Author(s): Farnazbanu

Originally published on Towards AI.

To maximize prediction accuracy, human participation is crucial during data preparation.

The first step in any ML activity starts with data. Data must be clean and properly preprocessed before working with it. Based on the type of learning, one can preprocess data. A human exploration is needed to understand the type and quality of the data and the relationships among data elements/attributes.

After the lead-in, the article walks through the typical preprocessing and modeling workflow: understanding data type and quality, exploring attribute relationships, detecting and remediating issues (like missing values), and applying preprocessing steps before training. It then introduces core dataset concepts (records/instances, attributes/features), explains major data types used in ML (qualitative/categorical vs quantitative/numeric, including nominal/ordinal and interval/ratio, plus discrete vs continuous), and shows how to inspect structure with tools like a data dictionary or library functions. The article continues with practical data exploration methods: handling outliers, using descriptive statistics for numerical data (central tendency, dispersion via variance/standard deviation, quantiles and box plots, and histogram interpretation), summarizing categorical data (unique values, counts, percentages, and modes), and visualizing relationships between variables using scatter plots and cross-tabulation for categorical pairs. It concludes with a brief wrap-up emphasizing that these fundamentals support effective, quality-controlled ML training.

Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI