on user psychology through artificial intelligence. When a user account is logged in to the website through a user terminal, An interface and a guide image for acquiring first data are provided to the user terminal, wherein the first data is data about the user's today's psychological state written through the user terminal, and the first data is provided through the user terminal. Obtaining and providing an interface and a guide image for obtaining second data to the user terminal, wherein the second data is data about the user's psychological characteristics written through the user terminal, and is provided through the user terminal. Obtaining second data, inputting the first data to a first artificial intelligence model pre-trained to predict a preferred clothing item based on data on today's psychological state to obtain a first output, and obtaining a first output, The data is input to a second artificial intelligence model that is pre-trained to predict preferred clothing items based on data on psychological characteristics to obtain a second output, and a first clothing item is provided to the user terminal based on the first output. A second clothing item is recommended based on the item and the second output, and when a clothing item is purchased through the user terminal, the first data and the purchased clothing item are input into the first artificial intelligence model for continuous learning ( Continual Learning), and the second data and the purchased clothing item can be input into the second artificial intelligence model for continuous learning.

In addition, the first clothing item and the second clothing item are each a plurality of clothing items, and the processor sends the user terminal a first clothing item based on the first output and a second clothing item based on the second output. When recommending an item, a plurality of first clothing items and a plurality of second clothing items may be displayed by sorting them in descending order of evaluation scores.

In addition, the first data is created in the same interface every time through the user terminal, and includes today's good things and their degree, difficult things and their degree, meal record, exercise record, sleep record, one-line review of each record, and the user terminal. Includes the color of each record specified through, and the second data provides different activities for each day of the week in order to obtain various psychological characteristics of the user through the user terminal, but is composed of one or more of pictures, colors, and text. Expressed, the above activities may include one of the following: expressing one's mind with colors, recording each life cycle and matching colors, writing a letter to oneself, writing a vision, checking habits, and freely expressing oneself through drawings.

In addition, the processor inputs review data for each clothing item into a third artificial intelligence model pre-trained to analyze reviews of clothing items, obtains a third output that is review analysis data for each clothing item, and obtains a third output, which is review analysis data for each clothing item. The artificial intelligence model includes first data acquired through a plurality of user terminals, a list of clothing items purchased through each user terminal on the date of acquiring the first data, ratings of all clothing items, and the third data of each clothing item. It is pre-trained with a dataset consisting of outputs, and the second artificial intelligence model includes second data acquired through a plurality of user terminals, purchase history of each user account, ratings of all clothing items, and third data of each clothing item. It can be pre-trained with a dataset consisting of output.

In addition, the first artificial intelligence model includes a multi-layer perceptron, a recurrent neural network (RNN), a convolutional neural network (CNN), and a recurrent convolutional neural network (R-CNN). It is a multi-input model based on one or more of the convolutional neural network) algorithm, and the second artificial intelligence model is an image processing model based on one of the CNN, Faster R-CNN, and YOLO algorithms; and an ensemble model including a multi-input model based on one or more of the multi-layer perceptron, recursive neural network, CNN, and R-CNN algorithms, and the third artificial intelligence model is SVM (Support Vector Machine) and Naive Bayes. , Neural Networks, or it may be a Sentiment Analysis model based on one of the Decision Tree algorithms.

In addition, the dataset for pre-training the first artificial intelligence model includes a list of clothing items viewed through each user terminal on the date the first data was acquired, and each user terminal on the date the first data was acquired. It further includes a list of clothing items contained in the shopping cart, and the dataset for pre-training the second artificial intelligence model is clothing items with the top 5 views among the list of clothing items viewed through the user terminal, and the user Among the list of clothing items in the shopping cart through the terminal, you can further include the top 5 clothing items that have been in the shopping cart for a long time.

In addition, the processor, before recommending a first clothing item based on the first output to the user terminal and a second clothing item based on the second output, recommends the first clothing item to the third artificial intelligence model. and input review data of the second clothing item to obtain review analysis data of the first clothing item and the second clothing item, and the current rating of the first clothing item is 2.0 or less, or the review analysis data has a negative sentiment. In this case, the rating of the clothing item with negative sentiment is set to 0, the first clothing item is re-predicted based on the first artificial intelligence model, and the current rating of the second clothing item is 2.0 or less, or review analysis data If is a negative emotion, the rating of the clothing item with negative emotion can be designated as 0, and the second clothing item can be re-predicted based on the second artificial intelligence model.

In addition, when the processor recommends to the user terminal a first clothing item based on the first output and a second clothing item based on the second output, the processor recommends a plurality of the first clothing items to the user terminal. A photo of the model wearing the item and a plurality of photos of the model wearing the second clothing item are provided, and the plurality of photos of the model wearing the first clothing item and the second clothing item are automatically switched in the order in which the model moves. It can be displayed.

In addition, when the processor recommends to the user terminal a first clothing item based on the first output and a second clothing item based on the second output, the processor recommends to the user terminal a person wearing the first clothing item. An image of a model moving and an image of a model wearing the second clothing item moving may be provided.

Additionally, the processor may transmit a writing notification message to the user terminal when 30 hours have elapsed since the first and second data were last acquired through the user terminal.

