H20 training can leverage Spark's many data connectors, munging capabilities etc. POJO/MOJO + Spark is sufficient for scoring. Sparkling water is great fo rthis. Batching or nightly Review collected by and hosted on G2.com.

Only works well int he Spark Ecosystem. H20 training in other environments not as portable. Will never be good for REST API. Does not work in real-time as one would want. Review collected by and hosted on G2.com.

Automated library and minimum user effort required. The best part is the outstanding acuracy it provides in terms of the results. Also, plotting graphs is very easy. It handles the missing data well. Also takes care of the unballance in the data. Review collected by and hosted on G2.com.

It is a black box, in spite the outstanding accuracy. There should be a better documentation of the algoritham running in the background. Review collected by and hosted on G2.com.

H2o is becoming better and better day by day. The JVM is fast in analysis. The number of operations from data preprocessing to model building have been increasing and very effective. I love the GUI interface and the speed at which the H2o df help in doing the operations like slicing, dicing , data cleaning and model building on Spark Cluster. Sparkling is the best i know for various Big data problems. Review collected by and hosted on G2.com.

Not all the Machine learning algorithms are supported by H2o.ai. Algorithms like XGBoost and Neural networks needs to be added to the list of algorithms. Review collected by and hosted on G2.com.