Digital Animation OpenToonz Classes for Jeju Global Education City

1. Purpose of an EDA Report

After cleaning, summarizing, detecting outliers, and measuring correlations, findings are packaged into a short EDA report for stakeholders. It turns raw tables into actionable insights and lets the client (domain SME) validate interpretation before modeling decisions are locked in.

2. Narrative + Visual Evidence

A strong report mixes concise bullets with charts/tables: quality issues, distributions, relationships, and recommended fixes. Keep it short—typically a focused 6–8 bullet executive summary—rather than dozens of pages.

3. Six Summary Dimensions

1. Data quality & missingness — rates and proposed remedies (e.g. median impute or drop)
2. Feature distributions — skewness and transforms (e.g. log1p for right-skewed Revenue)
3. Attribute relationships — correlation strengths that match (or challenge) business intuition
4. Categorical concentration — share of orders/revenue in top categories
5. Outliers & anomalies — extreme records to keep or remove with client input
6. Temporal / seasonal trends — cycles such as Q4 holiday spikes

4. Practical Snippets

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

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

missing_rate = df["Customer_Rating"].isna().mean() * 100
print(f"Missing Customer_Rating: {missing_rate:.1f}%")

print("Revenue skew:", df["Revenue"].skew())
df["Revenue_Log"] = np.log1p(df["Revenue"])

print("Qty–Revenue corr:", df["Quantity"].corr(df["Revenue"]))
print("Top-3 category share:", df["Category"].value_counts(normalize=True).head(3).sum())

Pair metrics with simple plots (e.g. raw vs log Revenue histograms) so stakeholders see the problem and the proposed fix.

5. Bridge to Modeling

Share the remediation checklist with the client before dropping rows, imputing, or filtering outliers. Agreement “freezes” preprocessing choices and anchors the modeling pipeline in domain-approved assumptions.