KR102675804B1 - Service providing apparatus and method for price analysis of game item - Google Patents

Abstract

The present invention relates to a service providing device and method for price analysis of game items. More specifically, the present invention relates to a service providing device and method for price analysis of game items. In more detail, the transaction data of items used in the game is obtained and the user wishes to purchase or sell through neural network analysis based on the transaction data. It relates to a service providing device and method for price analysis of game items that support optimal trading of items. The present invention not only allows the user to select an analysis model that can most appropriately analyze the value of the analysis subject item when requesting a price analysis for the analysis subject item, but also calculates a reasonable optimal price of the analysis subject item through the selected analysis model. This not only prevents indiscriminate spending on items by users using online games, but also helps users sell items at an appropriate price, greatly increasing user satisfaction and reducing the indiscriminate value of items. It has the effect of blocking the rise and supporting the smooth operation of the game by the game operator.

Description

{Service providing apparatus and method for price analysis of game item}

The present invention relates to a service providing device and method for price analysis of game items. More specifically, the present invention relates to a service providing device and method for price analysis of game items. In more detail, the transaction data of items used in the game is obtained and the user wishes to purchase or sell through neural network analysis based on the transaction data. It relates to a service providing device and method for price analysis of game items that support optimal trading of items.

Currently, various types of games are being provided along with the development of various game engines that provide precise physical effects and various graphics-related software and hardware to support them.

For example, in games of genres such as FPS (First-person shooter) or RPG (Role Playing Game), multiple users access a virtual game space provided online on a server and then manipulate the various characters provided by the game. It provides an active combat game where you achieve the goals set in the game.

In addition, games of this genre provide items that characters use in the game to compete with competitors in the game, and these items are set with various attributes to differentiate competitiveness between users within the game. It acts as a factor that increases interest.

Therefore, users strive to acquire items with higher ability attributes than items owned by other users within the game, and game service providers utilize the competitive elements using these items to provide an exchange for item trading within the game. I'm doing it.

These exchanges provide the ability to trade items between users through cyber money used within the game, and the more types of attributes set for an item or the higher the grade of the item determined in the game based on those attributes, the higher the level. Due to high demand, it is traded at a high price.

In addition, most of the exchanges are operated in an auction format, so if other users also want to purchase a specific item that a user wants to purchase, the user must bid a higher price than other users to win the bid and purchase the specific item.

At this time, in the process of purchasing an item, the user tries to prevent unnecessary expenditure of cyber money that must be paid when purchasing the item by determining whether the price offered by the seller is appropriate and the value level of the item. We want to acquire as much cyber money as possible by identifying the appropriate value.

However, since the exchange simply provides the past transaction history of other items similar to the item the user wants to buy or sell and does not provide any further information, it is difficult for the user to determine the appropriate value of the item, resulting in losses during the transaction process. In many cases, there is a problem of low satisfaction among users using the exchange.

Additionally, because the game continuously adds new items with various properties through updates, the past history provided by the exchange is useless in determining the appropriate value of the item.

Korean Patent No. 10-0972428

The present invention collects transaction data from an exchange operated on a game server, generates learning data based on such transaction data, creates a neural network model learned with the learning data, and allows users to purchase or sell through the neural network model. The purpose is to help users increase their satisfaction with item transactions by providing users with not only the appropriate price for the item they want, but also the timing of the transaction.

A service providing device for price analysis of game items according to an embodiment of the present invention includes item information including one or more attribute types and attribute values for each attribute type corresponding to an item to be traded according to an item transaction from a game server, and the transaction. A transaction data collection unit that collects transaction completion data containing transaction history information about the transaction history of the target item, and a unit that encodes categorical data included in the transaction completion data into numeric data to generate transaction information and then store it. A data pre-processing unit, a classification unit that classifies the plurality of transaction information into a plurality of groups based on one or more attribute types included in item information, and a classification unit for classifying the transaction information through the classification unit. A learning unit that trains the transaction information in an analysis model that is a neural network model corresponding to the group, and includes a plurality of analysis models that each correspond to the plurality of groups and in which a correlation between item information and transaction history information is learned. Transaction registration data containing item information about the analysis target item requested by the user is collected from the game server through the transaction data collection unit, and the transaction registration data is preprocessed through the data preprocessor to generate analysis target information. A specific group classified through the classification unit is identified for the analysis target information, and the analysis target information is applied to a specific analysis model corresponding to the specific group through the learning unit to determine transaction details predicted for the analysis target item. It may include an analysis unit that generates prediction information.

As an example related to the present invention, the attribute type is at least one of the name of the item, the type of the item, the grade of the item, the wearing part of the item, the type of character that can use the item, one or more option types set to the item, and the registration price. It may be characterized as including.

As an example related to the present invention, the transaction history information may be characterized as including the transaction price and transaction time.

As an example related to the present invention, each of the plurality of analysis models consists of a time series analysis model in which transaction patterns for time series changes in transaction history information for each item are learned, and the analysis unit is linked with the learning unit to model the specific analysis model. By applying the analysis target information to the specific analysis model, predicted pattern information for time series changes in transaction history information predicted for the analysis target item based on the current point in time is calculated, and the predicted pattern information is calculated based on the predicted pattern information. It may be characterized by calculating time series analysis information about the expected transaction time when the value of the item to be analyzed is maximized or minimized and the expected transaction price at the expected transaction time.

As an example related to the present invention, the transaction data collection unit collects event information about the event when an event occurs for a game to which the item is applied from the game server, and the learning unit links with the transaction data collection unit to collect the event information After the point of receiving the information, it is determined whether the event information is applied to the transaction information provided from the classification unit, and if the item according to the transaction information is an item subject to the event, the event information and the transaction information are matched. The analysis model corresponding to the transaction information is trained, and the analysis unit is linked with the transaction data collection unit and the learning unit to determine the specific event if a specific event subject to event application exists for the analysis target item according to the analysis target information. By applying specific event information to the specific analysis model together with the analysis target information, prediction pattern information considering price changes of the analysis target item according to the specific event is calculated through the specific analysis model, and the prediction pattern information is calculated. It may be characterized by calculating the time series analysis information based on .

As an example related to the present invention, the classification unit includes a preset classification model consisting of a neural network model, extracts data corresponding to item information from each of the plurality of transaction information, and then classifies the extracted data as learning data. The classification model may be trained to classify the plurality of transaction information into a plurality of different groups by training the model, and the transaction information may be classified into one of the plurality of groups through the learned classification model. there is.

As an example related to the present invention, the transaction information includes information on whether or not the item is on bid and, if the transaction information is transaction information on a bid on an item, bid history information on one or more bid prices for the bid on the bid. It includes, wherein the classification unit separately classifies the transaction information about the auctioned item among the plurality of transaction information into the auction transaction information, and the learning unit classifies the plurality of auction transaction information provided from the classification unit into a separate neural network model, the auction transaction information. The analysis model is trained for judgment, and when the analysis request information is received, the analysis unit interlocks with the transaction data collection unit to obtain an analysis target item currently being traded from the game server corresponding to the analysis request information based on the analysis request information. Collects transaction status data including progress information on the current transaction progress status, including item information about the item and a bid price for each user who participated in the purchase of the item to be analyzed, and performs the transaction through the data preprocessor. Status data may be pre-processed to generate transaction status information, and the transaction status information may be applied to the analysis model for determining a successful bid to calculate a probability of a failed bid for the analysis target item according to the analysis target information.

