You are here:  Home  >  Data Education  >  BI / Data Science News, Articles, & Education  >  BI / Data Science Articles  >  Current Article

Machine Learning Algorithms Today: Usage and Results

By   /  April 26, 2017  /  No Comments

machine learning algorithmsMachine Learning algorithms can predict patterns based on previous experiences. The overarching practice of Machine Learning includes both robotics (dealing with the real world) and the processing of data (the computer’s equivalent of thinking). These algorithms find predictable, repeatable patterns that can be applied to eCommerce, Data Management, and new technologies such as driverless cars. The full impact of Machine Learning is just starting to be felt, and may significantly alter the way products are created, and the way people earn a living. (For a primer in Machine Learning, see this article).

Machine Learning algorithms are trained with large amounts of data, allowing the “robot” to learn and anticipate problems and patterns. The Mars rover Curiosity uses a form of Machine Learning to traverse the Martian terrain, and there are plans to use the same algorithm for driverless cars. In the world of commerce “trend forecasting and analytics” rely on Machine Learning algorithms to anticipate shifts in purchasing behaviors, providing significantly better forecasts than had been done before the algorithm’s development.

Basic Algorithms

There are a variety of Machine Learning algorithms capable of assisting automated Data Modeling programs, and improving Data Management, eCommerce, and robotics. Listed below are some of the basic categories and related areas:

  • Deep Learning uses neural networks. Neural networks attempt to imitate how the human brain works. Interconnected “artificial” neurons are arranged in multiple processing layers (two is common with other Machine Learning systems). The additional processing layers provide higher-level abstractions, offering better classifications and more accurate predictions. Deep Learning is ideal for working with Big Data, voice recognition, and conversational skills.
  • Support Vector Machines (SVMs) are “supervised” learning models with appropriate learning algorithms. These algorithms analyze data, and are used for classification and for “regression analysis.” (Regression analysis uses statistics to estimate “the relationships among variables.”) It supports modeling and analyzing techniques using several variables, when the focus is on the relationship. More precisely, regression analysis helps in understanding how “criterion variables” change in value, when one of the independent variables change, while other independent variables remain fixed. SVMs are good at recognizing facial images and handwriting.
  • Probabilistic Models typically attempt to predict the best response by creating a model with a probability distribution. One of this model’s advantages is that it returns both the prediction, and the degree of certainty. A probabilistic model is meant to give a distribution of possible outcomes. It can describe all predicted outcomes and predict the probability of each. It is often used to provide “relevance” to search engine results.
  • Ensemble Learning algorithms act to combine the outputs from different Predictive Analytics models, and produce a “single” output. They are designed to help train “other” Machine Learning programs. Bootstrap Aggregating algorithms were the first effective Ensemble Learning algorithms. Bootstrap aggregating, also referred to as bagging, is an “Ensemble meta-algorithm” for Machine Learning, created to promote the accuracy and stability of programs used in regression and statistical classification. Although bagging is normally used with decision tree methods, it is adaptable to any type of Machine Learning method.

eCommerce Applications

Machine Learning algorithms are transforming major portions of the economy, altering how everything from online product marketing to customized search engines, and from self-driving cars to advanced medical imaging. The use of Machine Learning is broadening to include all aspects of eCommerce, the Internet, and technology. Consider the following:

  • Predicting “Market Right” Pricing:

Prior to price-oriented algorithms, online sellers competed with one another by slashing prices and minimizing profits. Now, eCommerce companies have fine-tuned the process, finding “just the right price” for bringing a customer and a particular product together. The retailer uses Machine Learning algorithms to ensure the optimum price for enticing the buyer, while also maximizing profits.

  • Improved Customer Segmentation:

Customer segmentation slots people into categories based on the patterns of their behavior. By identifying patterns among customers, and potential customers, e-commerce retailers can maximize their profits. Machine Learning algorithms can provide the information necessary for identifying new customers, and the opportunity to target specific customers with advertising.

  • Trend Forecasting and Analytics:

There is also an algorithm designed for analyzing and forecasting trends for online retailers. Prior to Machine Learning and Big Data, online retailers often experienced severe and chaotic shifts in fashion and trends. Purchased inventory would simply sit in storage, wasting investment capital and reducing profits. Currently, eCommerce merchants analyze and interpret as much data as possible, trying to anticipate the shifting tides of trends and fashion.

