Scholarlink

<p>This study aims to improve the accuracy and efficiency of predictive models for key agricultural indices in smart farming. Leveraging the growing role of technology, we developed a multi-output linear regression model and trained it on a multidimensional dataset that incorporates environmental and soil parameters, including air humidity, temperature, soil moisture, atmospheric pressure, pH, and light intensity. A notable aspect of this work is its ability to predict multiple indices simultaneously, capturing interdependencies (e.g., temperature and humidity) to support real-time agricultural monitoring and management. To ensure consistency, the time-series data were normalized and converted to Unix time, enabling flexible integration with various management systems. After preprocessing, the model achieved high performance (R² = 0.98, MAE = 0.76). We discuss the model’s strengths, limitations, and comparative evaluations against other algorithms, and suggest future directions, including data expansion, index enhancement, IoT integration, and real-time deployment.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Multi-Output Linear Regression Model for Real-Time Prediction of Agricultural Environmental Indices in Smart Farming Applications

  • 期刊论文

  • 发表日期：

    2026年6月16日

查看全文
  • Minh Son Nguyen,
  • Si Truong Do,
  • Thanh Q. Nguyen

摘要

This study aims to improve the accuracy and efficiency of predictive models for key agricultural indices in smart farming. Leveraging the growing role of technology, we developed a multi-output linear regression model and trained it on a multidimensional dataset that incorporates environmental and soil parameters, including air humidity, temperature, soil moisture, atmospheric pressure, pH, and light intensity. A notable aspect of this work is its ability to predict multiple indices simultaneously, capturing interdependencies (e.g., temperature and humidity) to support real-time agricultural monitoring and management. To ensure consistency, the time-series data were normalized and converted to Unix time, enabling flexible integration with various management systems. After preprocessing, the model achieved high performance (R² = 0.98, MAE = 0.76). We discuss the model’s strengths, limitations, and comparative evaluations against other algorithms, and suggest future directions, including data expansion, index enhancement, IoT integration, and real-time deployment.