In addition, when the processor acquires first data and second data every day for 28 days through the user terminal, the processor sorts all the first data and second data in order of creation date, produces a digital booklet, and stores it in the user terminal. A system may be provided that provides a discount coupon for a first clothing item and a second clothing item to the user terminal.

In addition, if the processor acquires the first data and second data obtained every day through the user terminal between midnight and 9 a.m. of the next day on which the first data and second data must be acquired, the first data and second data are acquired every day. It can be determined that the first data and the second data were acquired on the date they were supposed to be acquired.

In addition, after obtaining first data through the user terminal, the processor provides the user terminal with today's color emotion analysis result based on the color included in the first data, wherein the color emotion analysis result is is acquired using a color emotion analysis API, the color emotion analysis API is Brandfetch Color Psychology API or ColorTag API, and the processor pre-assigns keywords included in the color emotion analysis results to each of the entire clothing items. Clothing items that match a label and have a high match can be displayed on the user terminal.

In addition, the process calculates an evaluation score of each of the plurality of first and second clothing items based on [Equation 1],

[Equation 1]

The SCORE may be an evaluation score of the clothing item, x_rt may be the rating of the clothing item, x_vw may be the number of views of the clothing item, x_sl may be the sales volume of the clothing item, and x_rv may be the number of reviews of the clothing item.

According to one embodiment, by providing a product recommendation system based on user psychology through artificial intelligence, there is an effect of recommending clothing items based on the user's personal psychology that changes every day.

Additionally, it has the effect of recommending new clothing items that have no previous purchase history.

Additionally, it has the effect of continuously acquiring data on the user's psychological state and psychological characteristics and using it to recommend products.

Additionally, each time a user purchases a clothing item, the performance can be updated by training the artificial intelligence model.

Additionally, users can achieve psychological stability or psychotherapy effects by writing about their psychological state and psychological characteristics.

Additionally, it has the effect of providing users with dynamic photos or videos of clothing items being worn.

Additionally, it has the effect of providing multiple recommended clothing items sorted in order of highest evaluation.

Meanwhile, the effects according to the embodiments are not limited to those mentioned above, and other effects not mentioned can be clearly understood by those skilled in the art from the description below.

Figure 1 is a conceptual diagram schematically showing the configuration of a product recommendation system based on user psychology through artificial intelligence according to an embodiment.
Figure 2 is a diagram illustrating the configuration of a server that recommends products based on user psychology through artificial intelligence according to an embodiment.
Figure 3 is a flow chart to explain the operation process of a product recommendation system based on user psychology through artificial intelligence according to an embodiment.
Figure 4 is a flow chart to explain the process of re-predicting clothing products in a product recommendation system based on user psychology through artificial intelligence according to an embodiment.
Figure 5 is a flow chart to explain the process of analyzing and processing data related to clothing products collected by a product recommendation system based on user psychology through artificial intelligence according to an embodiment.

Hereinafter, embodiments will be described in detail with reference to the attached drawings. However, the embodiments will be described in detail below with reference to the accompanying drawings. However, various changes can be made to the embodiments, so the scope of the patent application is not limited or limited by these embodiments. It should be understood that all changes, equivalents, or substitutes for the embodiments are included in the scope of rights.

Specific structural or functional descriptions of the embodiments are disclosed for illustrative purposes only and may be modified and implemented in various forms. Accordingly, the embodiments are not limited to the specific disclosed form, and the scope of the present specification includes changes, equivalents, or substitutes included in the technical spirit.

Terms such as first or second may be used to describe various components, but these terms should be interpreted only for the purpose of distinguishing one component from another component. For example, a first component may be named a second component, and similarly, the second component may also be named a first component.

When a component is referred to as being “connected” to another component, it should be understood that it may be directly connected or connected to the other component, but that other components may exist in between.

The terms used in the examples are for descriptive purposes only and should not be construed as limiting. Singular expressions include plural expressions unless the context clearly dictates otherwise. In this specification, terms such as “comprise” or “have” are intended to designate the presence of features, numbers, steps, operations, components, parts, or combinations thereof described in the specification, but are not intended to indicate the presence of one or more other features. It should be understood that this does not exclude in advance the possibility of the existence or addition of elements, numbers, steps, operations, components, parts, or combinations thereof.

Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by a person of ordinary skill in the technical field to which the embodiments belong. Terms defined in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning in the context of the related technology, and unless explicitly defined in the present application, should not be interpreted in an ideal or excessively formal sense. No.

In addition, when describing with reference to the accompanying drawings, identical components will be assigned the same reference numerals regardless of the reference numerals, and overlapping descriptions thereof will be omitted. In describing the embodiments, if it is determined that detailed descriptions of related known technologies may unnecessarily obscure the gist of the embodiments, the detailed descriptions are omitted.

Embodiments may be implemented in various types of products such as personal computers, laptop computers, tablet computers, smart phones, televisions, smart home appliances, intelligent vehicles, kiosks, and wearable devices.

Figure 1 is a conceptual diagram schematically showing the configuration of a product recommendation system based on user psychology through artificial intelligence according to an embodiment, and Figure 2 is a schematic diagram of a server recommending products based on user psychology through artificial intelligence according to an embodiment. It is a drawing for explaining the configuration, and FIG. 3 is a flowchart for explaining the operation process of a product recommendation system based on user psychology through artificial intelligence according to an embodiment, and FIG. 4 is a diagram for explaining the operation process of a product recommendation system based on user psychology through artificial intelligence according to an embodiment. This is a flowchart to explain the process of re-predicting clothing products in a psychology-based product recommendation system, and Figure 5 shows an analysis of clothing product-related data collected by a product recommendation system based on user psychology through artificial intelligence according to an embodiment. This is a flow chart to explain the processing process.

Referring to FIG. 1, as an example, the product recommendation system based on user psychology through artificial intelligence according to the present invention may include a server 100 that recommends products based on user psychology through artificial intelligence.

The server can operate a website to recommend products based on user psychology through artificial intelligence and provide the website to multiple user terminals. The website allows a user to log in, receive data on today's psychological state and psychological characteristics, and display recommended clothing items predicted based on today's psychological state and psychological characteristics on the web page screen. The analysis results of today's psychological state are displayed on the web page screen, recommended clothing items matched based on the analysis results of the color selected by the user for today's psychology are displayed on the web page screen, and a model wearing the clothing item is displayed. It can display images dynamically, repeatedly play short videos of models wearing clothing items, and provide various functions and interfaces, such as purchasing clothing items. Additionally, clothing items can be posted on a web page by category, detailed information about each clothing item can be displayed, and functions for selling clothing items can be provided.

Additionally, referring to FIG. 2, as an example, the server 100 according to the present invention may include a processor 110 and a memory 120. The processor and memory can interact with each other to perform various functions to recommend products based on user psychology through artificial intelligence.

The processor 110 performs various functions such as data processing, manipulation, calculation, analysis, and algorithm execution, and the memory 120 can perform functions such as storing, reading, and writing data.

The processor 110 can operate a website that recommends products based on user psychology through artificial intelligence.

As an embodiment, referring to FIG. 3, when a user account is logged in to the website through the user terminal in step S301, the processor may provide an interface and a guide image for obtaining first data to the user terminal. And, first data can be obtained through the user terminal in step S302.

The first data may be data about the user's today's psychological state created through the user terminal. The first data is created in the same interface every time through the user terminal, and is specified through the user terminal, such as the good things of the day and their degree, the difficult things and their degree, meal record, exercise record, sleep record, one-line review of each record, and the user terminal. The color of each record may be included. The degree to which something was good or difficult can also be displayed using tools such as a slide bar. The meal record may include information such as the food eaten today, the exercise record may include information such as today's exercise record, and the sleep record may include information such as the sleep time last night. In addition, it is possible to leave a short self-evaluation for each record, but it would be desirable to guide the user to write it as positively as possible. And you can freely choose the color that comes to mind by looking at your exercise records, meal records, and sleep records.

The color selected by the user can be obtained by converting it to RGB values. Since the colors seen through the user's terminal may be different from the actual colors, guide colors such as red, yellow, blue, green, black, and white can be provided to allow the user to adjust the brightness or saturation of the terminal.

Additionally, the processor 110 may provide an interface and a guide image for acquiring second data to the user terminal in step S303, and may obtain second data through the user terminal in step S304.

The second data provides different activities for each day of the week in order to obtain various psychological characteristics of the user through the user terminal, and may be expressed in one or more of pictures, colors, and text, and the activities are expressed in colors. You can include one of the following: expression, recording and color matching for each life cycle, a letter to yourself, writing a vision, checking habits, or freely expressing yourself with a picture.

For example, choosing a color that can express me, choosing a color that I think I need, stretching after stretching or guided massage, recording the state of my body and mind after the massage, in the fetus, infancy, adolescence, adulthood, and old age. Briefly record your imagination or past information, select colors that come to mind for each life cycle, select life graphs and colors for each section, color a given non-colored picture using the coloring interface, check fatigue, check eating habits, sleep Checking your habits, checking your exercise habits, writing down the positive aspects of each habit, creating a mind map with yourself as the subject, writing a bucket list, finding the direction you want to go in the future, drawing a picture of yourself as a child, meditating, writing a letter to yourself, appreciating art. It may include the following.

The guide video may be composed to help you feel comfortable using the sights and sounds of nature, and may include a guidance voice to help you create first data or second data.

The second data may be data about the user's psychological characteristics created through the user terminal.

In one embodiment, the processor 110 inputs the first data in step S305 to a first artificial intelligence model that has been previously trained to predict a preferred clothing item based on data on today's psychological state to produce a first output. It can be obtained.

Additionally, in step S306, the processor may obtain a second output by inputting the second data into a second artificial intelligence model that has been pre-trained to predict a preferred clothing item based on data on psychological characteristics.

The processor may recommend a first clothing item based on the first output and a second clothing item based on the second output to the user terminal in step S307.

In addition, when a clothing item is purchased through the user terminal in step S308, the processor inputs the first data and the purchased clothing item into the first artificial intelligence model to perform continuous learning, and performs continuous learning in step S309. The second data and the purchased clothing item can be input into the second artificial intelligence model and continuously learned.

