DataMatrix — Interactive EDA

v0.3.0

One-call exploratory data analysis: overview, per-column profiles by inferred type, signed correlation structure, quality alerts, highlights and an embedded data explorer with stratified sampling.

Quick Start

eda.py

python

import viewx as vx

df = vx.load_dataset("titanic.csv")

dm = vx.DataMatrix(df)
dm.analyze()          # compute profiles, correlations, alerts
dm.save("eda.html")   # self-contained interactive report

analyze() returns the DatasetReport, also accessible later via dm.summary(). The HTML report works fully offline-except-CDN and includes an embedded data explorer with a stratified sample.

What the Report Contains

Overview

Shape, memory usage, duplicate rows and per-type column counts.

Column profiles

Numeric, categorical, datetime and boolean strategies: ranges, top values, missingness, cardinality.

Correlations

Signed correlation matrix and top correlated pairs.

Alerts & highlights

Data-quality warnings (high missingness, constant columns, skew) and notable findings.

Data Manipulation

DataMatrix also carries a small toolkit to prepare the data before analyzing:

prepare.py

python

dm = vx.DataMatrix(df)

dm.clean_data()                 # trim strings, drop empty rows/cols
dm.filter_rows("body_mass_g > 3000")
dm.select_columns(["species", "bill_length_mm", "body_mass_g"])

dm.analyze()
dm.save("eda.html")

API Reference

DataMatrix

DataMatrix(df, title='DataMatrix')

analyze

analyze(theme=None) -> DatasetReport

DatasetReport

summary / alerts / highlights

summary() / alerts() / highlights()

clean_data / filter_rows / select_columns

clean_data() / filter_rows(expr) / select_columns(cols)

save / show / render_html

save(path) / show() / render_html()

Large Tables

The embedded explorer does not inline the full DataFrame: it ships a stratified sample (with a configurable row cap) so reports stay lightweight even for datasets with millions of rows.