Target prediction method and system

Abstract

According to an embodiment of the present invention, a target prediction method for predicting a future outlook of a target, the method being performed by a computing device, comprises: when a user requests target prediction, collecting relevant structured and unstructured data; analyzing a relationship between a target and a variable affecting the target at a semantic level; and then obtaining a future outlook of the target.

Description

Target prediction method and system

[0001] The present invention is a method and system for predicting a target based on relationship information between target influence variables and targets at a semantic level.

[0002] With the recent emergence of pre-trained language models (Large Language Pretraining models, LLMs) on large-scale general domain data, various tasks previously handled manually are being replaced by artificial intelligence-based ones.

[0003] In particular, the technology of performing future prediction tasks using large-scale language models is an interesting and rapidly developing field of technology called futurecasting.

[0004] Futurecasting refers to the use of sophisticated algorithms to predict future trends, events, or behaviors, and can be applied to a wide range of fields, from predicting weather patterns to forecasting market trends and even anticipating social or political changes.

[0005] Machine learning, a subset of artificial intelligence, can play a pivotal role in this process. Some key aspects and technologies that machine learning models can replace in futurecasting are as follows:

[0006] Using past data, it is possible to predict future events. Machine learning models can identify patterns in large data sets and use them to make predictions. For example, in finance, these models can predict stock market trends, and in healthcare, they can predict disease outbreaks.

[0007] Time series analysis can be used in forecasting in fields such as meteorology, economics, and resource management. Machine learning models can analyze data points collected at successive time intervals to predict future points in the series.

[0008] In futurecasting, natural language processing (NLP) can be used to analyze news, social media, and other text data to gauge public sentiment or predict political or social trends.

[0009] However, the accuracy of futurecasting models can be limited by the quality and quantity of data, and they are particularly challenging when predicting numerical values ​​using both structured and unstructured data simultaneously.

[0010] The present invention proposes a target prediction method and system that accurately predicts the short-term or / and mid- to long-term prospects of a target based on data on various variables that may affect the target using a language model.

[0011] In detail, the target prediction method and system of the present invention can detect target influence variables related to a target at a semantic level, and accurately extract features that influence the target based on the detected target influence variables at the semantic level to predict the prospect of the target.

[0012] In addition, the present invention can provide a method and system for accurately predicting a target based on structured data and unstructured data for a target and target influence variables.

[0013] And the target prediction method and system of the present invention aims to predict the prospects of a target in the mid- to long-term based on various text data such as news and reports that can serve as a basis for predicting targets and target influence variables.

[0014] In addition, the present invention proposes a target prediction method and system that clearly presents a basis for predicting a target outlook based on target influence variables and their corresponding characteristics.

[0015] In detail, the target prediction method and system of the present invention can analyze target influence variables that affect the target at a semantic level and provide a basis for the predicted target based on a causal relationship with the characteristics of the target influence variables.

[0016] A target prediction method for predicting a future outlook of a target performed by a processor of a computing device according to an embodiment of the present invention, comprising the steps of: receiving a target prediction request from a user; determining a target prediction element to be predicted from the received target prediction request; retrieving a target outlook report for a target of the determined target prediction element, and generating relationship information between target-target influence variables at a semantic level based on the retrieved target outlook report; filtering structured data of features related to the target influence variables and unstructured data of text documents related to the target influence variables based on the target influence variables of the generated relationship information; calculating a future target outlook based on the filtered structured data and unstructured data; generating a basis for the calculated target outlook as relationship information at the feature level; and providing the calculated target outlook and the relationship information at the feature level to the user.

[0017] At this time, the step of receiving a target prediction request from the user may include a step of providing a chat interface to the user to receive a text including the target prediction request, and a step of analyzing the received text based on context to detect a context indicating the target prediction request.

[0018] In addition, the step of determining the target prediction element may further include a step of performing Named Entity Recognition on a text including the target prediction request to determine a keyword representing the target, a total outlook period, and a prediction unit period as the target prediction element.

[0019] At this time, the step of determining the target prediction element may further include a step of listing the plurality of recognized target keywords and providing them for the user to select, if a plurality of target keywords of a higher concept and a plurality of target keywords of a lower concept are recognized for the target of the target prediction element.

[0020] In addition, the step of generating the relationship information may include a step of defining target influence variables that influence the target at a semantic level, and a step of generating a causal relationship graph with the name of each defined target influence variable as a node name as relationship information.

[0021] At this time, the step of generating the relationship information may further include a step of indicating the causal relationship between target influence variables indicated by each node of the causal relationship graph using arrows.

[0022] In addition, the step of filtering the unstructured data of the above-mentioned features and the text documents related to the target influence variables may include the step of classifying the features stored in the data store into target influence variables defined at the semantic level, and the step of combining the structured data of the features classified into the target influence variables to create a structured data set.

[0023] In addition, the step of filtering the unstructured data of the above-mentioned feature and the text document related to the target influence variable may include a step of inputting the text document into a document classification prompt template and determining through a language model whether the text document is a document that influences the target.

[0024] At this time, the step of calculating the future target outlook may include a step of detecting a target outlook report that predicts the outlook of the target among the text documents, a step of performing sentiment analysis on sentences predicting the target in the target outlook report using a language model to classify the outlook of the target as positive, neutral, or negative for each target outlook report, a step of quantifying the level of the tone of the classified sentiment and returning it as prediction scoring data, and a step of listing the prediction scoring data of the target outlook reports in chronological order to generate quantitative data.

[0025] In addition, the step of calculating the future target outlook may include a step of combining the structured data and the quantitative data to create an integrated structured data set, and a step of inputting the created integrated structured data set into a prediction model to output a target outlook value.

[0026] Additionally, the step of calculating the future target outlook may further include a step of adjusting the target outlook based on relationship information between the target-target influence variables.

[0027] Additionally, the step of generating relationship information at the feature level may include a step of generating relationship information of features of target influence variables that serve as a basis for predicting the target outlook.