In addition, the first clothing item and the second clothing item may each be a plurality of clothing items, and the processor 110 provides the first clothing item and the second output to the user terminal based on the first output. When recommending a second clothing item based on a plurality of first clothing items and a plurality of second clothing items, a plurality of first clothing items and a plurality of second clothing items may be sorted and displayed in descending order of evaluation scores.

And the first data is created in the same interface every time through the user terminal, and is written through the day's good things and their degree, difficult things and their degree, meal record, exercise record, sleep record, one-line reviews of each record, and through the user terminal. May include color for each record specified.

The second data provides different activities for each day of the week in order to obtain various psychological characteristics of the user through the user terminal, and may be expressed in one or more of pictures, colors, and text.

The above activities may include one of the following: expressing one's feelings with colors, recording and matching colors by life cycle, writing a letter to oneself, writing a vision, checking habits, or freely expressing oneself through drawings.

The processor 110 inputs review data for each clothing item into a third artificial intelligence model that has been previously trained to analyze reviews of clothing items in step S401 to obtain a third output, which is review analysis data for each clothing item. You can.

The first artificial intelligence model includes first data acquired through a plurality of user terminals, a list of clothing items purchased through each user terminal on the date of acquiring the first data, ratings of all clothing items, and each clothing item. It is pre-trained with a dataset consisting of the third output of It can be pre-trained with a dataset consisting of the third output of .

The first artificial intelligence model is a multi-layer perceptron, RNN (recurrent neural network), CNN (convolutional neural network), and R-CNN (recurrent convolutional neural network). ) may be a multi-input model based on one or more of the algorithms.

A multilayer perceptron is a basic form of artificial neural network and can be composed of an input layer, a hidden layer, and an output layer. The input layer accepts data, the hidden layer processes the input data, and the output layer outputs the final result. Each node has weights and activation functions, allowing operations on input data to be performed and output values to be calculated.

RNN is a type of artificial neural network for processing time series data. It can process previous output values by considering them together with current input values, making it useful when processing sentences in the field of natural language processing. The most basic form of the RNN structure is a circular structure in which the input value at each point in time and the output value at the previous point in time are returned as the input value at the current point in time.

CNN is a type of artificial neural network used for image classification and processing. It operates by sliding input data through a filter of a certain size and extracting calculation results at each location. The calculation results of these filters are calculated again with other filters to obtain the final result. In this process, it is specialized for extracting image features, so it is widely used in the computer vision field.

R-CNN is a model that combines RNN and CNN, and is used in the fields of image classification and processing and natural language processing, and can also process time series data. R-CNN is a model that combines the characteristics of CNN with the advantages of RNN for processing time series data by enabling CNN's filter operations to be performed on time series data.

And the second artificial intelligence model is an image processing model based on one of the CNN, Faster R-CNN, and YOLO algorithms; and an ensemble model including a multi-input model based on one or more of a multilayer perceptron, recursive neural network, CNN, and R-CNN algorithm.

CNN, Faster R-CNN, and YOLO algorithms are algorithms for image processing. Among them, CNN is a representative deep learning algorithm in the image processing field. CNN is mainly used for feature extraction of images, and can automatically learn image features.

Faster R-CNN is an object detection algorithm that can find the location and class of objects in an image. Faster R-CNN provides fast object detection performance by combining CNN and RPN (Region Proposal Network).

YOLO, like Faster R-CNN, is an object detection algorithm that divides images into grids (cells) and detects and classifies objects at once, providing fast processing speed.

Multilayer perceptron, recursive neural network, CNN, and R-CNN algorithms can be used for multi-input models. A multilayer perceptron is a neural network consisting of an input layer, a hidden layer, and an output layer. It is the basic form of a general deep learning model and can be applied to nonlinear classification and regression problems.

RNN is a neural network for processing sequence data. It can identify patterns in sequence data by using the output of the previous state as the input of the current state, and is widely used in natural language processing and speech recognition, which require reading the flow of data.

Additionally, R-CNN is a model that combines CNN and RNN and can be used to process image sequence data.

An ensemble model is a model that increases prediction accuracy by using several different models together, including an image processing model based on one of the CNN, Faster R-CNN, and YOLO algorithms, multilayer perceptron, recursive neural network, CNN, and R -Multiple input models based on one or more of the CNN algorithms can be used together to produce a highly reliable final prediction result.

Lastly, the third artificial intelligence model may be a sentiment analysis model based on one of the SVM (Support Vector Machine), Naive Bayes, Neural Networks, and Decision Tree algorithms.

SVM is a supervised learning algorithm used in classification and regression analysis, and can operate by finding the boundary that classifies data and maximizing the distance between the closest data points to the boundary. Therefore, when new data comes in, it is possible to predict which class the data belongs to.

Naive Bayes is a probabilistic classification algorithm based on Bayes' theorem that can predict the probability that data belongs to a specific class, assuming that each feature is independent. Naive Bayes has the advantage of being fast in learning and working well even on small datasets.

