Flashcards
Overview of Data Mining Tasks
Cards in this deck (16)
- 1DATA MINING TASKS- Concept Description
- Association Analysis
- Classification and Prediction
-Outlier Detection
-Evolution Analysis - 2DATA MINING CATEGORIES- Data View
-Knowledge View
-Method View
-Application View - 3Data ViewThe different kids of data that can be mined:
-Transactional Database
-Date Warehouse
-Temporal database
-Object Relational Database
-Multimedia Database
-Text Databases
-The World Wide Web - 4Knowledge ViewDifferent kinds of knowledge to be discovered from the dataset.
- 5Method ViewThe different techniques utilised such as data oriented, statistics, machine learning.
- 6ClassificationProcess of constructing a model that assigns each object to a pre-defined class. A model can be classified both descriptive and predictive.
A descriptive and predictive model distinguishes the objects of the different classes and predicts the class label of a unknown instance. - 7Concept DescriptionAll objects in a class have same characteristics. The description method summarises the target class and compares similar characteristics.
- 8Association AnalysisSome patterns occur often in a dataset. Frequent item-set refers to items which are frequently purchased such as bread or milk. Interesting associations within the dataset can be discovered.
- 9Classification Methods- Decision Tree Induction
- Naïve Bayesian classification
- Rule-based method
- Artificial neural network
- Nearest-neighbours
- Logistic regression - 10Decision TreeA decision tree is constructed based on given data to predict a specific outcome. Association rules can be used within a decision tree.
- 11Cluster AnalysisA cluster is a collection of data objects that are similar to one and other and within the same cluster. Dissimilar objects are in other clusters.
Cluster Analysis is carried out when the class label is unknown, so its excluded from the training data. Data is grouped together to form new categories i.e clusters. - 12Approaches of Clustering- K-means clustering
- Density-based approaches
-Model-based approaches - 13Characteristics of high quality clusters- High intra-class similarities
- Low inter-class similarities
The quality of clustering method is discovered by the ability to discover some or all hidden patterns. - 14OutliersOutliers are data objects that do not comply with the general behaviour or the model of the data. Most data mining algorithms discard outliers as noise or exceptions.
Outliers can be of interest to detecting credit card fraud, also Network intrusions and Medical analysis. - 15Interesting patternsAn interesting pattern represents knowledge to be mined in data. Interestingness can be measure subjectively and objectively.
Subjective measures are based on user's belief in data. e.g
-Novel
-Unexpected
-Easily understood
-Potentially useful and actionable. - 16Data Mining Task Primitives- Task-relevant data; suitable attributes.
- Kind of knowledge to be mined e.g classification or cluster analysis
-Background knowledge of the domain to be mined.
-Pattern of interestingness measures; ability to find something interesting.
-Visualization of discovered patterns; to effectively convey discovered knowledge to users e.g tables, bar charts, decision tree etc.
Information
- Description
- Bachelors Degree Data Mining Flashcards on Overview of Data Mining Tasks, created by Jason Dsilva on 28/04/2016.
- Created by
- Jason Dsilva
- Views
- 1
- Rating