[0028] Additionally, the step of generating relationship information at the feature level may include a step of generating the relationship information including the numerical values ​​of features that influenced the predicted target outlook at one point in time.

[0029] In addition, when receiving a predicted environment change input from the user, the method may further include a step of performing a simulation according to the received predicted environment change input.

[0030] At this time, the step of performing a simulation according to the input of the predicted environment change received may include a step of changing the structured data of the feature according to the changed target influence variable when there is a change in the feature of the target influence variable, and a step of re-executing the process of interpreting the target forecast value and basis based on the changed structured data and unstructured data to output the target forecast value and basis information according to the assumption simulation.

[0031] In addition, the step of performing a simulation according to the received predicted environment change input may further include a step of detecting a case similar to the occurrence of the specific event when the occurrence of the specific event is received as a predicted environment change input from the user, and calculating a target forecast value based on the detected case.

[0032] A server computing system that receives a target prediction request from a user computing device and performs a target prediction task according to an embodiment of the present invention, the system comprising: a data store that stores target prediction-related data; a memory that stores commands and data for performing the target prediction task; and at least one processor that performs the target prediction task according to the commands and data of the memory, wherein the at least one processor receives a target prediction request from a user, determines a target prediction element to be predicted in the received target prediction request, retrieves a target outlook report for a target of the determined target prediction element, generates relationship information between target-target influence variables at a semantic level based on the retrieved target outlook report, filters structured data of features related to the target influence variables and unstructured data of text documents related to the target influence variables based on the target influence variables of the generated relationship information, calculates a future target outlook based on the filtered structured data and unstructured data, generates a basis for the calculated target outlook as relationship information at the feature level, and provides the user with the calculated target outlook and the relationship information at the feature level.

[0033] The target prediction method and system according to the present invention can accurately predict target prospects by predicting target prospects based on prediction basic data obtained by precisely performing data preparation necessary for target prediction.

[0034] Specifically, the target prediction method and system according to the present invention can precisely filter structured data and unstructured data related to target influence variables after defining target influence variables that influence targets at a semantic level.

[0035] In addition, the target prediction method and system according to the present invention can accurately predict target forecast values ​​through a time series prediction model by integrating filtered structured data and unstructured data.

[0036] Specifically, the target prediction method and system according to the present invention can generate a structured data set required for time series target prediction by quantifying unstructured data, accurately filtering only features related to target influence variables from structured data, and then generating an integrated structured data set.

[0037] In addition, the target prediction method and system according to the present invention can increase the reliability of the target forecast by providing basic data that serves as the basis for the forecast by interpreting the basis for the predicted target forecast.

[0038] In addition, the target prediction method and system according to the present invention has the advantage of being able to respond to various requests from users by performing simulations according to changes in the user's target prediction environment.

[0039] FIG. 1 illustrates an example of a block diagram of a computing system that performs a target prediction method according to an embodiment of the present invention.

[0040] FIG. 2 illustrates an example of a block diagram of a computing device, which is one of the components of a computing system that performs a target prediction method according to an embodiment of the present invention.

[0041] FIG. 3 illustrates an example block diagram of another aspect of a computing device, which is one of the components of a computing system that performs a target prediction method according to an embodiment of the present invention.

[0042] Figure 4 is a flowchart of a method for predicting the prospect of a target using a machine learning model according to an embodiment of the present invention.

[0043] FIG. 5 illustrates a meta-architecture for performing a method for predicting a target's prospect according to an embodiment of the present invention.

[0044] FIG. 6 is an example of a process for executing a target prediction task and determining causal relationship information with a target prediction variable according to an embodiment of the present invention.

[0045] FIG. 7 is an example of a process for performing data preparation based on relationship information between a target and a target influence variable according to an embodiment of the present invention.

[0046] Figure 8 is an example of a process for converting non-standard data into quantitative data according to an embodiment of the present invention.

[0047] Figure 9 is an example of a process for calculating a target outlook by integrating structured data and quantitative data according to an embodiment of the present invention.

[0048] Figure 10 is an example of a process for deriving a basis for a target outlook according to an embodiment of the present invention and a process for predicting an additional target outlook according to a user's simulation.

[0049] FIG. 11 is an example of a chart for a predicted target outlook according to an embodiment of the present invention.

[0050] Fig. 12 is an example of a causal relationship graph presented as basis data for a predicted target outlook according to an embodiment of the present invention.

[0051] Figure 13 is another example of a causal relationship graph presented as basis data for a predicted target outlook according to an embodiment of the present invention.

[0052] The present invention can be modified in various ways and has various embodiments. Therefore, specific embodiments are illustrated in the drawings and described in detail in the detailed description. The effects and features of the present invention, and the methods for achieving them, will become clear with reference to the embodiments described in detail below together with the drawings. However, the present invention is not limited to the embodiments disclosed below and can be implemented in various forms. In the following embodiments, the terms first, second, etc. are not used in a limiting sense, but are used for the purpose of distinguishing one component from another component. In addition, the singular expression includes the plural expression unless the context clearly indicates otherwise. In addition, the terms such as include or have mean that a feature or component described in the specification is present, and do not exclude in advance the possibility that one or more other features or components may be added.

[0053]

[0054] FIG. 1 illustrates an example of a block diagram of a computing system that performs a target prediction method according to an embodiment of the present invention.

[0055] Referring to FIG. 1, a computing system (1000) that performs target prediction according to one embodiment of the present invention includes a user computing device (110), a training computing system (150), and a server computing system (130), and each device and system are communicatively connected via a network (170).

[0056] According to an embodiment of the present invention, 1) a user computing device (110) can perform a target prediction method using a local or / and external machine learning model (120) or a machine learning model (140) provided by a server.

[0057] In addition, according to another embodiment of the present invention, 2) a server computing system (130) communicating with a user computing device (110) can provide a target prediction service to the user computing device (110) through an application or / and on the web in response to a user's request through the user computing device (110).

[0058] In addition, according to another embodiment of the present invention, 3) a user computing device (110) and a server computing system (130) may perform at least a part of a method of performing target prediction in conjunction with each other to provide a target prediction service to a user.

[0059] In addition, according to various embodiments of the present invention, the user computing device (110) and / or the server computing system (130) can learn the machine learning model (120 / 140) performed for target prediction through interaction with the training computing system (150) communicatively connected via a network (170). In this case, the training computing system (150) may be separate from the server computing system (130) or may be a part of the server computing system (130).

[0060] In some embodiments, the training computing system (150) may be part of a server computing system (130) or part of a user computing device (110).

[0061] In the following description, the description is based on the process of executing a target prediction task by accessing a server computing system (130) via a user computing device (110), collecting and analyzing data necessary for target prediction using a language model directly from the server computing system (130) or from a separate server, and performing target outlook prediction based on the collected and analyzed data. However, it can be understood that a part of the process described as being performed in the server computing system (130) is naturally included in the description of the present invention when it is performed in the user computing device (110).

[0062] The user computing device (110) may include any type of computing device, such as a smart phone, a mobile phone, a digital broadcasting device, a personal digital assistant (PDA), a portable multimedia player (PMP), a desktop, a wearable device, an embedded computing device, and / or a tablet PC.

[0063] Such a user computing device (110) includes at least one processor (111) and memory (112). Here, the processor (111) may be composed of at least one or a plurality of electrically connected processors among a central processing unit (CPU), a graphics processing unit (GPU), application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, micro-controllers, microprocessors, and / or other electrical units for performing functions.

[0064] The memory (112) may include one or more non-transitory / transitory computer-readable storage media, such as RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, and combinations thereof, and may include web storage of a server that performs a memory storage function on the Internet. The memory (112) may store data and instructions necessary for the at least one processor (111) to perform the operation of an application for performing target prediction.

[0065] In one embodiment, the user computing device (110) may store at least one machine learning model (120). For example, the user computing device (110) may be configured with a combination of various machine learning models, such as a plurality of neural networks (e.g., a deep neural network) that perform predictions on a target based on structured / quantitative data, or other types of machine learning models including nonlinear models and / or linear models.

[0066] For example, the predictive model may store a linear regression, decision tree, random forest, gradient boosting, pre-trained language model, or / and a deep learning model. The neural network may include at least one of feed-forward neural networks, recurrent neural networks (e.g., long short-term memory recurrent neural networks), convolutional neural networks, or / and other types of neural networks.

[0067] Additionally, the user computing device (110) may store a model to be used in each process and a prompt template that serves as the basis for input to the model in order to perform at least part of the process performed for target prediction through a large-scale language model (LLM).

[0068] For example, the user computing device (110) may store 1) a prompt for generating a query from a user's input, 2) a prompt for determining a causal relationship between a target and a target influence variable, 3) a prompt for identifying raw data related to the determined causal relationship, 4) a prompt template for quantifying unstructured data, etc.

[0069] That is, in one embodiment, the user computing device (110) may perform target prediction based on received data by requesting the language model of an external server to perform some of the execution steps in the target prediction task through prompts or the like.

[0070] In another embodiment, the target prediction task requested through the user computing device (110) may be performed in such a way that the server computing system (130) performs target prediction through at least one machine learning model (140) and a machine learning model of another server and provides predicted data to the user computing device (110).

[0071] Such a user computing device (110) may include at least one input component (121) that detects a user's input. For example, the user input component (121) may include a touch sensor (e.g., a touch screen or a touch pad, etc.) that detects a touch of a user's input medium (e.g., a finger or a stylus), an image sensor that detects a user's motion input, a microphone, a button, a mouse, and / or a keyboard that detects a user's voice input, etc. In addition, the user input component (121) may include an interface and an external controller when receiving an input to an external controller (e.g., a mouse, a keyboard, etc.) through an interface.

[0072] The server computing system (130) includes at least one processor (131) and memory (132). Here, the processor (131) may be composed of at least one or a plurality of electrically connected processors among a central processing unit (CPU), a graphics processing unit (GPU), application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, micro-controllers, microprocessors, and / or other electrical units for performing functions.

[0073] And the memory (132) may include one or more non-transitory / transitory computer-readable storage media such as RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. This memory (132) may store data and instructions necessary for the processor (131) to perform a task through the language model of the server computing system (130) or / and the language model of an external server, a prompt template, a machine learning model (140) for future casting, etc.

[0074] For example, the server computing system (130) may include a neural network or / and other multi-layer nonlinear models as a machine learning model (140) for futurecasting. Exemplary neural networks may include feedforward neural networks, deep neural networks, recurrent neural networks, and convolutional neural networks.

[0075] In one embodiment, the server computing system (130) may be implemented to include at least one computing device. For example, the server computing system (130) may be implemented to operate multiple computing devices according to a sequential computing architecture, a parallel computing architecture, or a combination thereof. Additionally, the server computing system (130) may include multiple computing devices connected via a network.

[0076] In an embodiment, the server computing system (130) may further include a data store computing system (1000) (hereinafter, "data store"), which is a storage for continuously storing and managing raw data that serves as the basis for futurecasting for a target. This data store may include various forms of data storage, ranging from file systems to cloud storage.

[0077] For example, a data store may include at least one of the following: a relational database that uses a structured query language (SQL) to define and manipulate data; a NoSQL database that is designed for flexibility and scalability and handles unstructured and semi-structured data; a data warehouse that centralizes large amounts of data from multiple sources and is optimized for querying and analysis, a data warehouse that stores large amounts of raw data in its native formats of structured, semi-structured, or unstructured data; and a local storage device or Network Attached Storage (NAS) that stores data in files in a format typically accessible by a computer operating system.

[0078] The training computing system (150) includes at least one processor (151) and memory (152). Here, the processor (151) may be composed of at least one or a plurality of processors electrically connected among central processing units (CPUs), graphics processing units (GPUs), application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, and / or other electrical units for performing functions. In addition, the memory (152) may include one or more non-transitory / transitory computer-readable storage media such as RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. The memory (152) may store data and instructions necessary for the processor (151) to train a futurecasting model.