Neural Networks is an algorithm that uses artificial neural networks to learn complex relationships between input data and output data. There are neural networks of various structures such as multi-layer perceptrons, recurrent neural networks, and CNNs. In emotional analysis, multi-layer perceptrons or CNNs are usually used.

Decision Tree is one of the classifiers that classifies data. It is expressed in a tree structure, where each node represents a condition for a specific characteristic, and each branch processes when the condition is True and False. represents. Decision trees can provide intuitive classification results.

Sentiment analysis model is a technology that analyzes the emotions of text in the field of natural language processing (NLP), and can be classified into positive emotions, negative emotions, and neutral emotions. This allows you to understand the emotional state of the review text.

As an example, the dataset for pre-training the first artificial intelligence model includes a list of clothing items viewed through each user terminal on the date the first data was acquired, and a list of clothing items on the date the first data was acquired. Each user terminal can further include a list of clothing items in the shopping cart.

The first artificial intelligence model includes first data acquired through a plurality of user terminals, a list of clothing items purchased through each user terminal on the date of acquiring the first data, ratings of all clothing items, and each clothing item. In addition to the third output of the learning data, a list of clothing items viewed through each user terminal on the date the first data was acquired, and a list of clothing items contained in the shopping cart through each user terminal on the date the first data was acquired. By using it, you can output more reliable results.

In addition, the data set for pre-training the second artificial intelligence model includes the top 5 most viewed clothing items among the list of clothing items viewed through the user terminal, and the shopping cart among the list of clothing items stored in the shopping cart through the user terminal. You can further include the top 5 clothing items with a long period of time.

The second artificial intelligence model is the second data obtained through a plurality of user terminals, the purchase history of each user account, the rating of all clothing items, and the third output of each clothing item, as well as the information retrieved through the user terminal. More reliable results by using the top 5 most viewed clothing items among the list of clothing items and the top 5 clothing items with the longest period of time in the shopping cart among the list of clothing items placed in the shopping cart through the user terminal as learning data. can be output.

Referring to FIG. 4, in one embodiment, the processor 110 recommends a first clothing item based on the first output and a second clothing item based on the second output to the user terminal in step S401. , input review data of the first clothing item and the second clothing item into the third artificial intelligence model, obtain review analysis data of the first clothing item and the second clothing item, and obtain the first clothing item and the second clothing item in step S402. If the current rating of the clothing item is 2.0 or less, or if the review analysis data has negative sentiment, the rating of the clothing item with negative sentiment can be set to 0, and the first clothing item can be re-predicted based on the first artificial intelligence model. In step S403, if the current rating of the second clothing item is 2.0 or less or the review analysis data has negative sentiment, the rating of the clothing item with negative sentiment is designated as 0, and based on the second artificial intelligence model The second clothing item can be re-predicted.

Since the ratings and reviews may differ from the results at the time of learning, before recommending the first clothing item based on the first output and the second clothing item based on the second output to the user terminal, the first clothing item and the second clothing item are recommended to the user terminal. 2 By analyzing the ratings and reviews of clothing items, clothing items with low ratings and many negative reviews can be excluded from the results and the first and second clothing items can be predicted again. If the rating falls below 2.0 differently from the time of learning, or if the review analysis data changes from positive or neutral sentiment to negative sentiment, the result may change if the rating is entered as 0.

In addition, when the processor 110 recommends a first clothing item to the user terminal based on the first output and a second clothing item based on the second output, the processor 110 recommends the first clothing item to the user terminal. Provides a plurality of models wearing photos and a plurality of models wearing photos of the second clothing item, wherein the plurality of models wearing photos of each of the first clothing item and the second clothing item are in the order in which the model's movements are followed. The display can be automatically switched.

A plurality of models wearing photos of each of the first clothing item and the second clothing item may be sequentially displayed in a slideshow format on a web page displayed on the user terminal. A photo of each model wearing the model can be displayed on the screen for a predetermined period of time and then automatically switch to the next photo.

In addition, when the processor 110 recommends a first clothing item to the user terminal based on the first output and a second clothing item based on the second output, the processor 110 recommends the first clothing item to the user terminal. An image of a model wearing a moving image and an image of a model wearing the second clothing item moving may be provided.

The video may be a video of a model wearing a clothing item in which a short video is played repeatedly.

The processor 110 may transmit a writing notification message to the user terminal when 30 hours have elapsed since the first and second data were last acquired through the user terminal.

The first and second data are recorded daily through the user terminal, allowing the user to reflect on themselves and calm their minds, providing psychological stability and comfort. A notification message can be sent when a predetermined time has elapsed so that the user can continuously look into his or her inner self and psychology.

In one embodiment, the processor 110 acquires first data and second data every day for 28 days through the user terminal, sorts all first data and second data in order of creation date, and produces a digital booklet. It is possible to provide a system for providing a discount coupon for a first clothing item and a second clothing item to the user terminal.

The second data may be 4 weeks for each day of the week, or a total of 28 days. For each day of the week, topics such as understanding the mind through colors, multitherapy, self-observation, writing a letter to myself, finding my vision, checking habits, and expressing through pictures are provided, and various activities are provided within the theme for each day of the week for a total of 4 weeks. can do.

Since the first and second data written consistently for 28 days can provide great psychological confidence or psychological strength to the user, the ability to produce a completed digital booklet into a real booklet may be provided. The user's first and second data produced as actual booklets can be delivered to the address entered by the user.

In addition, when the processor 110 acquires the first data and second data obtained every day through the user terminal between midnight and 9 a.m. of the next day on which the first data and second data must be acquired, It may be determined that the first data and the second data were acquired on the date they should be acquired. By flexibly judging the day, the user can be encouraged to write the first data and the second data every day.

In one embodiment, the processor 110, after obtaining first data through the user terminal, provides today's color emotion analysis results to the user terminal based on the color included in the first data, Color emotion analysis results can be obtained using the color emotion analysis API. At this time, the color emotion analysis API may be Brandfetch Color Psychology API or ColorTag API.

Brandfetch Color Psychology API and ColorTag API can extract color information from images or videos and analyze the extracted color information to identify emotions or psychological states. Brandfetch Color Psychology API extracts the main color information of an image and uses color psychology theory based on this to identify the emotions, perceptions, and behaviors evoked by the image. This API utilizes a color database to analyze color information extracted from images by associating them with the emotional and psychological meaning of each color.

Additionally, ColorTag API can classify the category to which the image belongs based on color information extracted from the image. This API can determine the category an image belongs to by combining color information extracted from the image and text analysis technology. For example, the category of a picture drawn by a user can be classified based on color information extracted from the picture.

In one embodiment, the processor 110 extracts keywords based on the color emotion analysis results, matches the extracted keywords with labels assigned to each clothing item sold on the website, and uses the user Clothing items with a high match can be displayed on the terminal. By labeling each clothing item with related categories and emotional keywords, clothing items can be recommended based on the color of the user's daily record. As a result, newly sold clothing items with no previous purchase history can be easily recommended.

In addition, in one embodiment, the process calculates the evaluation score of each of the plurality of first and second clothing items based on [Equation 1],

[Equation 1]

The score is the evaluation score of the clothing item, x_rt is the rating of the clothing item, x_vw is the number of views of the clothing item, x_sl is the sales volume of the clothing item, and x_rv is the number of reviews of the clothing item.

In [Equation 1], as the rating (x_rt) of the clothing item approaches 5, the evaluation score (score) of the clothing item may increase. Additionally, as the sales volume (x_sl) and number of reviews (x_rv) of a product increase, the evaluation score (score) can increase, and as the number of views (x_vw) increases, the impact on the evaluation score (score) can be reduced.

Therefore, [Equation 1] can be used to calculate an evaluation score by considering all of the product's rating, number of views, sales volume, and number of reviews.

Additionally, the number of reviews may exclude meaningless reviews and reviews with short text length. For example, you can exclude meaningless reviews such as ‘Like’ or ‘Thank you’, lists of repeated text such as “ㅇㅇㅇㅇㅇ” or “Like, Like, Like...”, and reviews with a text length of 10 or less.

As an example, referring to FIG. 5 , the processor may collect data related to clothing items using web crawling technology or an API provided by a clothing item sales site in step S501. Data related to the collected clothing items may include detailed size, material composition, thickness, category, etc.

In addition, in step S502, data such as detailed size, material composition, thickness, category, and elasticity of clothing items posted and sold on the website are compared with data related to the collected clothing items to extract similar or identical clothing items. In step S503, the ratings, reviews, and one-line comments of the clothing items extracted are analyzed using the third artificial intelligence model, and in step S504, analysis information on clothing items that are similar or identical to the clothing items being sold is posted on the website. , you can post it on your website and add it to the information about the clothing items you are selling. And when the corresponding clothing item is recommended to the user terminal, analysis data can be provided together, or it can be used to compile and analyze data on similar or identical clothing items.

In one embodiment, [Equation 2] can be used to compare data related to collected clothing items with data related to arbitrary clothing items being sold on a website.

[Equation 2]

The similarity(A,B) is the similarity between clothing items A and B, w_i is the weight, and d_i(A,B) may be the difference between the ith attribute value of clothing items A and B.

Under the assumption that clothing items A and B are clothing items of the same category, the ith attribute value of clothing items A and B may be detailed size, material composition, thickness, category, elasticity, etc. The value of the numerator is a value obtained by calculating the weighted sum of vectors A and B, then squaring the logarithmic function of the weight of each attribute and the difference value (d_i), and then summing it, which can indicate the degree of similarity between vectors A and B. In addition, the value of the denominator is calculated by taking a logarithmic function as the sum of the difference values of all the properties of vectors A and B by squaring them with the weight, squaring the value, and then calculating the square root to correct the size difference between vectors A and B. You can.

In addition, by performing an image search on a search portal site based on the image of the clothing item, data such as detailed size, material composition, thickness, category, and elasticity of clothing items with similar images can be collected.

Clothing items with similar or identical detailed information can be extracted from clothing items with similar images, and the average score of the extracted clothing items can be used when pre-training the first artificial intelligence model or the second artificial intelligence model, and the first artificial intelligence model When the average rating of a recommended clothing item based on the prediction results of the artificial intelligence and the second artificial intelligence is less than 2.0, the clothing item may be excluded and re-predicted.

