Data Mining and Knowledge Discovery in Large Databases

Data Mining and Knowledge Discovery in Large Databases

• 1.
  Outline Data Mining and „We are drowning in data, but we are starving for knowledge“ Knowledge Discovery Part 2: Clustering in Large Databases - Hierarchical Clustering - Divisive Clustering - Density based Clustering Erik Kropat University of the Bundeswehr Munich, Germany
• 2.
  Why “Data Mining”? •Companies are collecting massive amounts of data on customers, operations, and the competitive landscape. Firms can gain a competitive advantage from these data • But, there is far too much data − Online shops record purchase behaviours for millions of customers (sometimes with hundreds features for each customer) − Phone companies keep info on 100’s of millions of accounts (each with thousands of transactions) − Databases can often be hundreds of terabytes in size (this will be peanuts in the future).
• 3.
  Why “Data Mining”? „We are drowning in data, but we are starving for knowledge“ (John Naisbitt)
• 4.
  Knowledge Discovery inLarge Databases Process of finding valuable and useful patterns in datasets
• 5.
  Analysis of datasets from … • businesses & investments • finance & economics • science & technology • bioinformatics • telecommunication … or more complex data sets • multimedia & sound • images & video • automatic news analysis • social media analysis.
• 6.
  What are thedata sources? Consumer data − Credit card transactions data − Supermarket transactions data − Loyalty cards − Web server logs − Social media Variety of features − Name and address − History of shopping and purchases − Demographics − Credit rating − Quality & market share of products
• 7.
  Business Intelligence ‒ CustomerData Analytics & Market Analysis − customer segmentation − market basket analysis − target marketing − geo-marketing − cross-selling / up-selling − customer relation management
• 8.
  Market Basket Analysis‒ Cross Selling
• 9.
  Key Tasks Decision Trees Assocation Rule Learning Neural Networks Digital Forensics Automatic Derivation of Ontologies
• 10.
  Retail • Customer segmentation Identify purchase patterns of „typical“ customers Targeted advertisement, costumized pricing, cost-effective promotions • Market basket analysis Identify the purchase behaviour of groups of customers • Sales promotions Identify likely responders to sales promotions
• 11.
  Banking • Credit rating Given a large number names, which persons are likely to default on their credit cards? • Fraud detection − Credit card fraud detection − Network intrusion detection
• 12.
  Telecommunications Companies are facingan escalating competition and are forced to aggressively market special pricing programs aimed at retaining existing customers and attracting new ones. • Call detail record analysis Identify customer segments with similar use patterns. Offer attractive pricing and feature promotions. • Customer loyalty / customer churn management Some customers repeatedly „churn“ (switch providers). Identify those who are likely to switch or who are likely to remain loyal. Companies can target their spending on customers who will produce the most profit. • Set pricing strategies in a highly competitive market.
• 13.
  Big Data isBig Business Companies are using their data sets to aim their services and products with increasing precision. Business Intelligence − SAP AG is a German global software corporation that provides enterprise software applications. − SAP AG is one of the largest enterprise software companies. − In October 2007, SAP AG announced a $6.8 billion deal to acquire „Business Objects“. − Since 2009 „Business Objects“ is a division of SAP AG instead of a separate company.
• 14.
  Outline
• 15.
  Outline Part1: Introduction Part 4: Classification - What is „Data Mining“ ? - k-th Nearest Neighbors - Examples - Support Vector Machines Part 2: Formal Concept Analysis Part 5: Spatial Data Mining - Contexts and Concepts - DBSCAN - Concept Lattices - Density & Connectivity Part 3: Clustering Part 6: Regulatory Networks - Hierarchical Clustering - Eco-Finance Networks - Partitional Clustering - Gene-Environment Networks - Fuzzy Clustering - Graph Based Clustering
• 16.
  Questions ? For more information after today Email me at Erik.Kropat@unibw.de