[0079] For example, the training computing system (150) may include a model trainer (160) that trains a machine learning model stored in the user computing device (110) or / and the server computing system (130) using various training or learning techniques, such as backpropagation of errors.

[0080] For example, the model trainer (160) can update one or more parameters of a machine learning model for future casting using a backpropagation method based on a defined loss function.

[0081] In some implementations, performing backward propagation of errors may include performing truncated backpropagation through time. The model trainer (160) may perform a number of generalization techniques (e.g., weight reduction, dropout, knowledge distillation, etc.) to improve the generalization ability of the trained fusion casting model.

[0082] The model trainer (160) comprises computer logic utilized to provide the desired functionality. The model trainer (160) may be implemented as hardware, firmware, and / or software that controls a general-purpose processor. For example, in one embodiment, the model trainer (160) comprises a program file stored on a storage device, which may be loaded into memory and executed by one or more processors. In another implementation, the model trainer (160) comprises one or more sets of computer-executable instructions stored on a tangible computer-readable storage medium, such as a RAM hard disk or an optical or magnetic medium.

[0083] The network (170) includes, but is not limited to, a 3rd Generation Partnership Project (3GPP) network, a Long Term Evolution (LTE) network, a World Interoperability for Microwave Access (WIMAX) network, the Internet, a Local Area Network (LAN), a Wireless Local Area Network (Wireless LAN), a Wide Area Network (WAN), a Personal Area Network (PAN), a Bluetooth network, a satellite broadcasting network, an analog broadcasting network, and / or a Digital Multimedia Broadcasting (DMB) network.

[0084] In general, communication over a network (170) may be performed using any type of wired and / or wireless connection, using various communication protocols (e.g., TCP / IP, HTTP, SMTP, FTP), encodings or formats (e.g., HTML, XML), and / or protection schemes (e.g., VPN, Secure HTTP, SSL).

[0085]

[0086] FIG. 2 illustrates an example of a block diagram of a computing device, which is one of the components of a computing system (1000) that performs a target prediction method according to an embodiment of the present invention.

[0087] As illustrated in FIG. 2, the computing device (100) included in the user computing device (110), the server computing system (130), and the training computing system (150) includes a plurality of applications (e.g., Application 1 to Application N). Each application may include a machine learning library. For example, the applications may include a futurecasting application, a text messaging application, an email application, a dictation application, a virtual keyboard application, a browser application, a chat-bot application, a separate futurecasting application, and the like.

[0088] In an embodiment, the computing device (100) may include a model trainer (160) for training a futurecasting model, and may store and operate the futurecasting model to perform a target prediction task on input data.

[0089] Each application of the computing device (100) may communicate with a number of other components of the computing device, such as, for example, one or more sensors, a context manager, a device state component, and / or additional components. In one embodiment, each application may communicate with each device component using an API (e.g., a public API). In one embodiment, the API used by each application may be specific to that application.

[0090]

[0091] FIG. 3 illustrates an example block diagram of another aspect of a computing device, which is one of the components of a computing system (1000) that performs a target prediction method according to an embodiment of the present invention.

[0092] Referring to FIG. 3, a computing device (200) includes a plurality of applications (e.g., Application 1 to Application N). Each application may communicate with a central intelligence layer. For example, the applications may include an image processing application, a text messaging application, an email application, a dictation application, a virtual keyboard application, a browser application, etc. In one embodiment, each application may communicate with the central intelligence layer (and the models stored therein) using an API (e.g., a common API across all applications).

[0093] The central intelligence layer may include prompts utilizing multiple machine learning models and / or language models. For example, as illustrated in FIG. 3, each machine learning model and at least some of the models may be provided to each application and managed by the central intelligence layer. In other implementations, two or more applications may share a single machine learning model. For example, in some implementations, the central intelligence layer may provide a single model to all applications. In some implementations, the central intelligence layer may be included within the operating system of the computing device (200) or implemented differently.

[0094] The central intelligence layer may communicate with a central device data layer. The central device data layer may be a centralized data repository for the computing device (200). As illustrated in FIG. 3, the central device data layer may communicate with a number of other components of the computing device, such as, for example, one or more sensors, a context manager, a device state component, and / or additional components. In some implementations, the central device data layer may communicate with each device component using an API (e.g., a private API).

[0095] The techniques described herein may refer to servers, databases, software applications, and other computer-based systems, as well as actions taken and information transmitted to or from such systems. It will be appreciated that the inherent flexibility of computer-based systems allows for a wide range of possible configurations, combinations, and division of labor and functionality between and among components. For example, the processes described herein may be implemented using a single device or component, or multiple devices or components operating in combination. Databases and applications may be implemented on a single system or in a distributed system across multiple systems. Distributed components may operate sequentially or in parallel.

[0096]

[0097] Hereinafter, a target prediction method and system for collecting raw data using a language model, analyzing the collected raw data to predict the target's prospects, and providing causal relationship information that serves as a basis for predicting the prospects, using such a computing system (1000), will be described with reference to FIGS. 4 to 13.

[0098] First, a target prediction request can be received from a user computing device (110) of a computing system (1000), and a target prediction task can be executed according to the received target prediction request. (S101)

[0099] In an embodiment, the user computing device (110) may receive a text-based target prediction request from a user through a chat interface, transmit the text including the target prediction request to a server computing system (130), and execute a target prediction task of the server computing system (130).

[0100] The server computing system (130) can execute a target prediction task by detecting a pre-stored phrase for a target prediction request from text input through a chat interface or analyzing the text based on context to detect the context of the target prediction request.

[0101] And the server computing system (130) can recognize text including a target prediction request and determine target prediction elements for target prediction.

[0102] Here, the target prediction element may further include a target to be predicted and at least one of a total prediction length to be predicted and a prediction unit time.

