Topics: AI | Data Science | Data Visualization | Deep Learning | Machine Learning | NLP | Python | R | Statistics

KDnuggets Home » News :: 2013 :: Sep :: News, Software :: BigML Machine Learning Made Easy, see it in action ( 13:n23 )

BigML Machine Learning Made Easy, see it in action

I talked to BigML to learn about their "special sauce" and exciting advanced features such as decision forest ensembles, "Sunburst" visualization for decision trees, BigML PredictServer on AWS, text processing, and Multi-label prediction. See BigML it in action on Sep 25, and get a special KDnuggets discount.

Get 25% off BigML subscription with coupon code KDN25
See BigML in action on a September 25 webinar
BigML is passionate about democratizing Machine Learning, and provides data practitioners easy access to a hosted machine learning platform for predictive analytics and a variety of machine learning tasks. Within minutes just about anyone can upload a data source and begin creating actionable predictive models and ensembles. Traditionally this type of functionality would require significant investment in workers and IT infrastructure. With BigML, however, data analysts and data scientists alike can get started for as low as $30/month for unlimited modeling on datasets up to 64MB - with packages also available for larger sizes.
KDnuggetsKDnuggets community members can purchase a subscription for an
added 25% off, by using coupon code KDN25
Simply register here for free, and then choose a subscription plan from within your new account.

Interview with Francisco J. Martin

KDnuggets talked with BigML's co-Founder and CEO Francisco J. Martin to learn more about the company's background and vision. You can also attend a September 25 webinar from BigML showing how to use BigML's latest functionality to predict customer churn ( register here).

Q: Why did you create BigML?

A: I previously started two other Artificial Intelligence related companies. iSOCO focused on intelligent software components and Strands on recommender systems. Inspired by the Data Deluge issue from The Economist in 2010 I realized that there would be great demand for machine learning and predictive analytics by a broader audience who would want to bring the power of ML to make better decisions, build smarter products and predictive applications. BigML was founded in early 2011 with the objective of bringing the power of machine learning to just about anyone.

Q: What is BigML's "special sauce"?

A: There's no single thing; rather I think it's the combination of powerful machine learning algorithms mixed with an intuitive workflow, running on top a very sophisticated auto-scalable platform all packaged in a nicely designed product. This makes BigML special to new and expert users alike as it hides most complexities faced in data analysis.

Whereas marketing analysts can use our interface to discover new information and predictive relationships on their customer data, data scientists can leverage our RESTful API to quickly build advanced models and ensembles that can be downloaded for local usage and/or for incorporation into their own applications and services. One thing that stands out for BigML is the fact that with a single click you can build random decision forest ensembles based on 10s or 100s of models. This, of course, is the same methodology that people are using to consistently perform well in Kaggle competitions.

Q: What are some new features that KDnuggets readers should be excited about?

A: We're constantly rolling new innovations into our platform and interface. Earlier this summer we introduced a "Sunburst" visualization for decision trees - this has been a very popular feature for our users. Coming soon is the BigML PredictServer - a high-performance AWS server that will allow people to perform thousands of predictions per second that can be programmatically incorporated into their own systems and applications.

We're also releasing support for text processing which will empower users to create predictive models with a mix of categorical, numeric and text fields. Two more cool features are Excel export, which will enable users to get a predictive model directly exported into a spreadsheet and Multi-label Prediction support in our command line, which will allow users will to deal with problems when instances belong to more than one class (e.g., identify keywords and tags of blog posts).

We'll be showing all of these new features on our webinar later this month.

Q: What does the future hold for machine learning?

A: BigML believes that machine learning will become as pervasive in the business world as spreadsheets - moving from being the exclusive remit of technologists, to enabling virtually anyone in an organization to begin making data-driven business decisions, and empowering developers to add the power of predictiveness to their applications and services.

Sign Up

By subscribing you accept KDnuggets Privacy Policy