How Deep is your Data?
Machine learning and deep learning
Machine learning is an umbrella term for techniques that enable computers to perform tasks without being explicitly programmed for them. This is particular useful if requirements are not known beforehand or the circumstances constantly change. In the last years, machine learning gained momentum in the context of big data and predictive analytics where machine learning algorithms are used to find prior unknown patterns in large data sets to gain business insights or build mathematical models to predict the future.
Deep learning on the other hand is a sub-group of machine learning-algorithms (mostly variations of neural networks ) that try to move machine learning closer to the idea of artificial intelligence. Deep learning is at the moment mainly used to process highly unstructured content like images, videos or audio. For example if you talk to your smartphone, you are actually talking to an automatic speech recognition algorithm based on deep learning techniques.
How machine and deep learning fit into Your Analytical Ecosystem
Figure 1. Machine and deep learning in an analytical ecosystem
Figure 1 illustrates how machine and deep learning fits into an analytical ecosystem. On the one hand there are classical business intelligence approaches like reporting and OLAP that are mainly used to process and visualize structured data from spreadsheets or (relational) databases, e.g. your data warehouse. These techniques usually deal with descriptive issues (what happened) or – if you use techniques like slice and dice or data mining – also diagnostic questions (why did something happen).
However, if content becomes more unstructured, classical data mining techniques (e.g. outlier analysis or regression) are often insufficient or not even applicable. This is where machine and deep learning come into action. With complex mathematical and statistical methods they are able to process highly unstructured content like text documents or even audio and image files. Moreover, machine and deep learning-methods deliver predictive models that can (statistically) predict future behavior based on past observations. This idea is often referred to as predictive analytics. If such systems go one step further and not only predict what will happen but also give advice what to do about it, you talk about prescriptive analytics. For instance, if a prediction model identifies a customer who is likely to resign his contract, it can also suggest an action (e.g. give him a discount) based on his past behavior.
Implications for Your Business
Machine and deep learning can tremendously increase the amount of relevant data for business analytics. Gartner estimates that around 80% of the data in an organization is unstructured content like text documents, phone calls and the like . In most companies, this data is completely unused or poorly transformed into structured content (e.g. with forms that need to be filled out manually). With machine and deep learning it is possible to automatically process all this data and enrich your business analytics with it. For instance, you can process all your customer emails and calls to conduct sentiment analyses to get insight about your customers, your products or market trends.
Moreover, machine and deep learning can help you offer new data-driven services for customers. For example, if you sell machines you can offer an additional predictive maintenance service, where you monitor a machine with sensors to automatically predict (based on machine learning models) when and where a problem is likely to appear. With this, you are able to send a service worker to fix the problem before it actually occurs and thereby minimize downtime. This is especially valuable for industrial manufacturing where every second downtime means real money.
In summary, it can be stated that machine and deep learning can increase the scope of your business intelligence and analytics by allowing you to process new types of data. In addition to that, machine and deep learning can also enable new data-driven business models. Many big companies already use machine and deep learning technologies, but as the technology becomes more mature, it will become more accessible to small and medium sized companies. The current emergence of many cloud-based machine learning platforms will probably accelerate this change. [3, 4, 5, 6]
 IBM’s Watson Analytics: http://www-03.ibm.com/software/products/en/watson-analytics
 Google’s Tensor Flow: https://www.tensorflow.org/
 Microsoft’s Cortana Intelligence Suite: https://www.microsoft.com/en-us/server-cloud/cortana-intelligence-suite/
 Amazon Machine Learning: https://aws.amazon.com/machine-learning/