[0103] In the embodiment, the target includes information about a value that changes over time, and predicting the target includes predicting and producing a value of the target in the future by predicting unit periods, up to the total outlook period.

[0104] Specifically, the server computing system (130) may input the text of the target prediction request into a query generation prompt template that analyzes the text of the target prediction request to determine target prediction elements, input the text of the target prediction request into a language model, and return at least one or more of the target prediction elements as output to the language model to determine the target prediction elements.

[0105] For example, a query generation prompt template can be configured to input “text of target prediction request” into an interactive prediction request box as input, and to recognize values ​​corresponding to the target, total outlook period, and unit period based on Named Entity Recognition (NER) as an action, and return the target, total outlook period, and unit period of the query as output.

[0106] As a more specific example, when a user inputs a target prediction request text such as “Predict what the lithium price will be in the future on a monthly basis for the next 12 months,” the server computing system (130) may determine the target prediction element by inputting <<Input: Interactive prediction request “Predict what the lithium price will be in the future on a monthly basis for the next 12 months,” Action: Recognize the values ​​corresponding to the target, total outlook period, and unit period for the input text through NER, and generate and return the following query, Output: Query - {Target:, Unit period:, Total outlook period:}>> as a prompt to the language model to output the target prediction element as {Target: Lithium market price, Unit period: Monthly, Total outlook period: 12 months}.

[0107] At this time, the server computing system (130) may provide a separate future casting interface for inputting target prediction elements for target prediction when the target prediction elements are not specified or are abstract, and transmit the target prediction elements input through the provided future casting interface to the server computing system (130) to execute the target prediction task. That is, when the target is classified from a superordinate concept to a number of subordinate concepts according to a category, the server computing system (130) may list target keywords mapped to the superordinate concepts and provide them for the user to select.

[0108] For example, the future casting interface provides keywords of targets derived through entity name recognition sequentially from the upper concept to the lower concept for the user to select, thereby enabling the user to more accurately determine the target he or she wishes to predict.

[0109] Once the target prediction elements are determined, the server computing system (130) can determine relationship information between the target and the target influence variables. (S103)

[0110] First, the server computing system (130) can collect target analysis data for the target. This can be done by filtering data in a data store within the server computing system (130) or by crawling data on the Internet.

[0111] For example, the server computing system (130) can detect target analysis data by performing a keyword search based on keywords representing the determined target. Here, the target analysis data may be target analysis reports related to the target.

[0112] Specifically, the server computing system (130) can request a target analysis report collection prompt template preset in the language model to search for analysis data related to the target based on the target's keywords and return the search data through an analysis report.

[0113] More specifically, the server computing system (130) can obtain a target analysis report as output by using a target analysis report collection prompt template to <<Input: Target - Lithium market price, Action: Search for and return an analysis report with a title related to the target through a keyword search>>.

[0114] And the server computing system (130) can detect target influence variables that affect the target from the collected target analysis data, and analyze and generate relationship information between the target influence variables and the target.

[0115] In an embodiment, the relationship information may include information about target influence variables that influence the future prediction of the target, and information about the relationship between the target influence variables and the target.

[0116] More specifically, information on target influence variables refers to information defining target influence variables at a semantic level, and information on relationships between targets / target influence variables may refer to the causal relationship between targets and target influence variables and the influence proportions and weights between target influence variables.

[0117] In the following, the relationship information between the target and the target influence variable will be referred to as causal relationship information.

[0118] In an embodiment, the server computing system (130) can generate causal relationship information by analyzing a semantic causal graph as target-target influence variable correlation information at a semantic level based on collected target analysis data.

[0119] To this end, in one embodiment, the server computing system (130) may perform a topic-relevant terms recognition module on target analysis data to detect and annotate target influence variables associated with the target in the target analysis data.

[0120] And the server computing system (130) can generate a causal graph at a semantic level by inputting the target analysis data with the target and target influence variables annotated into a causal graph generation model that is trained to generate a causal graph between target and target influence variables.

[0121] Here, the causal relationship graph between target-target influence variables can include information defining the target and target influence variables at the semantic level in nodes along with the node name.

[0122] For example, information for determining target and target influence variables at the semantic level may include additional annotations such as the name, keywords, source, domain, region, location, and characteristics of the element.

[0123] And the causal relationship graph between target-target influence variables can include information about the causal relationship between each node (target and target influence variable) through arrows, such as whether they influence each other in a preceding or subsequent manner.

[0124] In one embodiment, the server computing system (130) may perform a process of collecting target analysis data based on context and outputting causal relationship information of target-target influence variables based on the collected target analysis data through a Retrieval Augmented Generation (RAG) model.

[0125] Through the process of generating causal information according to such embodiments, target influence variables can be clearly identified and defined at the semantic level by concepts, categories, topics, and / or specific criteria, thereby accurately determining the context and domain related to the target influence variables at the semantic level.

[0126] And, by annotating the information defined in this way to the target influence variables and utilizing it to perform data preparation at the semantic level later, the raw data required for target prediction can be accurately determined.

[0127] Once the causal relationship information of the target-target influence variables is determined, the server computing system (130) can perform data preparation based on the determined causal relationship information. (S105)

[0128] First, the server computing system (130) can collect raw data related to the target of causal relationship information and target influence variables for predicting the target's outlook.

[0129] In an embodiment, the server computing system (130) may collect unstructured data (e.g., news in text form, analysis reports, etc.) and structured data related to targets and target influence variables through keyword searches indicating targets and target influence variables, and store the collected raw data in a data store.

[0130] The server computing system (130) can determine whether raw data stored in the data store is related to target influence variables at the semantic level (Document Identification) and extract relevant data. At this time, the raw data can be filtered based on whether it matches the semantic definitions included in the aforementioned target-target influence variables, thereby obtaining the prediction base data necessary for target prediction.

