Set out the aims, techniques and examples of exploring data

Data exploration has three parts. Exploratory data analysis summarises and visualises the data: histograms, bar charts and box plots for one feature, scatterplots and correlation matrices for pairs, and word clouds for the most frequent text tokens. Feature selection keeps only features that help explain the target, reducing noise and overfitting. Feature engineering creates new features from existing ones, for example ratios, logs, or one-hot encoding a category into indicator variables. The aim is to understand the data and give the model the most useful inputs before training starts. The common confusion is between selection, choosing among existing features, and engineering, creating new ones from them.