Please rotate your device.

Our website uses cookies to ensure you get the best experience while you’re here.


An Introduction to Machine Learning

Abstract structure

The intersection of technical innovation with an ever-increasing competitive business environment are driving organizations to explore new approaches to analytics. Business leaders are now pushed to demand data value chains to acquire raw data, integrate, provide context, and produce predictions and recommendations at the speed of business. To facilitate this effort, computers must not only deliver what is explicitly instructed, but must also leverage innovative techniques and algorithms to perform autonomously. In other words: Machine Learning.

Machine Learning has excelled the capabilities of computers including the ability to read, write, translate languages in real-time, as well as applying critical thinking. At SEI we’ve worked with clients to combine Machine Learning with traditional analytical capabilities to pursue new opportunities and insulate from competitive threats. Usually the first questions our clients ask is, “What exactly is meant by the term Machine Learning?Machine Learning can be described as a subset of Artificial Intelligence (AI) that teaches computers to “learn” without explicit programming rules by creating a flexible model for predictive analytics. Artificial Intelligence focuses on a broad category of problems with the goal of approximating reasoning and intelligence which are easy for humans to understand, but difficult for machines.

Through Machine Learning, computers can develop the ability to learn through experience and search through data sets to detect patterns and trends. Instead of extracting information for human comprehension and application, it will use results to adjust its own program actions, by creating algorithms to act on input data without explicit programming instructions for prediction or classification. It automates analytical models and adjusts the models with additional input data. Machine Learning can be supervised – the correct answers are given to the algorithms in advance – or unsupervised – no answers are provided in advance and the algorithm must group data into different classifications. The objective is for the algorithms to process data and provide analytic output independently and change based upon execution scenarios. In other words, “to learn”.

There are endless possibilities in which businesses can apply Machine Learning in order to achieve broadened capabilities while furthering the reach of potential consumers of their products and services.  A few examples include scenarios within the Health Care and Financial industries. First, opportunities in Health Care may include continuous monitoring of chronic diseases such as diabetes.  Machine Learning could be applied to incoming data points to not only generate alerts based upon set thresholds, but also to combine with other model features to predict disease advancement or help detect comorbid conditions.  It might also be applied to assist physicians in diagnosis, prescription, and potentially customize patient treatment options.

Furthermore, the Financial industry can leverage Machine Learning in mitigating fraud.  Financial Fraud is a major concern for both financial institutions and consumers alike.  Machine Learning can be added to traditional industry benchmarks and complex rule sets to add a dynamic capability to detect new behavioral anomalies that may be indicative of fraud.  Machine Learning makes Fraud detection more robust and adaptable to changing conditions.  Technological advancements in High Frequency Trading and potential Blockchain micro financial services necessitate automated fraud detection mechanisms utilizing Machine Learning techniques. Machine Learning can be applied across industries to combine both historical customer data and brand interaction touch points with general macro trends to optimize product recommendation and dynamic pricing algorithms.  Machine Learning combines multichannel historical information with dynamic algorithms that could detect either a promotional opportunity or generate a proactive alert for customer service prior to a negative brand interaction.

Now that a general understanding of Machine Learning has been relayed, the following series of subsequent blogs outlined below will focus on getting started using Machine Learning within your own organization:

  • The Approach: Developing Machine Learning Capabilities requires identifying the challenge or desired capability, assembling the right team, defining the process, and applying the technology.
  • The Process: Delivering value quickly and building on successes using the proven standard CRISP-DM for data mining.
  • The Technology: identifying current technology trends, competing options, and a building a Machine Learning proof-of-concept (POC).

In the meantime, check out this TED Talk, by technologist Jeremy Howard, that is sure to give more context into the world of Machine Learning, and increase your desire to know more!