Data Mining Process and Iris Data Set
Chapter 1: Data Mining Process
- Data Sources
- Industrial Process Data: Captured at field and controller levels; also at operator and management levels.
- Business Data: Includes applications like shopping basket analysis and customer segmentation.
- Text Data: Derived from text documents, messages, and web documents.
- Image Data: Collected from sources like smartphone cameras and satellite imagery.
- Biomedical Data: Comprises genome data and lab results.
Definitions
- Data Mining (DM): The process of extracting meaningful knowledge from vast datasets.
- Knowledge: Defined as interesting patterns that are nontrivial, new, useful, or comprehensive.
- Knowledge Discovery (KDD): A structured process that includes:
- Preprocessing: Involves using a priori knowledge.
- Knowledge Extraction: The actual mining phase.
- Postprocessing: Evaluation of the mined knowledge.
- Data Analytics (DA): The application of computing systems to analyze large datasets for decision support.
- Intermediate Processes: DM, KDD, and DA are cyclical feedback processes requiring expert input.
- Related Areas: Integrates fields such as statistics, signal theory, pattern recognition, computational intelligence, machine learning, and operations research.
Knowledge Discovery Process
- Steps Include:
- Clustering: Grouping similar items together.
- Regression: Analyzing relationships among variables.
- Documentation: Keeping records of the process.
- Postprocessing: Final evaluations.
- Interpretation and Evaluation: Making sense of the results.
- Preparation, Planning: Organizing the approach to data collection.
- Feature Generation: Creating new variables from existing data.
- Data Selection: Choosing relevant data parts.
- Transformation: Altering data for analysis.
- Preprocessing: Data cleaning (filtering, cleaning) and ensures the data is prepared correctly.
- Analysis and Visualization: Studying data patterns visually to extract insights.
- Correlation & Forecasting: Understanding the interrelations between variables and predicting future trends.
- Standardization & Classification: Ensuring data is comparable and assigning categories.
Chapter 2: Data and Relations
- Example Topics:
- Scales: Understanding measurement scales in data analysis.
- Matrix Representation: Utilizing matrices to represent data hierarchically.
- Relations: Exploring connections between different data points.
- Distance Measures: Calculating dissimilarities between datasets.
- Clustering & Proximity Measures: Techniques to identify the closeness of data points.
- Sampling & Quantization: Methods for data collection and reducing information granularity.
Iris Data Set (Anderson 1935)
- Overview:
- Composed of n = 150 vectors with p = 4 dimensions concerning iris plants.
- Classes:
- Iris Setosa: 50 instances.
- Iris Versicolor: 50 instances.
- Iris Virginica: 50 instances.
- Components:
- Sepal Length
- Sepal Width
- Petal Length
- Petal Width
Typical Questions
- Which of the data might contain errors or false class assignments?
- What is the error caused by rounding the data off to one decimal place?
- What is the correlation between petal length and petal width?
- Which pair of dimensions is correlated most?
- None of the flowers in the data set has a sepal width of 1.8 centimeters. Which sepal length would we expect for a flower that did have 1.8 cm as its sepal width?
- Which species would an Iris with a sepal width of 1.8 centimeters belong to?
- Do the three species contain sub-species that can be identified from the data?