[0131] For example, the server computing system (130) can input a document to be judged as input and output the relevance to the target influence variable at the semantic level as an action, thereby extracting prediction basic data that is related to the target and the target influence variable at the semantic level from among the raw data.

[0132] To identify data related to target impact variables that influence these targets, past data analysis knowledge and domain expertise in the target-related field are important.

[0133] To supplement this, the server computing system (130) can derive events related to the target and unrelated events through a language model.

[0134] For example, the server computing system (130) can instruct the language model to return a plurality of associated events that affect a change in the target at a semantic level and a plurality of unassociated events that have no effect or affect the change below a threshold by sending a prompt to create associated / unassociated events that include phrases that instruct the language model to act as a domain expert for the target.

[0135] Specifically, by including information defining each target influence variable at the semantic level in the associated / unassociated event creation prompt, the language model can be instructed to distinguish associated and unassociated events that influence the target in the prediction base data.

[0136] And the server computing system (130) can create a document identification prompt for classifying and identifying prediction base data from raw data through the returned associated / unassociated events, and request a language model to classify documents for the raw data based on the created document identification prompt, thereby accurately extracting prediction base data related to the target and target influence variables.

[0137] Additionally, unstructured data related to the target's outlook can be detected from the predictive base data related to the target and / or target influence variables. That is, the server computing system (130) can classify documents related to the target and / or target influence variables from the raw data stored in the data store, and detect related events and / or sentences that affect the target from the documents.

[0138] For example, a document classification prompt may be configured to 1) instruct the target to act as an expert, 2) input at least one document included in the raw data to be identified as input data, 3) provide an action instruction to select one of a related event option that is relevant to the target's prediction in the document and a non-related event option that has no effect on the target, and 5) add a related event in the document that has an effect on the target to the related event option or add a non-related event that has no effect to the non-related event option.

[0139] As a specific example, the server computing system (130) can identify whether a document in the raw data is related to “lithium production” through a prompt consisting of <<1) Become a lithium expert. 2) Input: [document] 3) Classify [documents] related to the increase or decrease in lithium production. There are two options for your response. - Option 1: Highly relevant (list of related events), - Option 2: Not relevant (list of non-related events), 4) First, describe the reason for how the information provided related to lithium production increases or decreases. Then, place the option number on the last line. >>.

[0140] That is, the server computing system (130) collects raw data related to the target or / and the target influence variable, classifies the prediction basis data related to the target or / and the target influence variable from the raw data, determines the associated events and sentences that affect the target's outlook from the classified prediction basis data, and filters the sentences and associated events related to the target outlook from the raw data as unstructured data.

[0141] Next, the server computing system (130) can use a language model to identify and classify whether each feature stored in the data store belongs to a relevant target semantic variable, and generate a structured dataset consisting of structured data for the relevant features. Here, the feature refers to an attribute of data stored in a structured data format as various factors that can influence the target's outlook, and may include, for example, CSV, Excel files, or / and database tables.

[0142] For example, if the target is lithium price, target impact variables refer to variables that have a causal relationship with lithium price, such as “spodumene, lithium mine, lithium brine, lithium carbonate, lithium hydroxide, lithium battery”, and features can be structured data that belong to target impact variables and affect the outlook of the target, such as “Australian spodumene production, Australian spodumene exports, Chilean lithium hydroxide production, Chilean lithium hydroxide exports, Chinese spodumene imports, Chinese lithium carbonate imports, Chinese lithium carbonate production, Chinese lithium carbonate sales, lithium battery efficiency (km / wh), Chinese electric vehicle sales, and Chinese electric vehicle subsidy plan”.

[0143] That is, in an embodiment, a target influence variable may be a specific concept, topic, or category that influences the target outlook, and a feature may mean an attribute of structured data in a data repository related to the target influence variable.

[0144] And the server computing system (130) can filter out relevant features related to target influence variables among the features of the data store and integrate the filtered features to create a structured data set.

[0145] Specifically, the process of creating a structured dataset can be described by first listing features available in the data store by feature name. Furthermore, a description of each feature can be listed.

[0146] And the server computing system (130) can filter out features related to target influence variables that can affect the target among the listed features based on their association with target influence variables defined at the semantic level.

[0147] To this end, the server computing system (130) can utilize a machine learning model or language model that classifies the correlation between features and target influence variables.

[0148] In an embodiment, the server computing system (130) can map features classified into each target influence variable by listing feature names and descriptions of data stores, inputting keywords of target influence variables of causal relationship information into a word embedding model, and detecting feature names associated with the keywords of each target influence variable based on feature relevance. Here, word embedding refers to a model trained to classify features relevant to semantic target influence variables based on feature names and descriptions.

[0149] And the server computing system (130) can obtain tabular data corresponding to the names of classified features from a data store, organize and preprocess the obtained tabular data, and arrange it in a structured format so that it can be input into target prediction modeling, thereby generating a time-series structured data format (e.g., csv, excel, etc.).

[0150] In this way, the server computing system (130) can collect accurate raw data that serves as the basis for target prediction based on causal relationship information of target-target influence variables, and can precisely filter structured data and unstructured data required for target prediction from the collected raw data and use them as input data for target prediction modeling.

[0151] Next, the server computing system (130) can generate quantitative data by quantifying unstructured data (Text Processing for Forecasting). (S107)

[0152] First, the server computing system (130) can generate prediction scoring data by scoring the target prediction value for each target prospect report among the target prospect reports that predict the target prospect among the documents classified as unstructured data.

[0153] In detail, in the embodiment, the server computing system (130) inputs each target outlook report into a language model, performs sentiment analysis on associated sentences classified as predictions of the target outlook, classifies the target outlook into positive, neutral, and negative, and operates according to a target outlook scoring prompt that returns a numerical value for the level of tone, thereby listing the prediction scoring data in chronological order to generate quantitative data.

[0154] Specifically, the returning target outlook scoring prompt can be configured to, when a target outlook report (or, related sentences related to the target outlook extracted from the target outlook report) is input, classify opinions about the target outlook in the input text as positive / neutral / negative, and select a tone for the outlook opinion in the input text within a predetermined level range.