  • Fraud Detection and Prevention:

The amount of Internet payment fraud is increasing constantly, and, for the moment, seems to be unstoppable. Not too surprisingly, there has been a steady increase in e-commerce fraud each year, since 1993, with a 19% increase compared to 2013. For every $100 of product sold, fraudsters steal 5.65 cents.

The trend of increasing fraud against online retailers demonstrates the strong need for fraud detection and prevention. Machine Learning algorithms have great potential in the fight against criminals committing these crimes. Since detecting and identifying fraud requires a constant monitoring of online activity, Machine Learning is an ideal solution.

Retailers who understand and use Machine Learning to their advantage can substantially increase their profits. It has already become an important tool for large businesses, and as the technology becomes more affordable, online retailers of all sizes can take advantage of it. Machine Learning algorithms allow the correct product to be delivered at the correct time, at the correct price, and to the correct location.

Data Management Applications

Increasingly large amounts of data are collected, with the hopes of processing it into useful information. Indirectly, Machine Learning is providing great gains in the handling of Big Data. Machine Learning teaches Data Management programs how to handle these huge loads, in turn producing “Big Data.” As Data Science, and Data Management, continue to evolve, organizations can translate these huge amounts of information into useful questions and answers.

Online databases are expanding exponentially, and require a Data Management system to access the information quickly and efficiently. Machine Learning has helped significantly in the evolution of Data Management. The Cloud offers Machine Learning as an option in training Data Management programs. Although scalable Data Management is finally a reality, for more than three decades it was little more than a daydream. However, Cloud Computing comes with its own novel challenges, and these must be addressed to assure a successful Data Management system.

Master Data Management (MDM) is a big picture, overview system. Ideally today, it has access to a Cloud, and links all of an organization’s critical information to a single master file, which also acts as a common point of reference. A Master Data Management system should provide methods for collecting, consolidating, and distributing data within an organization. When setup properly, an MDM system streamlines the sharing of data with staff and managers.

Artificial Intelligence in Practice

Machine Learning can be used to teach robots real-world skills, and AI to make the best choice, through the use of algorithms and practice. These learning algorithms can include linguistic skills and sensorimotor skills, such as grasping, locomotion, and object categorization. Learning can take place by way of autonomous self-exploration or the guidance of a human instructor.

At the University of Washington, Maya Cakmak, has been working on a robot that can learn by imitation. This technique is called “programming by demonstration.” An instructor shows the robot a cleaning technique (within the robot’s vision) and the robot imitates the cleaning motion it watched the instructor perform. It is also trained to identify the “state of dirt,” both before and after the cleaning process.

According to a major report from South Australian industry analyst, Professor John Spoehr, robots and Artificial Intelligence will replace two of every five South Australian workers, over the next decade. Spoehr said,

“Our industry structure in South Australia makes us a bit more vulnerable to automation than other states, because of the types of occupations we have and dominance of mass manufacturing.”

He described how workers in the retail, transport, and hospitality services are most under threat, and added:

“Local accountants are being eclipsed by offshore provision of services, given the growth of the internet and online services. Even those types of sophisticated advisory roles are being rapidly automated.”

South Australia’s plight is due, in part, to the advances in Machine Learning algorithms along with allied technologies. Forbes Technology Council made similar predictions about jobs, on a broader scale. Certainly, the algorithms are here to stay and will continue to alter the Data Management and general workforce landscape far into the future


Photo Credit: Sergey Nivens/Shutterstock.com

About the author

Keith is a freelance researcher and writer.He has traveled extensively and is a military veteran. His background is physics, and business with an emphasis on Data Science. He gave up his car, preferring to bicycle and use public transport. Keith enjoys yoga, mini adventures, spirituality, and chocolate ice cream.

You might also like...

What’s the Difference Between a Data Warehouse and Data Lake?

Read More →
We use technologies such as cookies to understand how you use our site and to provide a better user experience. This includes personalizing content, using analytics and improving site operations. We may share your information about your use of our site with third parties in accordance with our Privacy Policy. You can change your cookie settings as described here at any time, but parts of our site may not function correctly without them. By continuing to use our site, you agree that we can save cookies on your device, unless you have disabled cookies.
I Accept