top of page

Strategic Partner

BigML

2018-12-12_14-01-05-816.png

Self Serve Machine Learning and Predictive Analytics

  Key Features

  • Easy to Get Started

  • Intuitive User Interface

  • Extensive Instruction Videos

  • Simple Source File Upload

  • 1-Click Dataset Builds

  • Real World Applications

  • Supervised Learning Models

  • Unsupervised Learning Models

  • Exportable Predictive Models

  • Collaborative Platform

  • REST API-Bindings/Libraries

  • Enterprise Security and Privacy

The company tagline for BigML is Machine Learning Made Easy. As an engaged user of BigML for the last five (5) years, DataEM can attest to this statement and we are proud to be a BigML Partner.

 

BigML's online tutorials will take you from 0 to 100 mph in an hour, using robustly-engineered Machine Learning algorithms, building supervised and unsupervised models, with predictions that solve real world problems.

It gets better from there because to get started, there is no software to install (cloud based) and up to 16 mg projects are FREE (no credit card needed).

If you are a Business Manager/Executive who is looking to excel your organization into data-driven thinking and improve KPIs, or the  Business Analyst who is looking to to go beyond basic business intelligence to predictive and prescriptive analytics, BigML is for you.

Technical Managers/Architects also use BigML to increase workforce productivity by offering a technically robust, scalable Machine Learning platform with traceable and repetable workflows. And Software Engineers will use BigML to build serverless smart applications with embedded models to enable real-time and local predictions.

 

If you get stuck building your machine learning project, BigML's customer service is extraordinarily responses, plus there is a robust knowledge base with detailed references and video tutorials. If you have a real challenging assignment, BigML has a team of data scientists ready to take your project on and make you a hero.

In addition to all of this, BigML is constantly adding new feature sets, bold enough to excite any experienced or budding data scientist.

 

The following is a "brief" overview of a few key features which BigML has to offer.  With this in mind, BigML has multiple layers of advanced features that are not illustrated here.

A Few Features

Easy to Get Started

2018-12-12_08-31-06-504.png
2018-12-12_08-38-11-978.png

BigML has no obstacles to getting started.  The sign up is Free and there are dozens of easy to follow, step by step, educational videos.

When you first enter the platform, BigML creates a customized experience based upon your Machine Learning proficiency level. From Absolute Beginners, to Novices familiar with the basic concepts of unsupervised learning, classification, overfitting, etc., to the Enthusiast who have used some Machine Learning packages like Weka, R, Scikit, etc., to the Practitioners that have built Machine Learning models in production, understanding the pros and cons of different techniques, to the Experts that have published papers at KDD, ICML, NIPS and devised their own algorithms.

Intuitive User Interface

2018-12-12_09-14-31-067.png
2018-12-12_09-21-02-608.png

In BigML, Sources are transformed into Datasets, which are the fundamental building blocks for BigML workflows. When BigML transforms a Source file into a Dataset.  it checks for missing files or errors in the data, provides summary statistics, and identifies fields that are "non-preferred". This is information that may not be useful for predictive analytics.

 

Datasets may be created automatically from Sources by using the 1-CLICK DATASET feature.  BigML datasets show field name, type, record count, missing data, errors and a distribution of data in a histogram.

In BigML, anything that you can create on the platform (datasets, models, scripts, etc.) are refereed to as Resources.

BIgML's intuitive user interface first displays the Sources of your data, has been uploaded. Uploading a new file can easily be added from your local machine or from external sources. This will be the first type of resource that you will create when using BigML.

2018-12-12_09-29-44-276.png
2018-12-12_09-48-15-528.png
2018-12-12_09-54-34-180.png

While BigML will use a complex set of algorithms to automatically build the datasets, almost everything is easily modifiable by the user, allowing processes to be overridden, however, the data itself cannot be changed. In that respect, the dataset is "immutable".  This is important when creating a dataset from a source file,  a sample from another dataset, or from a batch output.  BigML has a robust set of options for creating training samples from datasets, as well as a multi-functional dynamic scatter plot for examination.

Supervised Learning Models

2018-12-12_10-44-35-195.png
2018-12-12_12-23-53-923.png

With supervised learning, training data is loaded and one of the features in the data will be identified as the "outcome" or "objective" This is what you are trying to predict.  A supervised learning model is built by BigML and part of the data is used to test the accuracy of the models prediction.

 

BigML incorporates a number of algorithms that support Supervised Learning. This includes models / ensembles, logistic regression and time series.

An example of a BigML supervised learning model is the Decision Tree. The Decision Tree is flow chart style sequence of questions and possible answers about a data point, 

 

For example, a decision tree might be used to correlate the relationship between items purchased and the likelihood of an individual churning (not buying again). 

BigML will also generate Ensembles of Decision Trees, a technique called 'Bagging' or 'Random Decision Forest'. Here, a number of random data set are selected from the original data set, and a different tree is run on each of the samples. A prediction is made from each tree and are voted together to make a combined prediction. The outcome tends to be more reliable because the final prediction is not based on any one tree.

2018-12-12_12-44-25-206.png

More Supervised Learning Models

2018-12-12_13-02-13-689.png

BigML features include the ability to generate Supervised Learning Models from a dataset using 1-CLICK.

The 1-CLICK option provides the anticipated preferred fields with optimized settings.  As with other 1-CLICK selections, individual modifications may be applied and the models re-run.

Unsupervised Learning Models

2018-12-12_10-45-17-988.png

BigML supports unsupervised learning where the data provides examples, but no specific outcomes.  In this situation, BigML will try to find interesting patterns in the data. The goal is to provide discover rather than predict.

With this approach, BigML will look for Clusters of data that look similar, Anomalies that look different from the other data, Association Discovery where data looks alike with more than one condition.

Examples for applying unsupervised learning might include looking for customers that are similar (look-a-likes) which would be built from existing customer profiles.  Or, looking for unusual transactions, which would be built by modeling  historical transactions.

Another example, as illustrated below, looking for would be looking for products that are purchased together. This type of analysis would be trained on examples from previous purchase history.

2018-12-12_13-20-03-932.png
2018-12-12_13-19-49-754.png
2018-12-11_15-40-57-607.png

More Unsupervised Models

2018-12-12_13-02-22-869.png

BigML features include the ability to generate Supervised Learning Models from a dataset using 1-CLICK.

The 1-CLICK option provides the anticipated preferred fields with optimized settings.  As with other 1-CLICK selections, individual modifications may be applied and the models re-run.

BigML can facilitate unlimited predictive applications for agriculture, aerospace, automotive, electronics, energy, entertainment, financial services, food, healthcare, IOT, media, pharmaceutical, transport, telecommunications, and many more industries.  In addition to a friendly UX/UI, powerful and flexible models, BigML's Functionality Set Includes:

Interpretable /Exportable - Predictive Models on Local Offline or Real-Time Production Applications.

Automation - Bring Your Predictive Modeling Tasks To Production.

Collaborative Platform - Shared Workspace, Projects With Specific Roles And Permissions.

RESTful API - with Bindings & Libraries For Python, Node.Js, Ruby, Java, Swift, And More.

Flexible Deployments - on Virtual Private Cloud With Fully-Managed And Self-Managed Versions.

Security / Privacy - Granular Record Keeping And Transparency.

Security / Privacy - Assignable Access Privileges, Traceability And Reproducibility Workflows

Is BigML right for you? 

 

Contact DataEM or BigML today for a FREE assessment to help you determine if your organization will benefit from BigML.

2018-12-12_14-01-05-816.png
2018-12-11_11-59-01-416.png
bottom of page