By providing a product recommendation system based on user psychology through artificial intelligence according to the present invention, clothing items can be recommended based on the user's personal psychology that changes every day, and new clothing items with no previous purchase history can also be recommended. , It is easy to continuously obtain data about the user's psychological state and psychological characteristics, and each time the user purchases a clothing item, the performance can be updated by learning an artificial intelligence model, and the user can write about the psychological state and psychological characteristics. You can achieve psychological stability or psychotherapy effects by going out, you can provide users with photos or videos of dynamic clothing items worn, you can instill in users a sense of purpose for recording their own psychology, and you can receive multiple recommended clothing items. This has the effect of sorting and providing items in order of highest evaluation.

The embodiments described above may be implemented with hardware components, software components, and/or a combination of hardware components and software components. For example, the devices, methods, and components described in the embodiments may include, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, and a field programmable gate (FPGA). It may be implemented using one or more general-purpose or special-purpose computers, such as an array, programmable logic unit (PLU), microprocessor, or any other device capable of executing and responding to instructions. A processing device may execute an operating system (OS) and one or more software applications that run on the operating system. Additionally, a processing device may access, store, manipulate, process, and generate data in response to the execution of software. For ease of understanding, a single processing device may be described as being used; however, those skilled in the art will understand that a processing device includes multiple processing elements and/or multiple types of processing elements. It can be seen that it may include. For example, a processing device may include a plurality of processors or one processor and one controller. Additionally, other processing configurations, such as parallel processors, are possible.

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc., singly or in combination. Program instructions recorded on the medium may be specially designed and configured for the embodiment or may be known and available to those skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floptical disks. -Includes optical media (magneto-optical media) and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, etc. Examples of program instructions include machine language code, such as that produced by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

Software may include a computer program, code, instructions, or a combination of one or more of these, which may configure a processing unit to operate as desired, or may be processed independently or collectively. You can command the device. Software and/or data may be used on any type of machine, component, physical device, virtual equipment, computer storage medium or device to be interpreted by or to provide instructions or data to a processing device. , or may be permanently or temporarily embodied in a transmitted signal wave. Software may be distributed over networked computer systems and stored or executed in a distributed manner. Software and data may be stored on one or more computer-readable recording media.

Although the embodiments have been described with limited drawings as described above, those skilled in the art can apply various technical modifications and variations based on the above. For example, the described techniques are performed in a different order than the described method, and/or components of the described system, structure, device, circuit, etc. are combined or combined in a different form than the described method, or other components are used. Alternatively, appropriate results may be achieved even if substituted or substituted by an equivalent.

Therefore, other implementations, other embodiments, and equivalents of the claims also fall within the scope of the following claims.

100: server
110: Processor 120: Memory

Claims (15)

In a product recommendation system based on user psychology through artificial intelligence,
Includes a server that recommends products based on user psychology through artificial intelligence,
The server is,
processor; and
contains memory,
The processor,
We operate a website that recommends products based on user psychology through artificial intelligence,
When a user account is logged in to the website through the user terminal, an interface and a guide image for obtaining first data are provided to the user terminal, and the first data is the user's today's psychology created through the user terminal. This is data about the state,
Obtaining first data through the user terminal,
An interface and a guide image for obtaining second data are provided to the user terminal, wherein the second data is data about the psychological characteristics of the user created through the user terminal,
Obtaining second data through the user terminal,
Obtaining a first output by inputting the first data into a first artificial intelligence model that has been previously trained to predict a preferred clothing item based on data on today's psychological state,
Obtaining a second output by inputting the second data into a second artificial intelligence model pre-trained to predict preferred clothing items based on data on psychological characteristics,
recommending a first clothing item based on the first output and a second clothing item based on the second output to the user terminal,
When a clothing item is purchased through the user terminal, the first data and the purchased clothing item are input into the first artificial intelligence model for continuous learning, and the second data and the purchased clothing item are input to the first artificial intelligence model. Input to the second artificial intelligence model and continuously learn,
Collects data related to clothing items using web crawling technology or APIs provided by clothing item sales sites,
Compare the detailed size, material composition, thickness, category, and elasticity data of clothing items posted and sold on the website with data related to the collected clothing items to extract similar or identical clothing items,
The ratings, reviews, and one-line comments of the extracted clothing items are analyzed using a third artificial intelligence model,
Add analysis information on clothing items similar or identical to clothing items posted on the website and sold to the information on clothing items sold on the website;
[Equation 2] is used to compare data related to collected clothing items with data related to arbitrary clothing items sold on the website,
[Equation 2]

