Keşifsel Veri Analizi (EDA) Nedir?

In a data analysis process, building a model, producing charts or running statistical tests is usually the visible part of the work. But there is a much more fundamental step that comes before all of it: getting to know the data.

Exploratory Data Analysis (EDA) is the approach used to understand the structure of a data set, spot missing and unusual values, discover relationships between variables and extract as much information as possible from the data before the analysis begins.

“A good analysis starts with a good question, and a good question usually appears while you are exploring the data.”

Why Does EDA Matter?

Feeding a data set straight into a model can look tempting. But analyses made without understanding the structure of the data can lead to misleading results. EDA makes that risk visible before the analysis starts.

1. Getting to Know the Data Set

When starting EDA, the first thing to understand is what kind of data we actually have. On the Python side, pandas is a powerful and practical tool for this.

import pandas as pd

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

print(df.shape)
print(df.head())
df.info()
print(df.dtypes)

2. Looking at Basic Statistics

To see the general behaviour of a data set quickly, describe() is very useful. In a single line it gives an idea about central tendency, spread and extreme values.

df.describe()

# Including categorical columns
df.describe(include="all")

3. Exploring Missing Values

One of the most important checks during EDA is missing data analysis. Any calculation made without knowing how many gaps each column has can be misleading.

# Number of missing values per column
df.isna().sum()

# Seeing it as a ratio is easier to read
df.isna().mean().sort_values(ascending=False)

If a column such as income has a large number of missing values, for example, you need a separate decision about how that variable will be used in the analysis.

4. Checking Duplicate Records

The same record appearing more than once is one of the problems you meet especially when different data sources are combined.

# Number of duplicates
df.duplicated().sum()

# Display the duplicate records
df[df.duplicated()]

# Remove them if needed
df = df.drop_duplicates()

It is important to check whether two similar-looking records really are duplicates; two transactions that look the same can be different events.

5. Examining Outliers

Outliers are observations that clearly separate from the general distribution in a data set. But being statistically unusual does not mean the value is wrong.

Q1 = df["sales"].quantile(0.25)
Q3 = df["sales"].quantile(0.75)

IQR = Q3 - Q1
lower = Q1 - 1.5 * IQR
upper = Q3 + 1.5 * IQR

outliers = df[(df["sales"] < lower) | (df["sales"] > upper)]

6. Exploring the Distribution of Variables

Histograms, boxplots and density plots can be used to see how a variable is distributed. The mean alone does not describe the shape of a distribution.

import matplotlib.pyplot as plt

df["sales"].hist(bins=30)
plt.show()

7. Examining Categorical Variables

EDA is not only about numerical variables. Seeing the frequencies of categorical fields is also important for understanding the data set.

df["category"].value_counts()

# Proportional distribution
df["category"].value_counts(normalize=True)

This analysis shows which categories the data set concentrates in and whether the sample is balanced.

8. Discovering Relationships Between Variables

One of the most valuable sides of EDA is examining variables together rather than one by one. For example, we can ask whether there is a relationship between income and sales.

df.plot(kind="scatter", x="income", y="sales")
plt.show()

# Correlation between numerical variables
df.corr(numeric_only=True)

9. Analysis by Group

Totals sometimes hide the differences inside a data set. That is why analysing by categories or segments matters.

df.groupby("category")["sales"].agg(["count", "mean", "median", "std"])

The Questions I Ask Myself During EDA

A Simple EDA Flow in Python

All the basic checks can be combined into a small flow. This is the template I run first whenever a new data set arrives:

import pandas as pd

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

print(df.shape)
print(df.head())
df.info()
print(df.isna().sum())
print(df.duplicated().sum())
print(df.describe())
print(df["category"].value_counts())
print(df.corr(numeric_only=True))

EDA Is Not a Result, It Is a Beginning

The purpose of EDA is not only to say “the data is clean”. The real purpose is for the data to show us which questions we should be asking.

We may find a striking relationship between income and sales, for example. That observation raises new questions such as “Do sales grow as income grows?” or “Is this relationship the same in every segment?”.

This is why EDA moves beyond a simple data check and becomes the starting point of analytical thinking.

Conclusion

Exploratory Data Analysis is one of the most important stages of the data analysis process. Before building a model, running a statistical test or preparing a dashboard, we need to understand what the data in front of us really says.

Python and pandas offer strong tools for this. Even basic functions such as head(), info(), describe(), isna(), duplicated(), groupby() and correlation analysis can tell a lot about a data set.

EDA in short: Know the data → Find the problems → Examine the distributions → Discover the relationships → Form the questions → Move to analysis.