Data Science

Course Code: Data1001

Duration: 3 days

Course Outline

Module 1 – Introduction to Data Science

• Origins of Data Science and a brief history of the Big Data revolution.
• The Big Data landscape.
• How much data is there really, and does it matter?
• Un-siloing data: use paradigms for organisational data and public data.
• Descriptive, predictive and prescriptive analysis.
• From recommendations to insights: black-box and white-box analytic.

Module 2– Data as an Asset

• The V’s of Big Data: Volume, Velocity, Variability, Veracity.
• Data business strategies.
• Data sources, synergies and differentiation.

Module 3– Data Life Cycle

• The analytic value chain.
• Overview of the data analysis cycle, connecting data science to the business problems.
• Work cycle of a data scientist: wrangling, modelling and validation.
• Managing research.

Module 4 – Privacy and Ethics in Big Data

• The societal impacts of Big Data.
• Privacy in Australia and global perspectives.
• Big Data ethics: history and current thinking.
• Opportunities and risks for organisations and individuals

Module 5 – Data Engineering for Analysis

• Data Science engineering and its drivers for change.
• Data volumes, data structures, and how they vary.
• Data Science architectures: the common stages.
• The Usual suspects: Distributed File Systems, Map Reduce, Spark

Module 6 – Visualisation

• Practical and effective visualization: beyond bar charts.
• Finding the unexpected: the role of visualization in exploratory analysis.
• Communicating findings: the role of visualization in communicating Data Science outputs.
• Standard tools: R, Tableau, D3.

Module 7 –Data Wrangling and Exploratory Analysis

• Determining data quality. Data cleansing.
• Entity matching.
• Imputation.
• Background modelling.
• Exploratory analysis.

Module 8 – Fundamentals of Statistics

• Types of data: numerical, categorical, ordinal.
• Statistical summaries: mean, standard deviation, quantities, correlation.
• Simple data visualization: histograms, boxplots, time plots and scatterplots.
• Cross-tabulations.
• Causality vs. association, independence.
• Randomisation and random sampling.
• Statistical inference using bootstrapping.

Module 9 – Model Creation and Validation

• Prediction: linear regression, nonparametric regression, k-NN.
• Forecasting: auto.arima and Error-Trend-Seasonal exponential smoothing algorithms.
• Hold-out sets, cross-validation, AIC.
• Classification: logistic regression, classification trees, SVM.
• Clustering: k-means, hierarchical clustering.
• Supervised vs. unsupervised vs. semi-supervised learning.
• Dimension reduction: principal components.
• Languages and environments (e.g. R, Python, MATLAB or even Excel) and standards (PMML).

Module 10 –Operationalisation and the Model Life Cycle

• Determining the needs: on how much data must decisions be taken, how often and how quickly must they be made, how often must models be refreshed?
• Plugging into existing data paths and choosing appropriate technologies.
• Stale models and  model refreshing.
• Operationalisation from a business perspective: determining value and making Data Science outputs part of standard business and decision-making processes.

Module 11 – Panel: Building a Data – Driven Enterprise

• Data Science as a process, rather than as a point event.
• The role of high-level management in enabling data-driven decisions.
• The role of direct management: on the un-Gantt-ability of research.

Module 12 – Case Study

• Operational efficiency by predictive analytics
• Architectural choices for integration and efficacy