Weekend STEM Robotics Bootcamp for Kids in Jurong East, Singapore

1. Why Exploratory Data Analysis (EDA) Matters

Raw client data is rarely clean. EDA is the structured process of exploring, visualizing, and summarizing a dataset before modeling—to assess quality, spot patterns and anomalies, and build domain intuition.

EDA also supports client communication: findings and hypotheses in an EDA report can be checked against real business knowledge so modeling assumptions stay realistic.

2. Primary Goals of EDA

1. Identify patterns and trends — relationships between business variables
2. Detect anomalies and outliers — extreme values that may be real events or errors
3. Assess data quality — missing values, duplicates, remediation plans
4. Classify feature types — numeric, categorical, datetime, text
5. Formulate and validate hypotheses — ask why trends exist and review with the client

3. Essential Pandas Diagnostics

import pandas as pd

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

print("Dimensions (Shape):", df.shape)
print("\nColumn Data Types:")
print(df.dtypes)
print("\nMissing Values per Column:")
print(df.isna().sum())
print("\nDescriptive Summary:")
print(df.describe())
print("\nCategory Distribution:")
print(df["Category"].value_counts())

• df.shape — rows and columns
• df.dtypes — type mismatches (e.g. Sales as object → text mixed in)
• df.isna().sum() — missing counts per column
• df.describe() — mean, spread, min/max for numeric columns
• value_counts() — frequency of categorical classes

4. Key Questions to Ask

1. How many missing values exist in each column?
2. What is the primary target variable (e.g. Revenue)?
3. Which features are numerical vs categorical?
4. What is the distribution across product categories?
5. Which products drive the highest sales volume?
6. Are extreme transaction amounts legitimate or errors?
7. How do purchases vary by region and season?

5. Structuring an EDA Report

• Define goals upfront
• Identify target and predictor features with stakeholders
• Anchor tables and charts to specific questions
• Document assumptions, anomalies, and quality fixes for client review before modeling