Advanced Predictive Modeling Blossoms with 2014 BigML Spring Release

Janet Wagner, Contributing Writer
May. 14 2014, 01:18PM EDT

BigML, a leading machine learning and predictive modeling platform provider, has just announced the 2014 BigML Spring Release which includes new features and improvements including cluster analysis, additional filter options, segment-based dataset creation in models, new BigMLer command line tool features, and much more. The 2014 BigML Spring Release is a continuation of the company's efforts to boost predictive modeling which was reported by ProgrammableWeb earlier this year.

BigML now features segment-based dataset creation in models

Cluster analysis is a new feature that allows BigML users to create a cluster with one click from a dataset. Instances that are the most similar in a dataset can then be automatically grouped together into clusters. BigML will automatically apply scaling to all the numeric fields so that each field will have "roughly equivalent influence." Users can configure clusters modifying field scaling and weighting if needed. It should be noted that the platform auto-scale option is referred to as balance_fields in the BigML API. A cluster can be used to find the closest centroid for a new data point and the cluster analysis feature of the BigML platform is currently in beta.

BigML has added additional filtering options to the Flatline language provided with the platform. Datasets can now be filtered using functions such as comparison, equality, missing value, and statistics. Segment-based dataset creation in models has also been added to the platform which makes it possible for users to "create a new dataset for further analysis from a specific node in a tree."

New features have been added to the BigMLer command line tool including a new subcommand "bigmler analyze" which allows advanced predictive modeling by performing smart feature selection and node threshold selection. These new selections will be further explained in a future BigML blog post.

The 2014 spring release also includes many improvements to the BigML API which provides programmatic access to platform functions such as transformations, models, weights, predictions, and evaluations, which can be integrated into applications.

For more information about the BigML platform and API, visit

See also ProgrammableWeb's coverage of BigML's 1-million-predictive-models milestone and new BigML platform features.

Janet Wagner is a freelance writer and contributor to ProgrammableWeb covering breaking news, in-depth analysis, and product reviews. She writes well-researched, in-depth articles about machine learning, computer vision, GIS, maps, and other technologies. You can contact her on Twitter, Google+, or send her an email.