- Discover reason behind success and failures of schemes and projects - Understand customers and products better - Plan for future actions - Experiment meaningfully - Improve performance What is Analytics/Data Science Data wrangling (munging), retrieval and storage
Statistics
Big data
Data mining and
machine learning
- The application of scientific methods of reasoning to data driven decision
making - Includes hypotheses, experiments, facts, logical reasoning as well as data engineering -Focus: Machine Learning/Data Mining (Exploration of data, pattern finding and prediction/scoring) - Fancy definition The extensive use of data, statistical and quantitative analysis, exploratory, predictive models, and fact based decisions to drive decisions and actions
Why the need for advanced computational methods
- Small data can be easily analysed in MS Excel however once data grows beyond a certain size it cannot be analysed efficiently by Excel - Excel does not have a ready built function for clustering based on similar factors as well as factor finding. Reasons behind the rising popularity of analytics - Greater amounts of data available about consumers due to more data collecting technologies (wearables) - Greater capacity for data warehousing - Reduction in computing costs - Greater emphasis on customer relationship management with the advent of globalisation and increased competitiveness - More readily available commercial data analytics softwares - Change in organisational culture: product based to consumer based - Commercialisation of information and increased numbers of knowledge brokers Ways to Classify Analytics 1. Based on Method and Purpose a. Descriptive: Includes gathering, organising, tabulating and depicting the characteristics of what is being studied. Historically this kind of analytics was called reporting; describes but does not provide explanations or propose potential actions. E.g. Dashboards b. Predictive: Uses historical data to predict trends and identify phenomena and thus likely future outcomes c. Prescriptive: Analysing multiple options and stratgies (e.g. ideal product features using a cost-satisfaction matrix); includes methods such as experimental design, optimization, simulation and recommends a course of action 2. Based on Field of Application Marketing, Operational, Accounting, Customer, Social, Mobile, Fraud, etc Main Classes of algorithms in analytics Clustering: Discover natural groupings Classifiers: Prediction of which class a case belongs to Regression: Predict numerical outcomes Similarities/Recommenders: Explore associations and cooccurences Ensembles: Mixture of the above.