As an example related to the present invention, the analysis unit increases the attribute value by a preset value for an attribute type whose attribute value can be increased according to a preset standard among one or more attribute types according to item information included in the analysis target information. Generate virtual item information for a virtual item that updates the item information by increasing it, and apply the virtual item information to the specific analysis model into which the analysis target information is input to obtain transaction details predicted for the virtual item information. After calculating the predicted transaction information, recommendation information for a change in the properties of the analysis target item including the predicted transaction information and virtual item information may be generated and provided.

As an example related to the present invention, the transaction data collection unit includes seller information including the identifier and game level of the seller who sold the item corresponding to the transaction information, and the identifier and game level of the buyer who purchased the item. Transaction party information including buyer information is further collected from the game server, and the learning unit trains learning data matching the transaction information and transaction party information to an additional neural network model configured separately from the plurality of analysis models to The correlation between level changes and time intervals from the time of the user's item purchase to the time of item resale according to learning of transaction history information is learned in the additional neural network model, and the analysis unit collects the analysis target information through the transaction data collection unit. Yuri Choi collects user information about the level of the corresponding user, applies the user information and the analysis target information to the additional neural network model, and predicts the time when the analysis target item will be resold the most based on the user's level. It may be characterized by calculating and providing optimal purchase information for the time of purchase.

A service provision method for analyzing the price of a game item by a service provision device according to an embodiment of the present invention includes one or more attribute types corresponding to an item to be traded according to an item transaction from a game server and an attribute value for each attribute type. A step of collecting transaction completion data including information and transaction history information about the transaction history of the transaction target item, and encoding categorical data included in the transaction completion data into numeric data to generate transaction information and then storing it. A step of classifying the plurality of transaction information into a plurality of groups based on one or more attribute types included in item information for the plurality of transaction information, and a neural network corresponding to the group classified for the transaction information A step of learning the transaction information into an analysis model, which is a model, to create a plurality of analysis models that correspond to the plurality of groups and in which the correlation between item information and transaction history information is learned, and providing information on the analysis target item requested by the user. Transaction registration data containing item information is collected from the game server, the transaction registration data is pre-processed to generate analysis target information, and then the analysis target information is classified to identify a specific group, and a specific group corresponding to the specific group is classified. It may include the step of applying the analysis target information to an analysis model to generate prediction information about transaction details predicted for the analysis target item.

The present invention collects transaction-related information about a plurality of items provided in an online game, classifies similar items into a plurality of groups based on the type of attribute constituting the item, and builds a neural network model for each of the plurality of groups. By learning the characteristics based on the attribute types of the items in the group and the transaction history between users in a neural network model, the most important attributes among users are determined according to the types of important attributes that users value in the group to which the item belongs and changes in attribute values for each important attribute type. Analysis models with reasonable valuation criteria that can calculate the preferred optimal price can be created for each group, and based on this, when a user requests a price analysis for an item to be analyzed, the value of the item to be analyzed can be most appropriately analyzed. Not only can you select an analysis model that can be analyzed, but you can also calculate a reasonable optimal price for the item being analyzed through the selected analysis model and provide it to the user, thereby preventing indiscriminate spending on items by users using online games. Not only can it greatly increase user satisfaction by helping users sell items at an appropriate price, but it also has the effect of preventing indiscriminate increases in the value of items, thereby supporting the smooth operation of the game by game operators.

In addition, the present invention provides information on the expected transaction time when the value of the analysis target item is maximized or minimized and the expected sales price at the expected transaction time through time series analysis using an analysis model corresponding to the item to be analyzed, thereby providing information to the buyer. In the case of sellers, information can be provided on the transaction points in which the item subject to analysis can be sold at the maximum price, so that buyers who request item analysis can be provided. It has the effect of satisfying both customers and sellers.

1 is a configuration diagram of a service providing system including a service providing device for price analysis of game items according to an embodiment of the present invention.
Figure 2 is a configuration diagram of a service providing device for price analysis of game items according to an embodiment of the present invention.
Figure 3 is an operation example of a learning process of a service providing device for price analysis of game items according to an embodiment of the present invention.
Figure 4 is an operation example of an item value analysis process of a service providing device for price analysis of game items according to an embodiment of the present invention.
Figure 5 is an example of an operation for analyzing the probability of winning a bid by a service providing device for price analysis of a game item according to an embodiment of the present invention.
Figure 6 is an operation example of a virtual item creation and value evaluation process of a service providing device for price analysis of game items according to an embodiment of the present invention.
Figure 7 is a flowchart of a service providing method for price analysis of game items according to an embodiment of the present invention.

Hereinafter, detailed embodiments of the present invention will be described with reference to the drawings.

1 is a configuration diagram of a service provision system including a service provision device (hereinafter referred to as service provision device) for price analysis of game items according to an embodiment of the present invention.

As shown, the service provision system according to the present invention includes a game server 10 that provides an online game, a user terminal that connects to the game server 10 and participates in the online game, and a communication network. It may be configured to include a service providing device 100 that communicates with the game server 10 and the user terminal.

At this time, the service providing device 100 and the game server 10 may communicate with a plurality of different user terminals through a communication network.

In addition, the online game described in the present invention may include games of various genres such as FPS (First-person shooter), RPG (Role Playing Game), MMORPG (massive multiplayer online role playing game), etc. Hereinafter, online games are referred to as online games. We decide to call it a game.

Additionally, the communication network described in the present invention may include a wired/wireless communication network, and examples of such wireless communication networks include Wireless LAN (WLAN), DLNA (Digital Living Network Alliance), and Wireless Broadband (Wibro). , WiMAX (World Interoperability for Microwave Access: Wimax), GSM (Global System for Mobile communication), CDMA (Code Division Multi Access), CDMA2000 (Code Division Multi Access 2000), EV-DO (Enhanced Voice-Data Optimized or Enhanced Voice) -Data Only), WCDMA (Wideband CDMA), HSDPA (High Speed Downlink Packet Access), HSUPA (High Speed Uplink Packet Access), IEEE 802.16, Long Term Evolution (LTE), LTE-A (Long Term Evolution) -Advanced), Wireless Mobile Broadband Service (WMBS), 5G mobile communication service, Bluetooth, LoRa (Long Range), RFID (Radio Frequency Identification), Infrared Data Association (IrDA) , UWB (Ultra Wideband), ZigBee, Near Field Communication (NFC), Ultrasound Communication (USC), Visible Light Communication (VLC), Wi-Fi , Wi-Fi Direct, etc. may be included. In addition, wired communication networks include wired LAN (Local Area Network), wired WAN (Wide Area Network), Power Line Communication (PLC), USB communication, Ethernet, serial communication, and optical/coaxial. Cables, etc. may be included.

