I have choosen dataset of Iris and is perhaps the best known database to be found in the pattern recognition literature. Fisher's paper is a classic in the field and is referenced frequently to this day. (See Duda & Hart, for example.)
The dataset features three classes of Iris plants with 50 examples per class. Crucially for classification studies, it presents a challenge where one flower type can be easily distinguished from the rest two using a straight line, but the remaining two classes exhibit significant overlap and cannot be separated linearly.
This inherent challenge necessitates a deep understanding of Python's visualization capabilities, transforming their effective use into a fundamental requirement for mastering data analysis and thus becomes a cornerstone in gaining expertise to tackle such complex, real-world classification problems.
- sepal length in cm
- sepal width in cm
- petal length in cm
- petal width in cm
- class: -- Iris Setosa -- Iris Versicolour -- Iris Virginica
We will use five libraries for the tutorial:
- pandas: Data Loading, Cleaning, and Manipulation.
- matplotlib: The Visualization Foundation and Customization Engine.
- sklearn: Data Sourcing and Machine Learning Utilities.
- numpy: Numerical Computing and Array Operations.
- seaborn: High-Level Statistical Data Visualization.
Matplotlib and Seaborn are the premier Python tools for data visualization, a crucial process that drives insight, pattern recognition, and deeper understanding in data analysis. While both follow a three-phase methodology, their implementation differs.
Phase 1: Data Preparation and Understanding
This initial phase is identical for both libraries and is necessary to ensure data quality and clarity. Define the Goal: What specific relationships or distributions are you trying to communicate? Load and Inspect Data (Pandas): Use df.head(), df.info(), and df.describe() to check structure, data types, and potential issues (like missing values). Data Transformation: Aggregate, filter, or clean the data to prepare it for the specific chart type (e.g., calculating means for a Bar/Radar chart).
Phase 2: Visualization and Exploration This phase is where the approaches diverge. Matplotlib requires more steps to build the visual elements, while Seaborn handles most of the statistical mapping internally.
| Library | Primary Focus | Typical Syntax | Key Methodology |
|---|---|---|---|
| Matplotlib | Flexibility and Customization | ax.plot(), ax.scatter() | Explicit Plotting: You manually map data arrays (X, Y) to visual elements (color, marker) on a manually created axes object. |
| Seaborn | Statistical Graphics & Aesthetics | sns.scatterplot(), sns.histplot() | Declarative Mapping: You tell Seaborn the DataFrame (data=df), the variable for the x-axis (x='feature'), and the grouping variable (hue='category'). Seaborn handles aggregation and styling automatically. |
Phase 3: Refinement and Communication This final phase relies heavily on Matplotlib's foundational layer for fine-tuning.
| Step | Purpose | Matplotlib/Seaborn Tool |
|---|---|---|
| Labeling | Add titles, axis labels, and legends. | plt.title(), ax.set_xlabel(), ax.legend() |
| Aesthetics | Adjust colors, line styles, and grids. | Matplotlib commands (ax.grid(), ax.tick_params()) or Seaborn color palettes (palette=...). |
| Clarity | Prevent overlapping elements. | plt.tight_layout() |
| Saving | Finalize the image for presentation. | plt.savefig('output.png', dpi=300) |
Following Steps are essential in data visualization using python
The process begins by importing the Matplotlib library, usually its pyplot module, which offers a MATLAB-like interface for plotting:
import matplotlib.pyplot as plt
Visualization in a Jupyter Notebook: Matplotlib magic command ensures inline rendering:
%matplotlib inline
- Loading and Inspecting Data Data can be imported from sources such as CSV, Excel, or SQL databases using libraries like pandas:
import pandas as pd
data = pd.read_csv('data.csv')
Initial glance of data
| Command | Purpose |
|---|---|
| df.head() | Displays the first 5 rows of the DataFrame. Crucial for seeing the actual data format. |
| df.tail() | Displays the last 5 rows. Useful for checking if any end-of-file artifacts or summaries exist. |
| df.shape | Returns a tuple (rows, columns), telling you exactly how large the dataset is. |
| df.columns | Lists the names of all columns (features). |
Structure and Data Types
| Command | Purpose |
|---|---|
| df.info() | Provides a summary of the DataFrame: the number of entries, column names, the count of non-null values for each column, and the data type (dtype) of each column. |
| df.dtypes | Lists the data type for every column. Essential for ensuring numerical columns are treated as numbers (e.g., float64 or int64) and categories are objects. |
Statistical Summary
| Command | Purpose |
|---|---|
| df.describe() | Generates descriptive statistics for all numerical columns, including: count, mean, standard deviation, minimum, quartiles (25%, 50%, 75%), and maximum. |
| df.describe(include='object') | "Generates descriptive statistics for all categorical (object/string) columns, including: count, unique values, top value, and its frequency. |
Uniqueness and Specific Values
| Command | Purpose |
|---|---|
| df['column_name'].unique() | Lists all unique values in a specific column. |
| df['column_name'].nunique() | Returns the count of unique values in a specific column. |
| df['column_name'].value_counts() | Returns a Series showing the frequency of each unique value, ordered from most to least frequent. Excellent for checking class imbalance in target variables. |
- Selecting the Appropriate visualization charts depends on your objective, data type, and relationship you want to highlight.
