2
$\begingroup$

Currently I need one data mining platform (or solution) meets such requirements:

Data Source

I can import data from: CSV, MySQL, PostgreSQL, HDFS.

Optional: NoSQL, MongoDB.

Interface

I can assemble models and datas with drag & drops (like Weka).

Algorithms

I can model with common machine learning algorithms: Logistic Regression, Random Forest, SVM, word2vec, etc.

It would be awesome if it had deep learning support.

Optional: PMML

Customization

I can implement my own algorithms (under certain standard).

Optional: I can implement my own APIs (under certain standard).

Programming language is not limited.

Visualization

Simple visualization is okay.

License

Commercial solution is acceptable, open-source solution would be better (I can do some tweaks in order to meet my requirement).

There are two software I've investigated: Weka and SPSS Modeler. Weka doesn't support HDFS, while SPSS Modeler doesn't support customized algorithms.

$\endgroup$
5
  • $\begingroup$ have you found anything that gets close to your requirements? Where have you looked and what data mining platforms have you already rejected? $\endgroup$
    – Spacedman
    Commented Feb 15, 2017 at 7:55
  • $\begingroup$ Thanks for your reply. Weka doesn't support HDFS and SPSS Modeler is hard to custom my own algorithms. $\endgroup$
    – Icyblade
    Commented Feb 15, 2017 at 8:56
  • $\begingroup$ Sounds like R can everything you listed. What's one thing R can't do? $\endgroup$
    – SmallChess
    Commented Feb 16, 2017 at 23:33
  • $\begingroup$ Orange, SAS Enterprise Guide and maybe alteryx would be some other options $\endgroup$
    – oW_
    Commented Feb 16, 2017 at 23:51
  • $\begingroup$ @StudentT As far as I know, I can't D&D in R. $\endgroup$
    – Icyblade
    Commented Feb 17, 2017 at 0:35

1 Answer 1

2
$\begingroup$

As @oW_ mentioned, Orange seems to be a good choice.

Data Source

Orange3-spark provides Hive/Spark support, cool!

Interface

A little laggy, but acceptable.

Algorithms

Full algorithm supports (based on scikit-learn). I can implement deep learning stuff myself.

Customization

Highly supported. Widgets & addons can be developed.

Visualization

Not bad.

$\endgroup$
1
  • $\begingroup$ We decided to develop a new one based on Orange, thanks for the help! $\endgroup$
    – Icyblade
    Commented Mar 12, 2017 at 10:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.