The game server 10 provides game-related data in which items are used to a user terminal connected to the game server 10 through a communication network, and the user terminal executes the game based on the game-related data to participate in the game. You can participate.

At this time, the item described in the present invention may mean item content used in the game.

In addition, the game server 10 may transmit exchange-related data (or exchange site) for trading items used in the game to the user terminal, and the user terminal participates in the exchange based on the exchange-related data. Afterwards, items registered by users of other user terminals can be purchased at the exchange using cyber money obtained by participating in the game or cyber money purchased through payment on the game server 10.

In addition, the game server 10 may generate and store transaction data for items traded or registered between a plurality of different user terminals through the exchange, and include the transaction data in exchange-related data to enable the user terminal It may also be provided to .

At this time, the transaction data may include transaction completion data for items for which a transaction has been completed, transaction status data for items in which a transaction is in progress, transaction registration data for items registered for transaction, etc.

In addition, the game server 10 described in the present invention may be configured as an exchange server that provides an exchange related to the game or may be configured to include an exchange server. If the game server 10 is configured as an exchange server, the exchange server can be linked with the game operation server that provides the game.

Through the above-described configuration, the game server 10 provides an exchange so that users participating in the game can purchase items with high ability values (stats) measured in the game through the exchange and then apply them to the game to compete with other users. We can help you gain the upper hand.

Additionally, the game server 10 may provide an exchange based on an auction (competitive sale) method or an exchange based on an agreement between transaction parties (seller and buyer).

According to the above-mentioned configuration, when a user wants to purchase an item through the exchange, the higher the ability value of the item, the higher the demand, so the item can be won only by presenting the highest asking price (cost to pay), and the user can bid on the item through the exchange. If you want to sell an item, you must offer an appropriate price in order for the buyer to respond to the transaction and sell the item.

However, previously, users had to independently judge the appropriate value of the item and present a bid or sell price, and the data that could be used as a basis for such judgment was provided by the exchange only in relation to items similar to the item desired to be purchased or sold. It is quite difficult for users to judge the appropriate value of an item because it is based solely on existing transaction history.

As a result, users refer to existing transaction history, estimate the value, and present the asking price. Therefore, when excessive competition occurs with other buyers for the item to be purchased, they pay a higher price than the actual value of the item, resulting in damage or lower than the appropriate value. Selling items at a low price can result in losses, which can greatly reduce satisfaction with using the exchange. This problem can also cause problems that lower satisfaction with the game, which can also cause damage to the game operator.

To this end, the service providing device 100 according to the present invention collects transaction data generated through the exchange provided by the game server 10, preprocesses and classifies the transaction data, and then performs deep learning. By learning the based neural network model and analyzing the appropriate value for the item requested by the user through the neural network model, the result can be generated and provided to the user, which will be described in detail with reference to the drawings below.

Figure 2 is a configuration diagram of the service providing device 100 according to an embodiment of the present invention, and Figures 3 and 4 are operational configuration diagrams of the service providing device 100 according to an embodiment of the present invention.

As shown, the service providing device 100 may be configured to include a communication unit 110, a storage unit 120, and a control unit 130.

The communication unit 110 may communicate with the game server 10 and one or more user terminals through a communication network.

In addition, the storage unit 120 can store various types of information, and the storage unit 120 can be configured in various forms such as HDD (Hard Disk Drive), SSD (Solid State Drive), etc., and can be configured as a DB. It may be possible.

In addition, the control unit 130 performs an overall control function of the service providing device 100, and the control unit 130 may include RAM, ROM, CPU, GPU, and a bus. GPUs, etc. can be connected to each other through a bus.

Additionally, the communication unit 110 and the storage unit 120 may be included in the control unit 130.

Additionally, the control unit 130 may perform an overall control function for analyzing appropriate values for items of the service providing device 100, which will be described in detail.

At this time, the control unit 130 and at least one of the various components constituting the control unit 130 may communicate with at least one of the game server 10 and the user terminal through the communication unit 110. Hereinafter, The communication configuration through the communication unit 110 will be omitted.

As shown, the control unit 130 includes a transaction data collection unit 131, a data pre-processing unit 132, a classification unit 133, a learning unit 134, and an analysis unit 135. You can.

3, when explaining the learning process of the control unit 130, the transaction data collection unit 131 communicates with the game server 10 to exchange information on the exchange (or exchange site) provided by the game server 10. You can access and collect transaction completion data according to item transactions from the exchange.

At this time, the transaction completion data may include item information including one or more attribute types and attribute values for each attribute type corresponding to the transaction target item according to the item transaction, and transaction history information regarding the transaction history of the transaction target item. there is.

The attribute type described in the present invention may be an attribute (or attribute name or attribute type) defined for an item (or an object that is an item or an object corresponding to an item) in the game algorithm of the game server 10, The attribute value may mean a value set for the attribute in the game algorithm or item contents.

At this time, the game algorithm-related data and item content-related data may be stored in the game server 10. Additionally, the item information may be composed of the item content.

In other words, the attribute type is used to define the object corresponding to the item, and in order to distinguish one attribute constituting the object from other attributes constituting the object, the type of attribute or name of the attribute defined in the game algorithm ( or variable name).

In addition, the attribute value may mean a value input to an attribute type (attribute) or a value mapped to an attribute type, and may be a value set in the game server 10 to distinguish different item objects. . At this time, the attribute value may be set in advance to the game algorithm or item content corresponding to the item (or object that is an item) and applied to the game algorithm.

Alternatively, the properties constituting the item (object corresponding to the item) may be comprised of the above property type and property value, and accordingly, the object corresponding to the item is comprised of one or more different properties, and the item (object corresponding to the item) may be comprised of the above property type and property value. The attribute type and attribute value for each of the one or more attributes constituting the corresponding object may be set in advance in the game server 10.

For example, one or more attribute types described in the present invention include the name of the item, the type of the item, the grade of the item measured in the game, the wearing part of the item, the type of character that can use the item, and the type of one or more options set to the item. , may include the registration price (the price registered by the seller of the item or the selling price of the item), etc.

In addition, as an example of an attribute value, attribute values such as guns, swords, shields, armor, etc. may be matched to the type of item that is the attribute type, and hand, head, and foot for the wearing part of the item that is another attribute type. Attribute values such as etc. may be matched.

In addition, the option type may refer to an attribute of an item that can change the data constituting the user's character or enemy character according to the use of the item by the game algorithm during the game progress, for example, reinforcement level, damage, etc. , may include various attributes applied in relation to the character in the game, such as accuracy, strength, movement speed, resistance, agility, stamina, etc.

Alternatively, the option type may refer to the properties of an item that interacts with a game character.

At this time, item use may refer to various character actions defined in the game in relation to the item, such as being worn on the character or attacking an enemy character.

For example, in the case of damage applied to an item, the physical strength of an enemy character can be reduced depending on the use of the item, and in the case of physical strength, the physical strength of the user's character can be increased depending on the use of the item.

Additionally, the transaction history information may include information about the transaction price of the item (final transaction price or successful bid price), the bid price when the item is rejected, the timing of the transaction of the item, etc.

At this time, the bid price may mean the highest price among the bid prices presented by the buyer.

