Data Analysis Flashcards
Cluster analysis
Cluster analysis is an unsupervised machine learning technique that identifies inherent groupings within data. Unlike supervised learning where you have predefined categories, cluster analysis works on unlabeled data, automatically discovering structures. It achieves this by analyzing data points for similarities based on various features. Imagine a bunch of data points like customers at a store. Cluster analysis might group customers who buy similar products together, even without pre-existing labels like “high-value customer” or “budget buyer.” This grouping helps in tasks like customer segmentation in marketing or anomaly detection in fraud analysis. There are different algorithms for clustering, and the key is to choose the right one based on your data and the desired outcome. By uncovering these hidden structures, cluster analysis provides valuable insights for further analysis and decision making.
Cohort analysis
Cohort analysis is a subset of behavioral analytics that focuses on understanding how specific groups of users (cohorts) change and interact with a product or service over time. A cohort is defined by a shared characteristic, usually a common timeframe like the month they signed up. Instead of looking at overall trends, cohort analysis tracks the behavior of these groups longitudinally. This helps reveal retention patterns, how different user segments engage differently, identify points of drop-off in the customer journey, and measure the impact of changes within a product or marketing strategy. It offers a more granular understanding of user behavior than general analytics, helping businesses tailor their strategies to optimize engagement and customer lifetime value.
Factor analysis
Factor analysis is a statistical method used to describe variability among observed, correlated variables in terms of a potentially lower number of unobserved variables called factors. For example, it is possible that variations in six observed variables mainly reflect the variations in two unobserved variables. Factor analysis assumes that the correlations between your observed variables can be largely explained by their shared dependence on a smaller set of unobserved latent factors.
Factor analysis is like a detective searching for hidden patterns in a messy dataset. Imagine you have data on many variables (like student scores on different subjects, survey responses, etc.). Factor analysis tries to uncover a smaller number of underlying, unobserved variables (“factors”) that explain the relationships between the observed variables.
For example, if scores on math, physics, and chemistry exams are highly correlated, factor analysis might suggest a hidden factor like “quantitative ability” influencing them all. It’s a way to simplify complex data by identifying the core, potentially hidden, drivers that explain much of the variation you see.
Correlation Analysis: It starts by analyzing the correlation matrix of your observed variables.
Factor Extraction: Statistical techniques (e.g., principal component analysis) are used to extract factors that explain the maximum amount of variance in the data.
Factor Rotation: Often, extracted factors are rotated to improve interpretability, aiming to make each factor load highly on only a few of the original variables.
