Your idea of machine learning may be different than you might expect.
No, sadly, Machine Learning doesn’t have anything to do with Skynet. It will not help you to have a meaningful conversation with your laptop. It’s probably not going to be responsible for a new gadget you absolutely need this holiday season.
Machine learning is a collection of tools used to try and answer questions we thought were unpredictable. Itis an inductive attempt to create a model (algorithm) based on a historic data set. Algorithms
In a traditional program, the programmer follows an algorithm to create some functionality. That algorithm can be expressed as a flowchart, equation or some other UML (Unified Modeling Language
) manor. Machine learning comes into play when that’s not possible. Getting a computer to identify and categorize what’s in a picture or predict what will happen in the stock market tomorrow is beyond traditional programming, but that hasn’t stopped computer and data scientists from trying.
Any good data scientist has an abundance of tools in their toolbox. For example, natural phenomenons such as coastal erosion or annual rainfall may have a distribution that aligns with a statistical distribution. Visualizing this data using linear regression may allow the scientist to predict a future event with a reasonable amount of certainty. Making a movie recommendation based on viewing history requires an entirely different approach, like a decision tree
. Data can be included/excluded, switched between algorithms, or manipulated in other ways to reveal underlying connections. Models
Models can build built for different goals. The goal for a self-driving car may be to find the safest path, a stock trader may be focused on getting the greatest return, and a company with a large fleet of vehicles may want to find the most fuel-efficient way to distribute resources. Each could require its own approach to building the model.
Training is crucial part of the process. A data set is often divided into sections. The model will be built with different parts of the older data and then that model is tested against a newer set to see how it performs. The data is massaged, and multiple models will be built and tested to find the best likely candidate to use in real time. From Data Mining to Machine Learning
Ten to fifteen years ago, as these technologies were emerging, they were commonly referred to as data mining. The term has largely been dropped, but I personally prefer this term to machine learning. Taking the deep dive into the data available can give a significant business edge, but doesn’t guarantee any profound insight. A data scientist’s hope for correlation or entropy may not be realized, even if it exists. Sitefinity and Machine Learning
In addition to other tools like personalization and testing, Sitefinity’s Digital Experience Cloud leverages machine learning to analyze user traffic and makes recommendations to better engage users with your site. Vanguard can leverage Sitefinity's Digital Experience Cloud and its machine learning to provide you and your users with more value. If you have business records and think there is something more to learn from them. A better way to service your members and users, let us know.