In addition, the transaction data collection unit 131 collects the transaction completion data using an API (application programming interface) provided by the game server 10 or obtains meaningful information through web crawling. You can collect transaction completion data using a parsing module, or you can collect transaction completion data through image recognition or OCR (Optical Character Recognition), which recognizes exchange data in image form.

In addition, the preprocessor receives transaction completion data from the transaction data collection unit 131, encodes the categorical data included in the transaction completion data into numeric data to generate transaction information, and then generates transaction information in the storage unit ( 120).

In other words, the transaction completion data may include categorical data in text form, which may cause errors in the learning process of the neural network model, so the categorical data is converted to numeric data consisting of 0 and 1. Errors can be prevented.

At this time, the preprocessor may generate transaction information by preprocessing the transaction completion data through any one of onehot encoding, target encoding, and word2vec.

Meanwhile, the classification unit 133 divides the plurality of transaction information into a plurality of groups based on one or more attribute types included in the item information for the plurality of transaction information stored in the storage unit 120 by the preprocessor. It can be classified as:

For example, the classification unit 133 may classify a plurality of transaction information corresponding to a plurality of items into a plurality of different groups by classifying a plurality of different items having the same option type into the same group. there is.

Alternatively, the classification unit 133 may be configured to include a classification model that is an unsupervised learning-based neural network model, and after extracting data corresponding to item information from each of a plurality of transaction information stored in the storage unit 120, The classification model can be trained to automatically classify the plurality of transaction information into a plurality of different groups by training the extracted data as training data.

For example, the classification model learns a plurality of different images of monkeys, lions, and tigers based on feature points extracted from the images without labeling, and automatically generates an algorithm that can automatically classify monkeys, lions, and tigers into different types. When learning multiple images of monkeys, just as an algorithm is created to automatically classify monkeys into various types such as gibbons, lemurs, Japanese monkeys, etc. according to the type of monkey through learning the feature points of the images, the method used for the above-described learning Clustering a plurality of different items corresponding to a plurality of different transaction information into a plurality of different groups by learning the one or more attribute types constituting the item corresponding to the feature point of the image and the attribute value for each attribute type. ) can be classified as:

At this time, the classification model can independently generate one or more classification criteria for a plurality of different items through unsupervised learning based on transaction information, and can classify a plurality of items into weapons and armor according to the one or more classification criteria. It is classified into a plurality of different categories such as items, helmets, guns, knives, high-priced items, low-priced items, movement speed increase items, stamina increase items, etc., and the plurality of items are classified into a plurality of different categories and each corresponds to a plurality of different categories. It can be classified into multiple groups.

Additionally, the classification model may be composed of one or more neural network models.

At this time, the neural network model described in the present invention may be composed of an input layer, one or more hidden layers, and an output layer, and the neural network model includes a deep neural network (DNN), a deep neural network (RNN) Various types of neural networks such as Recurrent Neural Network), CNN (Convolutional Neural Network), etc. can be applied.

Additionally, the classification model may be composed of an Auto Machine Learning (AutoML)-based model that automatically generates a neural network model.

In addition, the classification unit 133 inputs (applies) the transaction information provided from the transaction data collection unit 131 to the classification model that has been learned when the classification model is completed, and through the classification model. A group identifier for the group corresponding to the input transaction information can be calculated, and through this, the transaction information can be classified into one of the plurality of groups.

Meanwhile, the learning unit 134 trains the transaction information into an analysis model, which is a neural network model corresponding to the group classified for the transaction information, through the classification unit 133 to each correspond to the plurality of groups and item Multiple analysis models can be created in which the correlation between information and transaction history information is learned.

At this time, the analysis model may be composed of one or more neural network models, and the learning unit 134 may include a plurality of analysis models.

As an example of this, the learning unit 134 may link with the classification unit 133 to include (generate) a plurality of analysis models each corresponding to a plurality of groups classified by the classification unit 133. , a group identifier corresponding to the transaction information and the transaction information classified by the classification unit 133 may be received for each plurality of different transaction information from the classification unit 133.

In addition, the learning unit 134 identifies an analysis model corresponding to the group to which the transaction information belongs based on the group identifier corresponding to the transaction information received from the classification unit 133, and identifies the analysis model corresponding to the group to which the transaction information belongs. Transaction information can be trained on the identified analysis model, and through this, each of a plurality of different analysis models can be trained with a plurality of transaction information corresponding to the analysis model.

Accordingly, the learning unit 134 can generate a plurality of different analysis models in which the correlation between item information and transaction history information is learned.

At this time, the plurality of analysis models may differ in at least one of the plurality of attribute types that are mutually learning targets according to the group classification of the classification unit 133.

As described above, the learning unit 134 is a plurality of groups in which correlations between transaction details and items belonging to a plurality of groups, each corresponding to the group classification of the transaction information of the classification unit 133, are learned. An analysis model can be created. For example, one of the plurality of analysis models may learn the correlation between weapon-related item information and transaction history information, and the other may learn the correlation between armor-related item information and transaction history information. A correlation may be learned, and another may be a correlation between item information related to a high-priced item and transaction history information.

Through this, the learning unit 134 can classify a plurality of items into similar items and train the classification model to use reasonable evaluation criteria for similar items, thereby providing optimal value evaluation for items with various attribute types. Multiple analysis models can be created to automatically select and analyze criteria.

On the other hand, when the above-described learning is completed, the analysis unit 135 controls each component of the control unit 130 to receive information about the analysis target item whose value is desired to be evaluated from the user's user terminal. The result of evaluating the appropriate value of the item can be provided to the user, which is explained with reference to FIG. 4.

First, the analysis unit 135 can control at least one of the transaction data collection unit 131, data pre-processing unit 132, classification unit 133, and learning unit 134.

In addition, the analysis unit 135 may receive analysis request information for analysis target items for which the transaction requested by the user has not been completed, and may receive analysis request information based on the analysis request information through the transaction data collection unit 131. By searching the exchange of the game server 10, transaction registration data, which is data registered for an analysis target item corresponding to the analysis request information, among data registered in the exchange, can be collected.

For example, when receiving analysis request information including item identification information or transaction identifier for an item to be analyzed from the user terminal, the analysis unit 135 provides the analysis request information to the exchange data collection unit and collects the exchange data. Under the control of the analysis unit 135, the unit searches the game server 10 or the exchange of the game server 10 based on the item identification information or transaction identifier included in the analysis request information to determine the game server 10. Alternatively, transaction registration data registered in the exchange for the analysis target item corresponding to the analysis request information may be collected from the exchange of the game server 10.

At this time, the transaction registration data may include item information about the item to be analyzed.

In addition, the analysis unit 135 controls the transaction data collection unit 131 to provide the transaction registration data to the preprocessor, and controls the data preprocessor 132 to provide the data preprocessor 132. Through this, categorical data included in the transaction registration data can be encoded into numeric data and preprocessed, and analysis target information can be generated through preprocessing of the transaction registration data.

At this time, the analysis unit 135 controls the data pre-processing unit 132 to extract item data corresponding to item information from the transaction registration data and then pre-processes the item data through the data pre-processing unit 132. Analysis target information can be generated.

