DS302 Data Visualization Quiz

Fundamentals of Data Science

• Course title: DS302
• Instructor: Dr. Nermeen Ghazy

Reference Books

• Data Science: Concepts and Practice, Vijay Kotu and Bala Deshpande, 2019
• DATA SCIENCE: FOUNDATION & FUNDAMENTALS, B. S. V. Vatika, L. C. Dabra, Gwalior, 2023

Lecture 4: Data Visualization

• Data visualization is a crucial technique for data discovery and exploration.
• While not strictly a data science technique, visual mining and pattern discovery are increasingly employed in data science, particularly in business.
• Data visualization represents data in an abstract visual form, enabling easy comprehension of complex data with multiple attributes and their relationships.

Data Visualization Motivations

• Comprehension of dense information: Visual charts easily display thousands of data points, offering a broader understanding than numerical representations, allowing for insights into trends.
• Relationships: Visualizing data in Cartesian coordinates allows exploration of relationships between attributes. While three or more attributes can't be easily displayed, creative solutions using size, color, shape, and flow maps can help illustrate complex relationships.

Univariate Data Visualization

• Exploration starts by investigating one attribute at a time using univariate charts.
• Techniques provide an understanding of the distribution and shape of data.

Histogram

• A basic visualization to understand the frequency distribution of values.
• Shows the distribution by plotting the frequency of occurrences.
• The horizontal axis represents the attribute of interest, and the vertical axis represents the frequency.

Class Stratified Histogram

• Histograms can be stratified to show the distribution of a specific variable across different groups.
• These provide an insightful view of the relationship between the variable and the groups/classes.
• Visually shows the distribution of attributes divided by classes.

Quantile Plots

• Box and whisker plots displaying the distribution of data for one variable, stratified by classes.

Multivariate Data Visualization

• Multivariate visual explorations consider more than one attribute in the same visual and examine relationships between attributes.
• These show the relationship between two to four attributes simultaneously.

Class-Stratified Quartile Plots

• Similar to box plots, but specifically displayed with stratification, showing the distribution of a variable across different classes/groups.

Distribution Plots

• Display the probability density of a variable, specifically useful for showing the distribution across different class types.

Scatter Plots

• Show the relationship between two variables; data points are displayed in Cartesian coordinates, enabling examination of the relationships between attributes.

Scatter Multiple Plots

• Similar to scatter plots but with multiple variables, presenting the relationships between multiple attributes, useful when dealing with a large amount of data, or multiple variables that influence the outcome variable/target variable

Multiple Scatter Matrices

• Matrices showing relationships among multiple attributes.

Bubble Plots

• Bubble plots show relationships between multiple variables using size to denote another variable.

Density Plots

• Dense plots of the underlying density for various attributes, demonstrating relationships between attributes.

Parallel Charts

• Parallel charts display multiple series along parallel lines, which display relationships between multiple attributes and various classes.

Deviation Charts

• Charts that show the distribution and deviations of multiple attributes over several classes

Andrews Curves

• Visually representing the relationships between several attributes in different classes using a mathematical equation.

Roadmap for Data Exploration

• A structured approach to explore and analyze a new dataset, with steps including organizing data, finding central points, understanding spread and distribution, outliers, relationship between attributes and high-dimensional visualizations.

Classification

• Classification, a type of data science, uses past records to predict and categorize upcoming events as class or numerical prediction.
• Uses information from predictors to categorize data into classes.
• Includes class prediction and numeric prediction.

Classification Algorithms

• Includes decision trees, rule induction, KNN, naive Bayesian, neural networks, and support vector machines.

Decision Trees

• A supervised learning algorithm suitable for classification and regression problems.
• Has a branching structure (tree) with potential answers to questions.

Decision Tree Metrics

• Information Gain (Entropy) is a key metric for training decision trees, measuring how much information is gained (or entropy reduced) by partitioning data, in a tree that is a data-mining method and algorithm.
• Gini impurity is another measure of a split's purity in decision trees. It measures inequality and helps split the data further.
• Misclassification error is the fraction of misclassified instances, used in decision trees to evaluate the quality of a split.

Decision Tree Implementation

• The steps to build a decision tree model are typically data organization, calculation of entropy, determination of information gain and building the tree for final analysis.
• The complete workflow involves training data, a learner (model), model evaluation, prediction, and outputting a complete analysis based on the output/results.

Decision Tree Implementation Considerations

• The use of splitting variables to create decision-making points or parts of the decision tree, or process.
• Recognizing when a decision tree is complete or stopping conditions for the decision tree, such as meeting an information gain threshold, reaching a maximum depth, or not having enough data in a split for proper analysis.
• Understanding how the values from a split affect the outcome variable.
• Evaluating the output or results of the decision tree model and examining its accuracy.

Additional Information for Note Taking

• Additional topics are available including a detailed discussion, and figures
• The documents provide a detailed introduction to data visualization methods. There is also content related to using these methods in classification processes.

Test your knowledge on data visualization concepts in the DS302 Fundamentals of Data Science course. This quiz will cover essential techniques and motivations behind effective data representation. Gain insights into how visual tools aid in data comprehension and analysis.