AutoEDA: Iterative Data Focusing and Exploratory Analysis Based on Attribute Frequency

The paper proposes an automated data exploration and analysis method based on Attribute Frequency Statistical Feature Ratio (AFSFR). It integrates AutoVis and Data Preprocessing Methods to design and develop AutoEDA-Segment. Addressing the Concentrate on Field Sequences (CFS) problem in data exploration and analysis, this study employs various classification models and combines AFSFR with field type and the Elbow Inflection Point (EIP) of index features to design a field type identification and field value assessment method. For evaluating the effectiveness of focus analysis, the approach provides clustering visualization effects and an analysis scheme based on Field Type Search Tree (FST) and cluster comparison profiles, using a custom CFS approach. Additionally, to enhance the value of focused subset analysis data, the approach introduces a Parallel Coordinates-Based Data Filter (PCF), forming an EDA feedback loop to achieve Iterative Exploratory Data Analysis for User-inferred Cognition (IEDA-UC). Finally, we engaged graduate students with varying levels of experience in visualization research for collaboration and discussion, validating the effectiveness and feasibility of the approach using structured data from Kaggle.