In addition, the analysis unit 135 controls the data pre-processing unit 132 to provide the analysis target information from the data pre-processing unit 132 to the classification unit 133, and the classification unit 133 Control can be performed to identify a specific group classified for the analysis target information through the classification unit 133.

At this time, the analysis unit 135 applies the analysis target information as input to the classification model through the classification unit 133 and calculates a specific group corresponding to the analysis target information through the classification model. (Print) can be done.

In addition, the analysis unit 135 controls the analysis target information and the group identifier of the specific group to be provided from the classification unit 133 to the learning unit 134, and controls the learning unit 134 to provide the learning unit 134 with the analysis target information and the group identifier of the specific group. Among the plurality of analysis models included in the learning unit 134, the analysis target information is applied as input to a specific analysis model corresponding to a specific group identified based on the group identifier (by inputting) the analysis target information through the specific analysis model. It is possible to calculate (generate or output) predictive information about the transaction details predicted for the analysis target item corresponding to the analysis target information.

At this time, the prediction information may include information on the expected transaction price (expected sale price, expected purchase price, or expected successful bid price), expected transaction time, successful bid rate, etc., predicted for the analysis target item according to the analysis target information, The expected transaction price may be the optimal transaction price (appropriate price).

In addition, the analysis unit 135 works in conjunction with the learning unit 134 to generate analysis result information including the prediction information and item information for the analysis target item, and then sends the analysis request information to the user terminal. By transmitting, it can be provided so that the user of the user terminal can determine the appropriate value of the analysis target item based on the analysis result information.

At this time, the analysis unit 135 may extract item information about the analysis target item from the transaction registration data collected through the transaction data collection unit 131 and include it in the analysis result information. Alternatively, the analysis unit 135 may include the transaction registration data in the analysis result information.

As described above, the present invention collects transaction-related information about a plurality of items provided in an online game, then classifies similar items into a plurality of groups based on the type of attribute constituting the item, and uses a neural network for each of the plurality of groups. After building the model, the characteristics based on the attribute types of the items in the group and the transaction history between users are trained in a neural network model to determine the types of important attributes that users value in the group to which the item belongs and changes in attribute values for each important attribute type. Accordingly, an analysis model with reasonable valuation standards that can calculate the most preferred (reasonably judged) optimal price among users can be created for each group, and based on this, users can request price analysis for the item being analyzed. Not only can you select an analysis model that can most appropriately analyze the value of the item subject to analysis, but you can also calculate a reasonable optimal price for the item subject to analysis through the selected analysis model and provide it to users, allowing users who play online games Not only can it significantly increase user satisfaction by preventing indiscriminate spending on items and helping users sell items at an appropriate price, but it can also support smooth game operation for game operators by blocking indiscriminate increases in the value of items. there is.

Meanwhile, based on the above-described configuration, the present invention configures each of the plurality of analysis models configured in the learning unit 134 as a time series analysis model, and responds to time series changes in transaction history information for each item in the analysis model composed of the time series analysis models. By learning the transaction pattern, the analysis model can be used to calculate and provide time series analysis information on the expected transaction time when the value of the item to be analyzed is maximized or minimized and the expected transaction price at the expected transaction time. This is explained in detail.

First, the plurality of analysis models configured in the learning unit 134 may be composed of time series analysis models. As examples of such time series analysis models, various models such as auto regressive model, multivariate model, autoregressive multivariate model, etc. may be applied. You can.

In addition, the learning unit 134 can generate an analysis model in which transaction patterns for time series changes in transaction history information for each item are learned by training the transaction information in an analysis model composed of the time series analysis model.

Accordingly, the analysis unit 135, in conjunction with the learning unit 134, applies the analysis target information to the specific analysis model identified for the analysis target information to the analysis target item based on the current time. Calculate predicted pattern information about time series changes in predicted transaction history information, and based on the predicted pattern information, the expected transaction time at which the value of the item to be analyzed is maximized or minimized, and the expected transaction price at the expected transaction time (expected Time series analysis information on the winning bid price) can be calculated.

Additionally, the analysis unit 135 may generate analysis result information including the time series analysis information and transmit it to the user terminal.

In addition, in the above-described configuration, the analysis unit 135 calculates time series analysis information on the time series change in item value at each transaction time based on the prediction pattern information, and provides analysis result information including the time series analysis information. It can also be created and provided.

According to the above-described configuration, the present invention provides information about the expected transaction time at which the value of the analysis target item is maximized or minimized and the expected transaction price at the expected transaction time through time series analysis through an analysis model corresponding to the analysis target item. By providing buyers with information on the transaction time when they can purchase the item under analysis at the minimum price, and providing sellers with information about the transaction time when they can sell the item under analysis at the maximum price, it is possible to analyze the item. It can satisfy both buyers and sellers who request it.

In addition, based on the above-described configuration, the service providing device 100 according to the present invention considers the impact of such events on the price of the item when a game-related event occurs in the game server 10 and determines the optimal transaction time and price. Information can be provided, which is explained in detail.

First, the transaction data collection unit 131 may collect event information about the event when an event occurs for a game to which the item is applied from the game server 10.

At this time, examples of events may include various events such as game balance correction, addition or update of maps used in the game, adjustment of attribute values for each attribute type of the item that is the event target, etc.

Additionally, the event information may include information about one or more items to which the event is applied.

Accordingly, the learning unit 134, in conjunction with the transaction data collection unit 131, determines whether to apply (apply) the event information to the transaction information provided from the classification unit 133 after the point of receipt of the event information. target or not) is determined based on information about one or more items subject to event application included in the event information, and if an item according to the transaction information is an item subject to the event, the event information and the transaction information are matched. It can be trained on the analysis model.

Through this, the learning unit 134 can generate a plurality of different analysis models in which the correlation between item information and transaction history information is learned according to the event type of the event information.

In addition, the analysis unit 135 is linked with the transaction data collection unit 131 and the learning unit 134, and if there is a specific event subject to event application to the analysis target item according to the analysis target information, By applying (inputting) specific event information as input to the specific analysis model identified through the classification unit 133 for the analysis target information together with the analysis target information, the price of the analysis target item according to the specific event Predictive pattern information considering variation can be calculated (output) through the specific analysis model.

In addition, the analysis unit 135 generates a time series for the expected transaction time at which the value of the item to be analyzed is maximized or minimized and the expected transaction price at the expected transaction time reflecting the impact of the specific event based on the prediction pattern information. Analysis information can be calculated.

For example, the analysis unit 135 checks the event information collected in conjunction with the transaction data collection unit 131 at the time of receiving the analysis request information corresponding to the analysis target information and analyzes the analysis target information according to the analysis target information. It can be determined whether there is specific event information that sets the item to an event application target.

Alternatively, the transaction data collection unit 131 may store the event information in the storage unit 120 each time the event information is collected, and the analysis unit 135 may store the event information stored in the storage unit 120 upon receiving the analysis request information. Among them, it may be determined whether the analysis target item according to the analysis target information corresponding to the analysis request information has specific event information set as an event application target.

In addition, if there is specific event information associated with the analysis target information, the analysis unit 135 applies (inputs) the analysis target information and the specific event information as input to a specific analysis model corresponding to the analysis target information to determine the specific event information. Predictive pattern information about time series changes in transaction history information predicted for the analysis target item reflecting a specific event according to event information can be calculated through the specific analysis model, and as described above based on the predictive pattern information. Time series analysis information can be calculated (output).

Additionally, the analysis unit 135 may generate analysis result information including the time series analysis information and transmit it to the user terminal that transmitted the analysis request information.

In addition, the analysis unit 135, in conjunction with the transaction data collection unit 131, periodically collects transaction completion data for the analysis target item (or analysis target item associated with an event) and then collects transaction completion data as described above. Periodically calculates predictive pattern information, generates analysis result information including time series analysis information on time series changes in item value at each transaction time of the analysis target item each time the predicted pattern information is calculated, and transmits the analysis request information. It can be transmitted to the user terminal.

This configuration is not applied only to the analysis target item, and the analysis unit 135 analyzes the new item when collecting event information on new item registration from the game server 10 in conjunction with the transaction data collection unit 131. After self-determining the target item, time series analysis information can be calculated as described above, and analysis result information including time series analysis information for the new item can be provided to the user terminal that has requested value change analysis for the new item. there is.

Additionally, the analysis unit 135 may set each of the items supplied daily or weekly by the game server 10 as an analysis target item, calculate the analysis result information, and then provide it.

In this way, the present invention builds an analysis model by learning events that cause price changes of items in online games into a neural network model, and determines the predicted transaction price change for the item to be analyzed by reflecting the impact of the event through the analysis model. Information can be provided.

Meanwhile, based on the above-described configuration, the present invention can separately collect information on rejected items and provide information on the probability of successful bids for items to be analyzed by including them in the analysis result information, see Figure 5. This is explained in detail.

First, the transaction data collection unit 131 may collect transaction completion data for items that have been traded in an unbidden state from the game server 10.

Additionally, the data pre-processing unit 132 may pre-process transaction completion data for items that have been rejected to generate transaction information and then provide it to the classification unit 133 or store it in the storage unit 120.

At this time, the transaction information may further include information on whether the bid is successful or not, and may further include quote history information regarding the bid price for each user who participated in the auction of the item.

In the above-described configuration, if the game server 10 does not provide the transaction completion data including the successful bid information, the transaction data collection unit 131 provides a list of items registered for transaction from the game server 10. After collecting the first list and the second list of the list of items for which transactions have been completed, one or more bid information on the bid items can be obtained by comparing the first list and the second list and removing duplicate items. .

In addition, the transaction data collection unit 131 searches the exchange of the game server 10 based on the invalid bid information for each invalid bid information and collects transaction registration data for the invalid bid item corresponding to the invalid bid information. Set (convert) the transaction registration data corresponding to the rejected bid information to transaction completion data, and automatically (automatically) generate information on whether or not the bid has been set to be a rejected item, and the transaction completion data for the rejected item (in the rejected bid information) It can be included in the corresponding transaction completion data) and then provided to the data pre-processing unit 132.

In addition, the classification unit 133 identifies transaction information about a bid item among a plurality of transaction information stored in the storage unit 120 based on the bid status information, and divides the transaction information about the bid bid item into a bid transaction information. It can be classified separately as information.

In addition, the learning unit 134 trains the plurality of successful bid transaction information provided from the classification unit 133 to a separate neural network model configured separately from the plurality of analysis models, so that the plurality of successful bid transaction information is learned. It is possible to create an analysis model for judgment of validity, which is a neural network model of .

In addition, when receiving the analysis request information, the analysis unit 135 is linked with the transaction data collection unit 131 to obtain a current value corresponding to the analysis request information from the exchange of the game server 10 based on the analysis request information. Transaction status data that includes item information about the analysis target item in which a transaction is in progress and progress information about the current transaction progress status, including the bid price for each user (per buyer) who participated in the purchase of the analysis target item. It can be collected from the exchange of the game server 10.

Additionally, the analysis unit 135 may preprocess the transaction status data through the data preprocessing unit 132 to generate transaction status information.

In addition, the analysis unit 135 applies (inputs) the transaction status information as an input to the analysis model for determining a successful bid, and determines the item to be analyzed according to the transaction status information and the object to be analyzed through the analysis model for determining a failed bid. Based on one or more bid prices currently presented for the item, the predicted auction probability for the analysis target item according to the analysis target information may be calculated (output).

In addition, the analysis unit 135 may further include the calculated bid probability of the analysis target item in the analysis result information generated through the analysis model for the analysis target item for which a transaction is in progress, and the bid probability may be The included analysis result information can be transmitted to the user terminal that transmitted the analysis request information.

As described above, the service providing device 100 according to the present invention learns the transaction information about the auctioned items in a separate neural network model to determine the bid probability for the analysis target item requested by the user at the optimal level for the analysis target item. It can be provided along with the price to help users easily determine the appropriate price for the item being analyzed.

In addition, when the service providing device 100 according to an embodiment of the present invention receives an analysis request for the optimal selling price of the analysis target item from the user terminal of the seller who wants to sell the analysis target item, the service providing device 100 configures the analysis target item. After creating a virtual item by increasing the attribute value of the attribute type that can be increased (or strengthened) among the above attribute types, the predicted transaction price for the virtual item is provided, so that the seller can strengthen the item being analyzed and sell it at a higher price. We can provide support to consider selling, which is explained in detail with reference to FIG. 6.

As shown, the analysis unit 135 targets an attribute type whose attribute value can be increased according to a preset standard among one or more attribute types according to the item information included in the analysis target information, and selects a preset attribute type according to the standard. Virtual item information for a virtual item whose item information has been updated can be generated by increasing the attribute value by the numerical value.

At this time, the analysis unit 135 identifies one or more attribute types whose attribute value can be increased according to a preset standard among one or more attribute types corresponding to the analysis target item, and preset for each identified one or more attribute types. One or more virtual item information corresponding to the one or more attribute types may be generated by increasing the value by a preset value according to a standard.

For example, the analysis unit 135 increases the attribute value of the first attribute type that can increase the attribute value corresponding to the analysis target item by 1 based on the item information of the analysis target item to generate first virtual item information. Generates, based on the item information, increases the attribute value of the second attribute type that can increase the attribute value corresponding to the analysis target item by 1, and the attribute value of the first attribute type is the same as the item information. 2 Virtual item information can be created.

Alternatively, the analysis unit 135 may increase the attribute value of each of the plurality of attribute types that can be increased by a preset value according to the above standard to generate virtual item information in which the attribute values of each of the plurality of attribute types are all increased. .

In addition, the analysis unit 135, in conjunction with the learning unit 134, applies (inputs) the virtual item information as an input to a specific analysis model into which the analysis target information is input to produce a predicted result for the virtual item information. After calculating (outputting) the predicted transaction information about the transaction history through the specific analysis model, recommendation information on changes to the properties of the analysis target item including the predicted transaction information and virtual item information can be generated and provided.

At this time, the predicted transaction information may include information on the expected transaction price, expected transaction time, successful bid rate, etc. for the virtual item according to the virtual item information.

