Data mining libraries that easily integrate with Free Pascal?

Question

I have been experimenting with different approaches for pattern matching, analyzing, and predicting time-series data.

Since I have no professional experience in data mining or related fields I came up with my own methods from scratch.

After a few months of sporadically skimming data mining articles (the few that I understood) I realized that my methods are very basic.

For example, the way I implemented pattern matching is through a multidimensional k-Nearest Neighbor approach.

I am just now starting to understand that there are common techniques such as Decision Trees, Principal Component analysis, Covarience matricies, etc that I should be utilizing.

I have been experimenting with RapidMiner, which is a GUI data mining tool. RapidMiner lets you drag-and-drop various implementations of sophisticated techniques and connect them together without writing a line of code. It has been an awesome learning tool for me. It looks like this:

RapidMiner has an API but unfortunately it is in Java, and all of my current code is in Free Pascal and MySQL.

I am looking for some way to integrate RapidMiner into my application, so I could experiment with different data processing methods through the RapidMiner GUI and use them seamlessly with my current code.

If that is not feasible, I would settle for another data mining package, as long as it is easy to integrate into my current Free Pascal code.

I looked into R - and that seems about right, but it does not seem to be easy to integrate with Free Pascal either.

For clarity, my data sets are fairly large (500,000+ rows) and computation must occur in real-time. The software runs on Windows 7.

Answer 1

I don't know the package, but a few general options:

1. Try to get a plain C (native DLL) to this software from the authors.
2. Use JNI to communicate with a Java program that calls the API. Delphi and FPC JNI libraries are available.