ACCT10003 Lecture Notes - Lecture 11: Predictive Analytics, Logistic Regression, Predictive Modelling
17 May 2018
School
Department
Course
Professor
Data Analytics and Decision Making
•Information ecosystem of business (and society) fundamentally transformed by technology
-Non-Financial data, Big data
•Potential to create value for business and society through predictive analytics
•Significant practical and ethical challenges
•How do we use this information in decision making?
•How can we improve our use of information outputs?
Types of predictive analytics
•Analytics = pairing data with predictive models to arrive at a conclusion
•Business knowledge + Data mining = Predictive analysis → Value
Basic analytics
•Slicing and dicing
-Breaking down data into smaller sets
-Descriptive statistics
-Data visualisation
•Basic monitoring
-Monitor large volumes of data in real time
-e.g. social media response to a new product launch
•Anomaly identification
-Actual observation differs from expectation
-e.g. higher rate of defects in one machine
Advanced Analytics
•Algorithms for complex analysis of either structured on unstructured data
•Predictive modelling
•Text & Voice analytics
•Other statistical and data mining algorithms
-Classification tress
-Logistic regression
-Neural networks
-Clustering techniques like K-nearest neighbours
Operationalised analytics
•Analytics become part of the business process
•e.g. insurance company builds a model that predicts the likelihood of fraudulent claims;!
incorporate in claims-processing system to ‘red-flag’ claims
•e.g. predict customers who are good targets for ‘upselling’;!
sales staff are alerted to specific additional products/services during interaction
Monetizing analytics
•Beyond businesses using analytics to increase revenues from their own operations and
datasets may be valuable to other companies
•e.g. customer data of credit card providers, financial institutions, telcos
•On-selling data
•What about the ethics of this?
1
Document Summary
Data analytics and decision making: information ecosystem of business (and society) fundamentally transformed by technology. Types of predictive analytics: analytics = pairing data with predictive models to arrive at a conclusion, business knowledge + data mining = predictive analysis value. Monitor large volumes of data in real time. E. g. social media response to a new product launch: anomaly identi cation. E. g. higher rate of defects in one machine. Advanced analytics: algorithms for complex analysis of either structured on unstructured data, predictive modelling, text & voice analytics, other statistical and data mining algorithms. Case studies of data analytics: social media. Use of twitter to predict election results. Cambridge analytica - company that uses data to micro-target individuals: customer relationship management. Analysing products with maximum demand, now and in the future. Predicting the buying habits of customers: supply chain management, collection analysis. For nancial institutions and other businesses, improved prediction of defaulting customers: credit risk analysis, operational management.