[0155] Additionally, the server computing system (130) can generate an event list based on related events that affect the outlook of targets detected in documents when filtering unstructured data.

[0156] For example, the server computing system (130) can generate a list of events that quantitatively quantify the date of occurrence of an event that affects the outlook of a target, related characteristics, the value of the related characteristics, and the impact and influence that affected the outlook of the target.

[0157] Additionally, the server computing system (130) can encode each document classified as unstructured data into a latent vector using a language model encoder and return an embedding matrix. Specifically, the server computing system (130) can obtain an embedding matrix that captures the semantic essence of each document by encoding the document into a latent vector using a language model.

[0158] Specifically, the server computing system (130) can input documents, such as news, among unstructured data into the encoder of a language model, and generate document embedding metrics to model prevalent topics within each document. The document embeddings generated in this manner can identify prevalent topics within the documents using an algorithm such as Latent Dirichlet Allocation (LDA), thereby highlighting topics (variables, features) that may influence the target's future prospects.

[0159] Afterwards, the server computing system (130) can predict the target outlook based on the generated structured dataset and quantitative data. (S109)

[0160] In detail, the server computing system (130) can calculate the target forecast value for each forecast unit period during the total forecast period based on the quantitative data and structured data sets.

[0161] To this end, the server computing system (130) can create an integrated structured dataset by concatenating a structured dataset created based on structured data and a quantitative dataset created based on unstructured data.

[0162] Specifically, the server computing system (130) can first classify data according to the influence it has on the target, and then combine the data by assigning weights to the data. For example, the server computing system (130) can classify variables among the features included in the structured dataset that have an influence on the target exceeding a reference value as macro variables, and classify variables that have an influence below the reference value as micro variables. In addition, the server computing system (130) can time-series match the classified macro variables with quantitative data and then integrate them into a single macro time-series structured dataset, and can integrate data classified into micro variables into a single micro time-series structured dataset.

[0163] That is, in the embodiment, an integrated structured dataset including both structured data information and unstructured data information can be created by matching and combining the event list and predicted scoring data according to the time-series flow of the structured dataset.

[0164] The server computing system (130) can input the generated integrated structured dataset into a prediction model to produce a target forecast value for each prediction unit period over the entire forecast period. Here, the prediction model may include linear regression, decision tree, random forest, gradient boosting, a deep learning model, or / and a pre-trained language model.

[0165] In an embodiment, the server computing system (130) may additionally input causal information at a semantic level into the prediction model to induce prediction of a target outlook based on the causal information.

[0166] Additionally, in the embodiment, the server computing system (130) may input the aforementioned embedding metrics into a second prediction model that predicts a target outlook based on the embedding metrics, thereby reflecting unstructured target prediction information that is not in structured data into the prediction value.

[0167] Specifically, in the embodiment, the server computing system (130) can input an integrated structured dataset into a first prediction model to initially produce a first target forecast value.

[0168] And the server computing system (130) can regulate the first target outlook based on the semantic causal relationship graph to produce a second target outlook reflecting the causal relationship information between the target influence variable and the target.

[0169] Finally, the server computing system (130) can calibrate the generated second target forecast value based on the non-standard target forecast information to ultimately generate the final target forecast value.

[0170] Additionally, the server computing system (130) can generate basis information by interpreting the basis for the target outlook based on causal relationship information and a structured dataset. (S111)

[0171] In detail, referring to FIG. 10, the server computing system (130) can interpret the basis for the final target forecast based on the causal relationship information at the semantic level and the structured dataset, and output basis information.

[0172] Specifically, the server computing system (130) can generate a past causal relationship graph at the feature level based on the present, the existing target values ​​of the past, the structured dataset, and the semantic causal relationship graph in the structured dataset.

[0173] And the server computing system (130) can generate a future causal relationship graph at the feature level based on a causal relationship discovery model (Data-driven Causal Discovery) learned from the past causal relationship graph, a structured data set, and a semantic causal relationship graph based on the present, a future final target forecast, and a semantic causal relationship graph.

[0174] And the server computing system (130) can provide a future causal relationship graph by mapping it to a target outlook, thereby providing basis information on how the target outlook was calculated due to which characteristics have an influence on the target outlook to some extent.

[0175] For example, referring to FIG. 11, the server computing system (130) can provide a target outlook graph representing target outlook values ​​calculated for each prediction unit period during the total outlook period through the user computing device (110).

[0176] In addition, as shown in FIG. 12, the server computing system (130) can provide a causal relationship graph at the feature level that interprets the basis for predicting the target outlook as basis information to the user computing device (110).

[0177] In particular, referring to FIG. 13, the server computing system (130) can further improve user reliability of the target outlook by displaying specific numerical values ​​of features that have influenced the predicted target outlook at a specific prediction point in time.

[0178] Finally, the server computing system (130) may, after providing the final target forecast and supporting information, receive input for changing the prediction environment from the user, re-perform a what-if simulation based on the input change in the prediction environment, thereby providing the target forecast and supporting information in the changed environment. (S113)

[0179] Specifically, referring to FIG. 10, a user can input a change in the prediction environment by changing the characteristics of a target influence variable that affects the target outlook or by inputting the occurrence of a specific event through a user computing device (110).

[0180] In an embodiment, if there is a change in a target influence variable, the server computing system (130) may change the integrated structured data set according to the changed target influence variable and then re-execute the process of interpreting the target outlook and basis, and output the target outlook and basis information according to the simulation and provide them to the user computing device (110).

[0181] In another embodiment, the server computing system (130) may receive input indicating a change in the predicted environment due to the occurrence of a specific event. In this case, if the occurrence of a specific event can be quantitatively reflected in the event list, the server computing system (130) may produce changed quantitative data, then based on this, modify the integrated structured data set, and then re-execute the process of interpreting the target forecast and basis, thereby outputting the target forecast and basis information according to the simulation and providing them to the user computing device (110).