| Visualization Type | Description | Typical Use Case |
|---|---|---|
| Line Plot | Displays data trends or changes over a continuous interval or time period. | Tracking metrics over time (e.g., production, temperature). |
| Bar Plot | Represents categorical data with rectangular bars proportional to their values. | Comparing quantities across different categories. |
| Count Plot | Similar to a bar plot but automatically counts the occurrences of categorical variables. | Showing the frequency distribution of categorical features. |
| Histogram | Shows the frequency distribution of numerical data divided into bins. | Understanding distribution and skewness of continuous variables. |
| Density Plot | Smooth version of a histogram that estimates the probability density function of continuous data. | Visualizing data distribution and comparing multiple distributions. |
| Pie Diagram | Circular chart divided into slices to illustrate numerical proportions. | Showing parts of a whole (e.g., market share). |
| Donut Diagram | Similar to a pie chart but with a central hole, enhancing readability. | Emphasizing proportions while allowing space for central labels. |
| Scatter Plot | Displays relationship between two numerical variables using dots. | Detecting correlation or clustering patterns. |
| Pair Plot | Plots pairwise relationships between all numerical features in the dataset. | Quick multivariate analysis and correlation detection. |
| Facet Grid Plot | Creates a grid of subplots based on combinations of categorical variables. | Comparing distributions or relationships across multiple subsets of data. |
| Joint Plot | Combines a scatter plot with histograms or density plots for both axes. | Analyzing relationships between two variables with marginal distributions. |
| Regression Plot | Displays scatter data along with a fitted regression line. | Evaluating linear relationships and trends between variables. |
| 3D Scatter Plot | Extends scatter plot to three dimensions using x, y, and z axes. | Visualizing spatial relationships or high-dimensional data patterns. |
| Box Plot | Summarizes distribution using median, quartiles, and potential outliers. | Comparing spread and detecting outliers across groups. |
| Violin Plot | Combines box plot with a density plot to show data distribution and spread. | Understanding both summary statistics and data distribution shape. |
| Swarm Plot | Displays individual data points along a categorical axis with minimal overlap. | Showing all observations in a categorical variable while maintaining clarity. |
| Strip Plot | Similar to a swarm plot but allows overlapping of data points. | Visualizing raw data distribution for small datasets. |
| Bubble Plot | Extension of scatter plot where bubble size represents an additional variable. | Visualizing three variables simultaneously (x, y, and bubble size). |
| Radar Plot (Spider Chart) | Plots multiple quantitative variables on axes starting from a common center. | Comparing feature profiles across categories (e.g., species characteristics). |
| Correlation Heatmap | Visualizes pairwise correlation coefficients between variables using color gradients. | Identifying strength and direction of relationships between variables. |
| Plot Type | Matplotlib Function | Purpose |
|---|---|---|
| Relationship | plt.scatter() | Show correlation or cluster separation (e.g., Iris data). |
| Distribution | plt.hist()\ plt.boxplot()\ sns.violinplot() | Show the frequency, spread, and central tendency of a single variable. |
| Comparison | plt.bar()\ plt.plot() | Compare quantities across different categories or track trends over time. |
plt.scatter(data['x'], data['y'])
plt.title('Relationship between X and Y')
plt.xlabel('X Variable')
plt.ylabel('Y Variable')
plt.show()
- Customizing the Visualization
Matplotlib allows customization of almost every visual element to improve readability:
Titles and labels for context
Legends for category identification
Color maps and styles for aesthetics
Gridlines and annotations for emphasis
plt.plot(data['Year'], data['Sales'], color='teal', linestyle='--', marker='o')
plt.title('Annual Sales Trend')
plt.xlabel('Year')
plt.ylabel('Sales (in USD)')
plt.grid(True)
plt.show()
- Combining Multiple Plots
Subplots help compare multiple visualizations side by side for deeper analysis:
fig, ax = plt.subplots(1, 2, figsize=(10, 5))
ax[0].hist(data['Revenue'], bins=20, color='skyblue')
ax[1].boxplot(data['Revenue'])
plt.show()
- Interpreting and Communicating Insights
The final step involves interpreting patterns, anomalies, and relationships revealed by the plots. Clear labeling, color consistency, and concise legends help communicate insights effectively.
- Saving and Sharing Visualizations
Plots can be exported in multiple formats for reports or dashboards:
plt.savefig('visualization.png', dpi=300, bbox_inches='tight')
Included with the repository is Visualization jupyter notebook, jupyter notebook with output and python code of Iris dataset using Matplotlib.