In the age of data-driven decision-making, preprocessing and exploratory data analysis (EDA) are critical yet time-consuming stages of the machine learning lifecycle. SmartEDA is an intelligent, automated framework that simplifies and unifies data preprocessing, visualization, and model building for both regression and classification problems. The system dynamically detects the nature of the target variable and performs context-specific transformations, including outlier detection, encoding, scaling, and feature engineering. It generates interpretable EDA plots, downloadable reports, and trains baseline models—all through an intuitive web interface. Powered by open-source tools like pandas, scikit-learn, ydata-profiling, and generative AI via Gemini, SmartEDA bridges the gap between raw data and model-ready pipelines. Experiments on benchmark datasets reveal a reduction of preprocessing and analysis time by over 80% without compromising accuracy or interpretability, making it ideal for both novice users and experienced practitioners.

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

SmartEDA: Automating EDA and Model Building for Structured Data

  • Saurabh Shisode,
  • Sharayu Wankhade,
  • Sarthak Shah,
  • Sanika Rasal,
  • Vishakha Tayade,
  • Gaurav Shinde

摘要

In the age of data-driven decision-making, preprocessing and exploratory data analysis (EDA) are critical yet time-consuming stages of the machine learning lifecycle. SmartEDA is an intelligent, automated framework that simplifies and unifies data preprocessing, visualization, and model building for both regression and classification problems. The system dynamically detects the nature of the target variable and performs context-specific transformations, including outlier detection, encoding, scaling, and feature engineering. It generates interpretable EDA plots, downloadable reports, and trains baseline models—all through an intuitive web interface. Powered by open-source tools like pandas, scikit-learn, ydata-profiling, and generative AI via Gemini, SmartEDA bridges the gap between raw data and model-ready pipelines. Experiments on benchmark datasets reveal a reduction of preprocessing and analysis time by over 80% without compromising accuracy or interpretability, making it ideal for both novice users and experienced practitioners.