Exploratory Data Analysis Notes – Class 12 Data Science (844)

In these Exploratory Data Analysis Notes for Class 12 DS (CBSE), you’ll learn about EDA, Univariate Analysis, Bivariate Analysis, Multivariate Analysis, Graphical Analysis, and Clustering with simple, pointwise notes prepared as per the latest CBSE curriculum.

Introduction to Exploratory Data Analysis (EDA)

• Exploratory Data Analysis (EDA) is the process of performing an initial analysis of the available data to understand it better.
• The main purpose of EDA is to identify patterns, detect anomalies (unusual observations), and test any hypotheses or assumptions about the data.
• EDA uses summary statistics and graphical representations to analyze and understand the data.
• It is an approach for analyzing datasets and summarizing their key characteristics, often with the help of visual methods.
• In real-life applications, EDA mainly relies on graphical techniques because graphs help explore data quickly and identify patterns more effectively than statistical methods alone.
• EDA can be performed using various tools and methods, including:
  • Univariate Analysis: Analyses a single feature (variable) using summary statistics and visualizations.
  • Bivariate Analysis: Analyses the relationship between two variables using summary statistics and visualizations.
  • Multivariate Analysis: Analyses multiple variables together to understand the interactions between different features.
  • Graphical Analysis: Uses graphs such as histograms, probability plots, lag plots, mean plots, standard deviation plots, and box plots to visualize the data.
  • Clustering: Uses unsupervised learning techniques to identify groups (clusters) of similar data points. It is commonly used in applications like image compression and pattern recognition.

Univariate Analysis

• Univariate analysis is the simplest form of data analysis in which only one variable from the dataset is analysed.
• Since only one variable is studied, there is no need to analyze causes or relationships with other variables.
• The main purpose of univariate analysis is to describe the data and identify patterns present in a single variable.
• In univariate analysis, one variable is selected and analyzed in detail. For example, variables such as revenue or height can be studied individually.
• Univariate analysis uses both statistical and graphical methods to analyze data.
• Common statistical methods used in univariate analysis include:
  • Mean
  • Median
  • Mode
  • Range
  • Variance
  • Maximum and Minimum values
  • Quartiles
  • Standard Deviation
• Common graphical methods used in univariate analysis include:
  • Frequency Distribution Table
  • Bar Chart
  • Histogram
  • Frequency Polygon
  • Pie Chart
  • Scatter Plot (for a single variable)
  • Box Plot – Shows the quantile ranges of the variable and helps identify outliers.
  • Histogram – Shows the frequency distribution of data over different value ranges.

Bivariate Analysis

• Bivariate analysis is a type of multivariate analysis that involves analyzing two variables from a dataset.
• It is usually performed between the target variable and another feature of the dataset.
• The main objective of bivariate analysis is to determine whether a relationship exists between the two variables.
• Bivariate analysis is commonly performed using graphical methods such as:
  • Scatter Plot
  • Line Chart
  • Pair Plot
• These graphs help visualize and understand the relationship between two variables.
• Bivariate analysis is also used to measure the correlation between two variables.
• For example, a market survey may analyze the relationship between the price of a product and its sales.

Multivariate Analysis

• Multivariate analysis is a statistical technique used to analyze more than two variables in a dataset.
• It is more complex than univariate and bivariate analysis because it studies the interactions among multiple variables simultaneously.
• The method used for multivariate analysis depends on the objective of the analysis.
• Some common multivariate analysis techniques include:
  • Canonical Correlation Analysis
  • Cluster Analysis
  • Contour Plots
  • Principal Component Analysis (PCA)

Data Cleaning

• Data cleaning is the process of identifying incorrect, incomplete, and inaccurate data, and improving the dataset by removing or replacing such data.
• It is an essential step in data preprocessing and plays a fundamental role in data science.
• A well-cleaned dataset helps machine learning algorithms learn more effectively and produce better insights.
• Different types of data may require different cleaning techniques, but a systematic approach can be applied to most datasets.

Steps of Data Cleaning

• Remove Duplicate Observations: Duplicate records often occur when data is collected from multiple sources or combined from different datasets. Removing duplicates helps improve the accuracy of the model.
• Remove Irrelevant Observations: Eliminate rows or columns that do not contribute to solving the problem, so the model learns only from relevant data.
• Remove Unwanted Outliers: Outliers can negatively affect the performance of some machine learning models. Unwanted outliers should be removed, while valid outliers should be retained.
• Fix Data Type Issues: Ensure that each variable has the correct data type. For example, numerical or date values stored as text should be converted to the appropriate data type.
• Handle Missing Data: Missing values should be managed before building a model. Common techniques include removing the affected rows or replacing the missing values with values close to the mean or mode of the variable.