vizion 0.1.7

Smart EDA and model recommendation helper

Vizion

Vizion is a lightweight, notebook-friendly helper for quick exploratory data analysis (EDA).
It bundles the repetitive parts of dataset inspection—summaries, missing data checks, visualizations, outlier handling, and walkthrough tips—into a single class so you can focus on insights.

Installation

pip install vizion

or, for local development:

pip install -e .

Quick Start

import pandas as pd
import vizion as vz

df = pd.read_csv("data.csv")

# Simple summary in one line
vz.quick_summary(df)

# Visual EDA helpers
vz.plot_numeric_eda(df, plot_types=["hist", "box"])
vz.plot_categorical_eda(df, target="SalePrice")

# Data hygiene helpers
df_clean, report = vz.handle_missing(df)
df_no_outliers = vz.handle_outliers(df_clean)

# Guided workflow
vz.help_steps()

API Reference

All functions live under the Vizion class but are also re-exported at module level (vizion.quick_summary, etc.).

• quick_summary(df, show=True)
  Prints (or returns) dataset shape, numeric/categorical/datetime column counts, duplicated rows, and top missing columns.

• missing_value_summary(df)
  Returns a DataFrame detailing missing counts, percentages, dtypes, and suggested remediation per column.

• get_column_types(df)
  Identifies numeric, categorical, and missing columns, storing them in globals (num_col, cat_col, mis_col) for quick reuse.

• plot_numeric_eda(df, columns=None, plot_types=None, ..., show_corr=True, show_pairplot=False)
  Generates grid plots (hist, kde, box, violin, scatter, line, area) for numeric columns, with optional correlation heatmap and pairplot.

• plot_categorical_eda(df, columns=None, plot_types=None, ..., top_n=20)
  Creates count/pie plots for categorical features, and supports target-aware bar/box/violin visuals.

• handle_outliers(df, cols=None, method="auto", threshold=1.5, visualize=True, ...)
  Detects numeric outliers via IQR or IsolationForest, optionally caps/removes values and shows before/after boxplots.

• handle_missing(df, drop_threshold=0.75, numeric_strategy="median", categorical_strategy="mode", datetime_strategy="ffill")
  Drops ultra-sparse columns and fills the rest according to strategy, returning the cleaned DataFrame plus a fill/drop report.

• generate_doc(filename=None)
  Auto-builds Markdown docs for every Vizion method; prints to stdout or saves to a file.

• help_steps()
  Prints a recommended, step-by-step EDA workflow that chains Vizion helpers in a sensible order.

Contributing

Issues and PRs are welcome! See the open tasks in the GitHub repo or propose your own improvements (encoding, scaling, imbalance checks, etc.).

License

MIT © Milind Chaudhari