In addition, the analysis unit 135 may apply each of the one or more virtual item information to the specific analysis model to generate one or more predictive trading information corresponding to the one or more virtual item information through the specific analysis model. , prediction information for each of the one or more virtual item information and recommendation information including the one or more virtual item information may be generated.

In addition, the analysis unit 135 includes the recommendation information in the analysis result information calculated for the seller's analysis target item and transmits it to the seller's user terminal, or sends the recommendation information to the user terminal together with the analysis result information. can be transmitted to.

As described above, the present invention changes the attribute value for the type of attribute that can strengthen the ability of the item to be analyzed as the game progresses, so that the user can obtain one virtual item from the item to be analyzed as the game progresses in the future. By setting the above and suggesting the optimal price sold on the exchange for each virtual item through neural network analysis, we can greatly increase the satisfaction of sellers selling items by supporting sellers to strengthen the items being analyzed and sell them at a higher price. there is.

Meanwhile, the service providing device 100 according to the present invention learns the transaction patterns of users who purchase and then resell items for each item in a separate additional neural network model, so that the user who wants to purchase the item to be analyzed through the additional neural network model is the most It can provide information about when the item being analyzed can be purchased at a favorable price, which is explained in detail.

First, the transaction data collection unit 131 collects seller information including the identifier and game level of the seller who sold the item corresponding to the transaction information, and buyer information including the identifier and game level of the buyer who purchased the item. Transaction party information including may be further collected and stored in the storage unit 120.

In addition, the learning unit 134 generates a plurality of learning data matching transaction information and transaction information corresponding to the transaction information based on the information stored in the storage unit 120, and then generates the plurality of learning data. By training an additional neural network model configured separately from the plurality of analysis models, the correlation between the level change and time interval from the time of the user's item purchase to the time of item resale for each item can be learned in the additional neural network model.

At this time, the time interval may mean the time interval between the purchase point and the resale point.

In addition, the analysis unit 135 receives user information about the level of the user corresponding to the analysis target information from the game server 10 through the transaction data collection unit 131. Collects and applies the user information and the analysis target information to the additional neural network model through the learning unit 134 to predict the time when the analysis target item will be resold the most based on the user's level. Optimal purchase information for the point in time can be calculated.

Additionally, the analysis unit 135 may transmit the optimal purchase information to a user terminal that has requested analysis of the analysis target item, and may provide the optimal purchase information by including it in the analysis result information.

Through this, the present invention estimates and recommends the point in time when the item the user wishes to purchase is most resold or when the number of items with a high possibility of resale increases, so that the item the user wishes to purchase becomes cheap due to abundant supply of the item. We can help you purchase it at a low price.

Meanwhile, the analysis unit 135 may be configured to include an XAI (eXplainable AI) model, which is an explainable neural network model, and inputs the analysis model included in the learning unit 134 in conjunction with the learning unit 134. and output values are applied to the It can be calculated (output) through the XAI model.

Additionally, the analysis unit 135 may calculate the description information through the XAI model for each of the plurality of analysis models included in the learning unit 134.

At this time, the XAI model may include Partial Dependence Plot (PDP), Individual Conditional Expectation (ICE), Local Interpretable Model-agnostic Explanations (LIME), SHapley Additive exPlanation (SHAP), etc.

Through this, the service providing device 100 can provide explanatory information on why the item is overvalued or undervalued, so it can be used to discuss the game company's item development planning direction, the seller of the item, or You can identify the types of attributes buyers should keep in mind and provide them so they can focus.

Figure 7 is a flowchart of a service provision method for price analysis of game items by the service provision device 100 according to an embodiment of the present invention.

First, the service providing device 100 receives from the game server 10 item information including one or more attribute types and attribute values for each attribute type corresponding to an item to be traded according to an item transaction, and transaction details of the item to be traded. Transaction completion data containing transaction history information can be collected (S1).

Additionally, the service providing device 100 may generate transaction information by encoding categorical data included in the transaction completion data into numerical data and store the transaction information (S2).

Additionally, the service providing device 100 may classify the plurality of transaction information into a plurality of groups based on one or more attribute types included in the item information (S3).

In addition, the service providing device 100 learns the transaction information in an analysis model, which is a neural network model corresponding to the group classified for the transaction information, and corresponds to each of the plurality of groups and provides a link between item information and transaction history information. Multiple analysis models with learned correlations can be created (S4).

Thereafter, when learning of the plurality of analysis models is completed (S5), the service providing device 100 receives analysis request information for the analysis target item from the user terminal (S7), and an item for the analysis target item requested by the user is provided. Transaction registration data including information can be collected from the game server 10 (S8).

At this time, if learning of one of the plurality of analysis models has not been completed, the service providing device 100 may train it based on the transaction information (S6).

Additionally, the service providing device 100 may preprocess the transaction registration data to generate analysis target information and then classify the analysis target information to identify a specific group among the plurality of groups (S9).

In addition, the service providing device 100 may apply the analysis target information to a specific analysis model corresponding to the identified specific group to generate prediction information about transaction details predicted for the analysis target item ( S10).

As described above, the present invention analyzes items through a neural network model, automatically classifies items into a plurality of groups according to classification criteria, and supports evaluating the appropriate value of items belonging to the classified groups according to important attributes. , the most persuasive criteria for evaluating the value of an item can be automatically technically selected without the intervention of background knowledge of the game, and the price patterns that vary by group can be analyzed to support an accurate valuation of the item.

In addition, the present invention can induce users to improve commerce of game items and game immersion, and thereby reduce user abandonment by increasing understanding of games and items. Because of this, it can be of great help in item patching along with maintaining game service, and if game service maintenance continues, the in-game economic market can be activated and content can be diversified. Additionally, it can increase buyers' desire to pay and encourage buyers to pay within the game.

Components described in embodiments of the present invention include, for example, a storage unit 120 such as memory, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, and an FPGA. One or more devices, such as hardware such as a Field Programmable Gate Array (Field Programmable Gate Array), PLU (programmable logic unit), microprocessor, software containing an instruction set, or a combination thereof, or any other device capable of executing and responding to instructions. It can be implemented using a general-purpose computer or a special-purpose computer.

Anyone skilled in the art to which the present invention pertains can make modifications and changes to the above-described content without departing from the essential characteristics of the present invention. Accordingly, the embodiments disclosed in the present invention are not intended to limit the technical idea of the present invention, but are for illustrative purposes, and the scope of the technical idea of the present invention is not limited by these embodiments. The scope of protection of the present invention should be interpreted in accordance with the claims below, and all technical ideas within the equivalent scope should be construed as being included in the scope of rights of the present invention.

