KR-20260064376-A - AI-Based Agricultural Diary Analysis System

Abstract

This invention structures data from farming logs recorded by users, converts it into analyzable information, and provides technical information to enhance the efficiency of agricultural management based on this. The system automatically processes farming log data entered by the user, extracts key information (e.g., material usage, work details, dates, etc.), and stores it in a database. Subsequently, based on the information stored in the database, it provides comparative data on past records and farming methods of other farms, and includes a module that predicts expected economic performance (e.g., sales, net profit, etc.) based on resource usage. Through this, users can optimize current resource usage by comparing past farming methods and resource input details via the system. By predicting the economic performance of agricultural management and suggesting optimal resource allocation plans, the system effectively supports farmers in making more efficient decisions.

Inventors

• 이충우

Assignees

• 이충우

Dates

Publication Date

20260507

Application Date

20241031

Claims (1)

1. Regarding the method of analyzing farming logs using AI models, A step (S1) of extracting key information related to agriculture using a natural language processing model (NLP) to automatically refine text data from an input farming log; A step (S2) of generating comparative data that users can refer to by analyzing similarity using similarity analysis techniques and time series analysis models, using the data extracted in the above step, past farming records, and farming data from other farms; A step (S3) of learning the relationship between resource input and economic performance using an AI model and predicting future economic performance; A step (S4) of generating resource input scenarios, analyzing performance for various scenarios to derive an optimal resource usage plan, and providing it to the user; The process proceeds to the step (S5) of visualizing predicted performance result data and providing it to the user in an interface format with visualized data added to the farming log. Method for analyzing farming logs using an AI model.

Description

AI-Based Agricultural Diary Analysis System for Resource Optimization and Performance Prediction The present invention relates to a recording medium called a "farming log," and to a system that makes it easier to understand farming conditions by analyzing the farming log. A farming log is a medium for recording the status of farmland by date. By documenting work dates, land conditions, pesticides and fertilizers applied, and work details, it serves as a tool that allows users (farmers) to systematically monitor crop management. Current farming log management methods rely on manual entry or Excel spreadsheets; however, relying solely on these methods prevents users from viewing farming indicators at a glance and fails to provide significant assistance in decision-making. Figure 1 is a diagram showing the overall flowchart (flowchart) of the system. FIG. 1 illustrates the overall flow of a system according to the present invention. As illustrated, the system according to the present invention provides various functions to convert a farming log entered by a user into data, extract key information, compare it with farming methods of past and other farms, and predict resource usage and economic performance. First, when a user uploads a farming log file, the system stores it in a database (130) and passes it to a data analysis and structuring step (120). In the data analysis and structuring stage (120), various preprocessing operations are performed to refine text data and convert it into an analyzable format. In this stage, natural language processing (NLP) libraries such as spaCy or NLTK are utilized to remove unnecessary symbols (121) and stop words (122), and sentence tokenization (123) and word tokenization (124) are applied to divide the text into sentences and words. Through this, the text data is subdivided and prepared so that key information can be separated and structured. Additionally, tasks are performed to identify and normalize specific patterns, such as numbers, units, and dates, using regular expressions. Here, a regular expression is a string processing tool used to find specific patterns in a string and to extract or replace matching parts. The extracted key entities, such as crop name, material usage, and work date, are organized into systematized data by mapping them by field through an information extraction (125) process, and are converted into a structured format using a data processing library such as Pandas and stored in a database (130) such as MongoDB or PostgreSQL. The comparison module (140) generates comparison data between the current user's farming method and past records and the methods of other farms based on data stored in the database (130). In this process, similarity analysis techniques are used to identify differences in major work behaviors and material usage patterns in order to compare the amount of resources and work methods used at a specific time in the past with the current farming method. This comparison is implemented by comparing resource usage patterns in chronological order using a time series analysis library (e.g., statsmodels). The system displays the information as reference material depending on whether the past records and the methods of other farms meet the criteria, and presents it simply as comparison data without determining that a specific method is the correct answer. Within the comparison module, past record comparison (141) and other farm comparison (142) are performed. The resource usage and economic prediction module (150) predicts economic indicators such as expected sales, operating profit, and net profit based on the resource input history recorded by the user. To this end, it learns the relationship between the current resource input and production volume by reflecting market consumption data and consumption data from past farming records, and thereby predicts future consumption potential. In this stage, regression analysis and machine learning models are used to model the relationship between resource input and economic indicators, and tree-based non-linear prediction models such as Random Forest from the scikit-learn library are utilized to predict performance according to various resource usage scenarios. Additionally, based on past records and market data, it performs time series and specific period modeling between the two datasets to predict future resource consumption and production volume. It provides an optimal resource usage plan through resource input learning (151) and scenario generation and optimization (152). Finally, the prediction results are visualized through the performance prediction result visualization module (150), allowing the user to output a farming log including the analysis results. In this step, a visualization library such as Matplotlib or Seaborn is used to provide key indicators such as sales, costs, and net profit as graphs (160), which helps the user easily understand the economic performance of each scenari