The system where similarity(A,B) is the similarity between clothing items A and B, w_i is the weight, and d_i(A,B) is the difference between the ith attribute values of clothing items A and B. delete According to claim 1,
The first clothing item and the second clothing item,
Each is a plurality of clothing items,
The processor,
When recommending a first clothing item based on the first output and a second clothing item based on the second output to the user terminal,
A system for sorting and displaying a plurality of first clothing items and second clothing items in order of good evaluation scores. According to claim 1,
The first data is,
Created in the same interface every time through the user terminal,
Includes today's good things and their extent, difficult things and their extents, meal records, exercise records, sleep records, one-line reviews of each record, and the color of each record specified through the user terminal,
The second data is,
In order to acquire various psychological characteristics of the user through the user terminal, different activities are provided for each day of the week, expressed in one or more of pictures, colors, and text,
The above activities are:
A system that includes one of the following: expressing one's feelings through colors, recording and matching colors by life cycle, writing a letter to oneself, writing a vision, checking habits, and freely expressing oneself through drawings. According to claim 1,
The processor,
Input review data for each clothing item into a third artificial intelligence model that has been previously trained to analyze reviews of clothing items, and obtain a third output, which is review analysis data for each clothing item,
The first artificial intelligence model is,
A dataset consisting of first data acquired through a plurality of user terminals, a list of clothing items purchased through each user terminal on the date of acquiring the first data, ratings of all clothing items, and a third output of each clothing item. It is already learned,
The second artificial intelligence model is,
A system that is pre-trained with a dataset consisting of secondary data acquired through a plurality of user terminals, purchase history of each user account, ratings of all clothing items, and third output of each clothing item. According to clause 5,
The first artificial intelligence model is,
Based on one or more of the following algorithms: multi-layer perceptron, RNN (recurrent neural network), CNN (convolutional neural network), and R-CNN (recurrent convolutional neural network) It is a multi-input model with
The second artificial intelligence model is,
Image processing model based on one of the following algorithms: CNN, Faster R-CNN, YOLO; and an ensemble model including a multi-input model based on one or more of the multilayer perceptron, recursive neural network, CNN, and R-CNN algorithms,
The third artificial intelligence model is,
The system is a Sentiment Analysis model based on one of the SVM (Support Vector Machine), Naive Bayes, Neural Networks, and Decision Tree algorithms. According to clause 5,
The dataset for pre-training the first artificial intelligence model is,
Further comprising a list of clothing items viewed through each user terminal on the date the first data was acquired, and a list of clothing items contained in a shopping cart through each user terminal on the date the first data was acquired,
The dataset for pre-learning the second artificial intelligence model is,
Among the list of clothing items viewed through the user terminal, the top 5 most viewed clothing items, and the top 5 clothing items with the longest period of time in the shopping cart among the list of clothing items placed in the shopping cart through the user terminal, system. According to clause 5,
The processor,
Before recommending a first clothing item based on the first output and a second clothing item based on the second output to the user terminal,
Input review data of the first and second clothing items into the third artificial intelligence model to obtain review analysis data of the first and second clothing items,
If the current rating of the first clothing item is 2.0 or less, or if the review analysis data has negative sentiment, the rating of the clothing item with negative sentiment is designated as 0, and the first clothing item is selected based on the first artificial intelligence model. Re-forecast,
If the current rating of the second clothing item is 2.0 or less, or if the review analysis data has negative sentiment, the rating of the clothing item with negative sentiment is designated as 0, and the second clothing item is created based on the second artificial intelligence model. Re-forecasting system. According to claim 1,
The processor,
When recommending a first clothing item based on the first output and a second clothing item based on the second output to the user terminal,
Provide to the user terminal, a plurality of photos of models wearing the first clothing item and a plurality of photos of models wearing the second clothing item, wherein a plurality of models wearing each of the first clothing item and the second clothing item A system that automatically switches and displays photos in the order in which the model moves. According to claim 1,
The processor,
When recommending a first clothing item based on the first output and a second clothing item based on the second output to the user terminal,
A system that provides, to the user terminal, an image of a model wearing the first clothing item moving and an image of a model wearing the second clothing item moving. According to claim 1,
The processor,
A system that transmits a writing notification message to the user terminal when 30 hours have elapsed since the time the first data and the second data were last acquired through the user terminal. According to claim 11,
The processor,
When first data and second data are acquired every day for 28 days through the user terminal, all first data and second data are sorted in order of creation date, produced into a digital booklet, and provided to the user terminal,
A system that provides discount coupons for a first clothing item and a second clothing item to the user terminal. According to claim 12,
The processor,
If the first data and second data acquired every day through the user terminal are acquired between midnight on the date on which the first data and second data must be acquired and 9 a.m. the next day, the first data and second data A system that determines that it has been acquired on the date it should have been acquired. According to claim 1,
The processor,
After obtaining first data through the user terminal,
Provide today's color emotion analysis results to the user terminal based on the colors included in the first data, wherein the color emotion analysis results are obtained using a color emotion analysis API,
The color emotion analysis API is,
Brandfetch Color Psychology API or ColorTag API,
The processor,
A system that matches keywords included in the color emotion analysis results with labels predetermined for each of all clothing items, and displays clothing items with a match higher than a preset standard on the user terminal. According to clause 3,
The processor,
The evaluation scores of each of the plurality of first and second clothing items are calculated based on [Equation 1],
[Equation 1]

The SCORE is the evaluation score of the clothing item, x_rt is the rating of the clothing item, x_vw is the number of views of the clothing item, x_sl is the sales volume of the clothing item, and x_rv is the number of reviews of the clothing item.