Introduction to Analytics

This two-day course aims to provide an introduction to the world of analytics and how it is possible to analyse and visualise small and big data using the R open source statistical software.

The course combines theoretical aspects of descriptive and predictive analytics with a hands-on approach to R statistical software, using real life data from a range of different industries.

In the first day of the course we focus on data visualisation, transformation and exploratory analysis, while in the second day we explore patterns between multiple variables as well as prepare predictions using time series forecasting methods.

Content

The programme is organised across two days, with each day carrying four sessions. The content of each session is outlined below.

Day 1: Exploratory analysis

Introduction and managing data

The first session will start with a short introduction to the concept of analytics and aims/objectives of this programme. Then, we will introduce the audience to the main tools that will be used throughout the programme, R statistical software and RStudio interface, and we will use these for simple operations and demonstration of useful functions. We will explore some commonly used data structures and, finally, we will demonstrate how we can load data from external data files.

Data visualisation

R has several excellent systems for making graph, but in this workshop we will focus on one of the most commonly used, versatile and widely supported systems - ggplot2. ggplot2 implements the grammar of graphics - a coherent system for describing and building graphs. We will introduce the concepts of aesthetic mapping, faceting, and a range of basic geometrical objects that can be used with ggplot2 to represent data.

Data transformation

Visualisation is an important tool for insight generation, but it is rare that you get the data in exactly the right form you need. We will look at ways you can use R to filter, arrange, select and manipulate elements and variables of the data set you use, as well as the basics of using logical operators to transform your data.

Exploratory data analysis

In the final session of the day we will show you how to use transformation and visualisation to explore your data in a systematic way, a task that statisticians call exploratory data analysis (EDA). We will look at methods to obtain descriptive statistics and a range of ways to visualise variation and distribution, such as histograms, density plots, and boxplots.

Day 2: Exploratory analysis and prediction

Bivariate data and correlation

To start with explanatory analysis we will consider displays of bivariate data, which are instrumental in revealing relationships between variables. To do this we explore the application of visual methods of displaying data from day one, as well as scatterplots and quantile plots and methods to calculate correlation between two variables.

Simple regression

Quite often, we wish to model the effect of one variable to another. For example, we wish to measure the effect of advertising on sales or demand, so that we use any insights of such analysis to optimise advertising decisions. This session will demonstrate how we can model such relationships, through linear regression modelling, using R. At the end of the session, participants will be able to understand and explain the statistical output of regression and to use it for extrapolation and, subsequently, for decision making.

Multiple regression

Extending from the previous session, this session focuses on models where multiple independent variables are affecting (and can potentially be used for predicting) the dependent variable. Through an example on promotional modelling, we will explore the effects of different types of promotions on the sales of a product. We will demonstrate how it is possible to include additional variables, such as seasonal/holidays indicators and/or indicators for other special events. We will, also, show how we can decide on the inclusion/exclusion of important variables and build a model that have maximised predictive power.

Univariate predictive analytics

In many cases, the only data that are available are past observations of the variable under investigation, organised in the form of time series. This session focuses on univariate models for prediction of the future values of a sequence of observations (such as monthly demand of a particular product, hourly call centre arrivals every, or yearly revenues of a company) through capturing fundamental time series patterns, including trends and seasonal cycles. Simple (random walk and exponential smoothing) as well as more complex models (such as ARIMA) will be applied through the R statistical software in order to automatically produce predictions for the planning horizon.

Benefits

How will I benefit?

The participants of this course will benefit from a range of intellectual, practical and transferable skills, such as:

• acquire, clean, visualise, and analyse data
• identify correlations and patterns in data
• use software for statistical analysis and forecasting
• improved awareness of variability in data
• understanding of uncertainty in forecasts
• during this programme you will be using Analytics with R the preferred statistical software for business, however the principles learnt can be easily applied across other statistical software

How will my organisation benefit?

• Gaining understanding from visualising and analysing real world data
• Gaining business insights from quantitative data analysis
• Introduction to the benefits of open source software which could potentially be applied directly