BigML News, Issue #1,  January 2013

BigML - Machine learning made easy


Welcome <<Full Name>>,

There are many ways you can keep up to date with what is happening at BigML: read our blog, follow us on Twitter, Facebook or LinkedIn. And today we are adding to that list our monthly newsletter. It’ll keep you posted on the latest and hippest on BigML! This month:

Model of the month: Predictiong Kickstarter Success
Kickstarter success Predictive Model

Crowdfunding is an increasingly popular way of funding projects. Kickstarter is one of those crowdfunding organizations. But is there a way to predict a project’s success or failure? Kickstarter enthusiast Dan Misener has put together a site called The Kickback Machine that helps to organize the data from Kickstarter projects. We used that data to create this predictive model: Predicting Kickstarter Success!

A customer
Let's Date App.

Let’s Date is an app that helps you find people to date. The Let’s Date team took a different approach to matching. “We wanted to eliminate laborious surveys and questionnaires, fruitless repetitive messaging and endless search criteria and useless result lists. It was important for us to create something that requires less commitment, less effort and more results. You just casually browse people’s cards and let us know your reaction and our system learns how to deliver better cards to you.” Using BigML, the team has built a machine learning backend to learn from users likes and dislikes and increase the success of proposed dates. Check out Let’s Date for yourself.

Latest feature: Clojure bindings
BigML + Clojure

Just recently we published a Clojure library for our API in our constant effort to make machine learning simple for you. For more details and some examples, you can read more here.

Latest feature: Azure DataMarket browser
Windows Azure

There are many places where you can find great data. Windows Azure Marketplace DataMarket is one of those places. You can now browse datasets and get them into your BigML dashboard without leaving BigML or even without up- or downloading! Just click the ones that you want to work with and we’ll get it from the Windows Azure DataMarket to your dashboard. Check this blog post for more details.

Machine Learning explained: Training and Testing your model

What is the difference between Training and Testing a predictive model? Both activities start with datasets that contain the actual values of the feature you are trying to predict: your Objective. When training a model, a machine learning algorithm reads every instance of the training Dataset and infers the rules that predict the Objective. These rules combined make up your Model. For testing that Model, you take a test Dataset and for every instance in the Dataset, the Model will make a prediction. This is compared to the real value of the Objective that is in the test Dataset. This way we compute how ‘well’ your Model is doing. In BigML terms, training a model is ‘Create Model’ and testing a model is ‘Create Evaluation’. You can read more on Evaluations here.

This month questionaire

In every newsletter we’ll include a little questionnaire. A fun, interesting or informative set of questions. We will reward every completed questionnaire with free credits worth 10MB of predictive modeling. In this first one we'd like to know how to help you, so go ahead and take the questionnaire here:


Our mailing address is:
BigML, Inc
2851 NW 9th Street
Suite D, Conifer Plaza Building
Corvallis, Oregon 97330

Add us to your address book

Copyright © 2013 BigML, Inc, All rights reserved.