Data Science, Machine Learning and Predictive Analytics are currently hot topics in the analytics space. But what do these terms really mean? Is this the realm of PhD’s only? Or can all organisations benefit from advances in the field?
Data Science, Machine Learning and Predictive Analytics Defined
Data Science is a multidisciplinary blend of statistics, data analysis, algorithm development and technology, applied to extract knowledge or insight from data in various forms.
Within the field of Data Science, Machine Learning is one method by which these techniques can be applied. Using algorithms that iteratively learn from data, machine learning allows computers to find hidden patterns that are used to make a prediction, to segment data or to associate related information. The outputs of these models can be used to better inform business processes and decision making in a way not possible through simple human cognition. Predictive Analytics, Data Mining and Advanced Analytics are related terms often used as synonyms for Data Science and Machine Learning.
Machine Learning Surrounds Us Every Day
Machine Learning surrounds us in our everyday lives, perhaps without us realising it. Whether through Google’s search algorithm, when receiving a related product recommendation from our favourite online retailer, when receiving a personalised offer from our local Pizza chain, or when booking a hotel or flight. Machine Learning is being leveraged extensively by forward-thinking organisations to deliver a competitive edge, to optimise efficiency or to better engage their customers.
An IDC study found that the ROI of business analytics solutions that incorporate predictive analytics is about 250%, more than twice that of projects focused only on information access and productivity gains. A similar study performed by Nucleus Research found that 94% of customers had achieved a positive ROI, with an average payback period of 10.7 months.
Example Applications of Machine Learning
Machine Learning can be used to solve a variety of business problems. Examples include:
1. Yield Optimisation
- Demand Forecasting
- Price Elasticity / Sensitivity Modelling
- Dynamic Pricing
- Product Mix Optimisation
2. Customer Analytics
- Propensity to Respond
- Next Best Action
- Marketing Personalisation
- Customer Sentiment
- Propensity to Churn
3. Predictive Maintenance
- Predict Asset Failure
- Identify Operation Risk Factors
- Root Cause Identification
- Identify Poor Quality Parts
4. Threat & Fraud Detection
- Anomaly Detection
- Risk Scoring
Big Data Meets its Potential
The amount of data generated is growing exponentially each day, and will continue to grow. Whether from expanding operational data, telemetry from IoT (Internet of Things) devices or new types of data sources, these expanding data volumes present a huge opportunity for organisations to gain new insights. Machine Learning is the “secret sauce” promising to turn the hype of “Big Data” into reality. Organisations not leveraging their data are missing a huge opportunity to streamline and grow.
Toolsets used to build and deploy machine learning models are quickly evolving to the point that an average business or data analyst can build sophisticated models without the need for a PhD in Statistics. For example, tools such as SPSS Modeller and Microsoft Azure Machine Learning provide graphical user interface’s allowing users to implement machine learning models in a visual drag-and-drop interface, without the need to learn scripting languages such as R or Python. Data Visualisation tools are also quickly evolving, many now incorporating statistical algorithms and the ability to embed advanced analytic visualisations.
How to Get Started?
There are many potential applications for Machine Learning that leverage the typical data sets organisations already collect. It’s often possible to get started with core Sales, Finance, HR, Customer and Operations data, supplemented with contextual data that is easily obtained. With a little guidance, experimenting with a simple model over existing data sets can provide a low-risk way to get started with Machine Learning.