[0182]

[0183] Although the detailed description of the present invention has been described with reference to preferred embodiments of the present invention, it will be understood by those skilled in the art or having ordinary knowledge in the art that various modifications and changes can be made to the present invention without departing from the spirit and technical scope of the present invention as set forth in the claims below. Accordingly, the technical scope of the present invention should not be limited to the contents described in the detailed description of the specification, but should be defined by the claims.

[0184] The present invention is a method and system for predicting future prospects for a target by analyzing structured data and unstructured data at a semantic level, and thus has industrial applicability.

Claims

1. A target prediction method for predicting the future outlook of a target performed by a computing device. A step of receiving a target prediction request from a user; A step of determining a target prediction element to be predicted from the target prediction request received above; A step of retrieving a target outlook report for a target of the above-determined target prediction element, and generating relationship information between target-target influence variables at a semantic level based on the retrieved target outlook report; A step of filtering unstructured data based on the target influence variable of the generated relationship information and the structured data of features related to the target influence variable and the text document related to the target influence variable; A step of calculating future target outlook based on the above filtered structured data and unstructured data; A step of generating the basis of the above-mentioned target outlook as relationship information at the feature level; and A step of providing the user with the target outlook produced above and relationship information at the feature level. Target prediction method.

2. In paragraph 1, The step of receiving a target prediction request from the above user is as follows: A step of providing a chat interface to the user to receive a text including the target prediction request; A step of analyzing the received text based on context to detect a context indicating a target prediction request. Target prediction method.

3. In paragraph 2, The step of determining the above target prediction elements is: Further comprising a step of performing Named Entity Recognition on the text including the target prediction request to determine keywords representing the target, the total outlook period, and the prediction unit period as the target prediction elements. Target prediction method.

4. In paragraph 3, The step of determining the above target prediction elements is: If a plurality of target keywords of a higher concept and target keywords of a lower concept are recognized for the target of the target prediction element, a step of listing the plurality of recognized target keywords and providing them for the user to select is further included. Target prediction method.

5. In paragraph 1, The step of generating the above relationship information is: A step of defining target influence variables that affect the above target at a semantic level, A step of generating a causal relationship graph with the names of each defined target influence variable as node names as relationship information. Target prediction method.

6. In paragraph 5, The step of generating the above relationship information is: It further includes a step of indicating the causal relationship between target influence variables represented by each node of the above causal relationship graph using arrows. Target prediction method.

7. In paragraph 1, The step of filtering unstructured data from the structured data of the above features and the text documents related to the target influence variables is as follows: A step of classifying features stored in the data store into target influence variables defined at the above semantic level, A step of generating a structured data set by combining the structured data of features classified as the above target influence variables. Target prediction method.

8. In paragraph 1, The step of filtering unstructured data from the structured data of the above features and the text documents related to the target influence variables is as follows: A step of inputting the above text document into a document classification prompt template and determining through a language model whether the above text document is a document that affects the target. Target prediction method.

9. In paragraph 8, The step of calculating the above future target outlook is: A step of detecting a target outlook report predicting the outlook of the target among the above text documents, A step of performing sentiment analysis on sentences predicting the target in the target outlook report using a language model, and classifying the outlook value of the target as positive, neutral, or negative for each target outlook report; A step of quantifying the level of the tone of the above-mentioned classified emotions and returning it as predicted scoring data, A step of generating quantitative data by listing the predicted scoring data of the above target outlook reports in chronological order. Target prediction method.

10. In paragraph 9, The step of calculating the above future target outlook is: A step of combining the above structured data and the above quantitative data to create an integrated structured data set, A step of inputting the above-generated integrated structured data set into a prediction model and outputting a target forecast value. Target prediction method.

11. In paragraph 10, The step of calculating the above future target outlook is: Further comprising a step of adjusting the target outlook based on relationship information between the target-target influence variables. Target prediction method.

12. In paragraph 1, The step of generating relationship information at the above feature level is: A step of generating relationship information of the characteristics of target influence variables that serve as a basis for predicting the above target outlook. Target prediction method.

13. In paragraph 12, The step of generating relationship information at the above feature level is: A step of generating the relationship information including the numerical values ​​of the features that influenced the predicted target outlook at one point in time. Target prediction method.

14. In paragraph 1, When receiving a predicted environment change input from the user, the method further includes a step of performing a simulation according to the received predicted environment change input. Target prediction method.

15. In paragraph 14, The step of performing a simulation according to the predicted environment change input received above is: If there is a change in the characteristics of the above target influence variable, a step of changing the standard data of the characteristics according to the changed target influence variable, A step of re-executing the process of interpreting target forecasts and grounds based on the above-mentioned changed structured data and unstructured data to output target forecasts and grounds information according to the assumption simulation. Target prediction method.

16. In paragraph 1, The step of performing a simulation according to the predicted environment change input received above is: When receiving a specific event occurrence from the user as a predicted environment change input, the method further includes a step of detecting a case similar to the specific event occurrence and calculating a target forecast value based on the detected case. Target prediction method.

17. A server computing system that receives a target prediction request from a user computing device and performs a target prediction task, A data store that stores target prediction related data; A memory for storing commands and data for performing the above target prediction task; and At least one processor that performs the target prediction task according to the instructions and data of the memory, At least one processor above, Receive a target prediction request from a user, Determine the target prediction element to be predicted from the target prediction request received above, Retrieving a target outlook report for the target of the above-determined target prediction element, and generating relationship information between target-target influence variables at a semantic level based on the retrieved target outlook report, Based on the target influence variable of the relationship information generated above, the structured data of the feature related to the target influence variable and the unstructured data of the text document related to the target influence variable are filtered, Based on the above filtered structured and unstructured data, future target outlook is calculated, The basis of the above-mentioned target outlook is generated as relationship information at the above-mentioned feature level, A target prediction system that provides the user with the target outlook produced above and relationship information at the feature level.

Images