10: Game server 100: Service provision device
110: communication unit 120: storage unit
130: Control unit 131: Transaction data collection unit
132: Data preprocessing unit 133: Classification unit
134: Learning Department 135: Analysis Department

Claims (10)

Collecting transaction completion data including item information including one or more attribute types and attribute values for each attribute type corresponding to the transaction target item according to the item transaction from the game server, and transaction history information regarding the transaction history of the transaction target item. transaction data collection department;
a data preprocessor that encodes categorical data included in the transaction completion data into numerical data to generate transaction information and then stores it;
a classification unit that classifies the plurality of transaction information into a plurality of groups based on one or more attribute types included in item information;
The transaction information is trained in an analysis model, which is a neural network model corresponding to a group classified through the classification unit, and each of the plurality of groups corresponds to a plurality of learned correlations between item information and transaction history information. A learning unit containing an analysis model; and
Transaction registration data containing item information for the analysis target item requested by the user is collected from the game server through the transaction data collection unit, and the transaction registration data is preprocessed through the data preprocessor to generate analysis target information. A specific group classified through the classification unit is identified for the analysis target information, and the analysis target information is applied to a specific analysis model corresponding to the specific group through the learning unit to generate transaction details predicted for the analysis target item. Analysis department that generates predictive information about
Includes,
The transaction data collection unit includes transaction party information including seller information including the identifier and game level of the seller who sold the item corresponding to the transaction information, and buyer information including the identifier and game level of the buyer who purchased the item. collect more from the game server,
The learning unit trains the learning data that matches the transaction information and the transaction party information into an additional neural network model configured separately from the plurality of analysis models, from the time of the user's item purchase to the time of item resale according to learning of the transaction history information for each item. The correlation between the level change and the time interval is trained in the additional neural network model,
The analysis unit collects user information about the level of the user corresponding to the analysis target information through the transaction data collection unit, and applies the user information and the analysis target information to the additional neural network model to determine the level of the user. A service providing device for price analysis of game items, characterized in that it calculates and provides optimal purchase information for Choi Yuri's purchase time, which is predicted to be the time when the item to be analyzed will be resold the most.
In claim 1,
The attribute type is a game characterized in that it includes at least one of the name of the item, the type of the item, the grade of the item, the part where the item is worn, the type of character that can use the item, the type of one or more options set for the item, and the registration price. A service providing device for price analysis of items.
In claim 1,
A service providing device for price analysis of game items, wherein the transaction history information includes transaction price and transaction time.
In claim 3,
Each of the plurality of analysis models consists of a time series analysis model in which transaction patterns for time series changes in transaction history information for each item are learned,
The analysis unit applies the analysis target information to the specific analysis model in conjunction with the learning unit to generate predictive pattern information about time series changes in the transaction history information predicted for the analysis target item based on the current time point to the specific analysis model. A game item characterized by calculating time series analysis information about the expected transaction time and the expected transaction price at the expected transaction time when the value of the item to be analyzed is maximized or minimized based on the prediction pattern information. A service provider for price analysis.
In claim 4,
The transaction data collection unit collects event information about the event when an event occurs for a game to which the item is applied from the game server,
The learning unit determines whether to apply the event information to the transaction information provided from the classification unit after the point of receiving the event information in conjunction with the transaction data collection unit, and determines whether the item according to the transaction information is the item to which the event is applied. In the case where the event information and the transaction information are matched, the analysis model corresponding to the transaction information is trained,
The analysis unit is linked with the transaction data collection unit and the learning unit, and if there is a specific event subject to event application to the analysis target item according to the analysis target information, the analysis unit collects the specific event information for the specific event along with the analysis target information. By applying the analysis model to calculate the prediction pattern information considering the price change of the analysis target item according to the specific event through the specific analysis model, and calculating the time series analysis information based on the prediction pattern information. A service providing device for price analysis of game items.
In claim 1,
The classification unit includes a preset classification model consisting of a neural network model, extracts data corresponding to item information from each of the plurality of transaction information, and then trains the classification model with the extracted data as training data to determine the plurality of transactions. Providing a service for price analysis of game items, characterized in that the classification model is trained to classify information into a plurality of different groups, and the transaction information is classified into one of the plurality of groups through the learned classification model. Device.
In claim 1,
The transaction information includes information on whether or not the item is on bid, and if the transaction information is transaction information on a bid on an item, quote history information on one or more bid prices for the bid on the bid,
The classification unit separately classifies transaction information on items that are rejected among the plurality of transaction information as failed transaction information,
The learning unit trains a plurality of invalid transaction information provided from the classification unit into an analysis model for determining invalid bids, which is a separate neural network model,
When receiving analysis request information for an analysis request for the analysis target item, the analysis unit links with the transaction data collection unit to analyze the current transaction corresponding to the analysis request information from the game server based on the analysis request information. Collects transaction status data including item information about the target item and progress information about the current transaction progress including a bid price for each user who participated in the purchase of the analysis target item, and through the data preprocessor. A game characterized in that the transaction status data is pre-processed to generate transaction status information, and the transaction status information is applied to the analysis model for determining a non-bid decision to calculate a probability of a non-bid for the analysis target item according to the analysis target information. A service providing device for price analysis of items.
In claim 1,
The analysis unit updates the item information by increasing the attribute value by a preset value for an attribute type whose attribute value can be increased according to a preset standard among one or more attribute types according to the item information included in the analysis target information. Create virtual item information about a virtual item, apply the virtual item information to the specific analysis model into which the analysis target information is input, calculate predicted transaction information about transaction details predicted for the virtual item information, and then calculate the predicted transaction information. A service providing device for price analysis of game items, characterized in that it generates and provides recommendation information about changes in the properties of an analysis target item, including predicted transaction information and virtual item information.
delete In the service provision method for price analysis of game items of the service provision device,
Collecting transaction completion data including item information including one or more attribute types and attribute values for each attribute type corresponding to the transaction target item according to the item transaction from the game server, and transaction history information regarding the transaction history of the transaction target item. step;
Step of encoding categorical data included in the transaction completion data into numeric data to generate transaction information and then storing it:
Classifying the plurality of transaction information into a plurality of groups based on one or more attribute types included in the item information;
The transaction information is trained on an analysis model, which is a neural network model corresponding to the group classified for the transaction information, to generate a plurality of analysis models that each correspond to the plurality of groups and in which the correlation between item information and transaction history information is learned. steps; and
Transaction registration data containing item information for the analysis target item requested by the user is collected from the game server, the transaction registration data is pre-processed to generate analysis target information, and then the analysis target information is classified to identify a specific group. And applying the analysis target information to a specific analysis model corresponding to the specific group to generate predictive information about transaction details predicted for the analysis target item.
Includes,
The step of collecting the transaction completion data includes seller information including the identifier and game level of the seller who sold the item corresponding to the transaction information, and buyer information including the identifier and game level of the buyer who purchased the item. Further comprising collecting transaction party information from the game server,
The step of generating the plurality of analysis models involves training the learning data matching the transaction information and the transaction party information to an additional neural network model configured separately from the plurality of analysis models, so that the user can learn the transaction history information for each item. It further includes training the additional neural network model to determine the correlation between level changes and time intervals from the time of item purchase to the time of item resale,
The step of generating the prediction information includes collecting user information about the level of the user corresponding to the analysis target information, and applying the user information and the analysis target information to the additional neural network model based on the user's level. A method of providing a service for price analysis of game items, further comprising calculating and providing optimal purchase information regarding Choi Yuri's purchase time, which is predicted to be the time when the item to be analyzed will be resold the most.

Images