nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

Siemens Community Catalyst Program

The Siemens Community Catalyst program was co-created with our community to acknowledge technology leaders who consistently contribute to the Siemens Community. Nominations are accepted on a rolling basis.

Nominate Now

Should I Use a database or not?

siamak_want

Hi RM forum,

I am working on a relatively huge text data set (CSV file) with about 300,000 records and about 40,000 columns. I read the data with a "CSV Reader" and then process it via classification and clustering operator. The problem is that, as far as I know, the "CSV Reader" operator, first of all, loads the entire CSV file into main memory. The CSV file is several gigabytes and it uses more than a half of my main memory. So many algorithms starve from insufficient amount of memory.

As a workaround, I converted my data set to the fantastic "RM sparse format". I don't know how the "sparse reader" delivers the data but this could alleviate the problem of insufficient amount of main memory. Nevertheless the problem does still exist.

Now, in order to handle the problem of insufficient main memory, I have a new Idea about using a database (like postgre SQL):)! I think that a database can fetch the proper amount of examples into main memory and deliver the proper number of examples to other operators (i.e. classification operators). please explain if I am right about this. I need to know the advantages and disadvantages of using a database vs using simple files.

Any help would be appreciated.

Find more posts tagged with

AI Studio

Accepted answers

All comments

MariusHelf

I don't see any disadvantages of databases vs. text files, and you already stated the disadvantages of text files. So for any reasonable amount of data I would suggest to import the data into a database.

Btw, the sparse data format only works well if you have a lot of zeros in your data - otherwise it will use even more memory than the standard data format.

Best, Marius

siamak_want

Hi Marius,

Thanks to your straightforward answer (as always). So, for sure I will use a database as you stated. But there is an important issue about this I think:

A DBMS (for example PostgreSQL) can handle tables with less than 2000 columns. I am working on text data set. so I need to deal with about 80,000 features which does not fit in a regular data base table. Do you know any workaround?

Another thing that I am curious about is that, what should I do when the table is so big that it does not fit in main memory at all? Is the DBMS or RM able to load the data incrementally?

Thanks again Marius.

MariusHelf

If you have less examples than features you could transpose the dataset before storing it in the DB, and re-transpose it after loading.

For big tables, you can of course only read parts of the table from the DB and work on the subset (which is the usual way of dealing with large datasets - learn on a subset which fits into memory, then apply to large datasets incrementally).

For reading data automatically incrementally from the DB there is the Read Database (Incrementally) operator in RapidMiner, but I have to admit that I did not use it until now. Usually I rather load the subset of data I need manually with the normal Read Database operator.

Best, Marius

siamak_want

Hi the nice guy, Marius,

I think you mentioned a nice discussion...

learn on a subset which fits into memory, then apply to large datasets incrementally

would you please explain more? Do you mean I should use a sampling method to select some examples manually and then I should apply it on all of the data set? Then how about clustering?

Thanks in advance.

MariusHelf

siamak_want wrote:

would you please explain more? Do you mean I should use a sampling method to select some examples manually and then I should apply it on all of the data set? Then how about clustering?

Basically yes. if you choose a clustering algorithms which results in a cluster model (e.g. k-Means or any other centroid-based clustering), you can:
- create a representative sample of your data which fits into main memory
- create the cluster model and store it in the repository
- create a process which loads the model, iteratively reads parts of your data, applies the model and writes the result back to the database

Hope this helps,
Marius

siamak_want

Thanks to Marius

Quick Links

Upcoming Events

Advancing power electronics design with Simcenter PSIM and High-Performance Computing
Join us for a community conversation on how High-Performance Computing (HPC) and Simcenter PSIM work together to meet the increasing demands of modern power electronics simulation. During this one-hour technical session, you'll discover how leveraging HPC can speed up simulations, improve scalability, and empower engineers…
Getting Started with Faster Motor Building in Simcenter Simlab
Join us for a community conversation on accelerating your motor simulation workflow with Simcenter Simlab. In this beginner-friendly session, we'll introduce the platform's automated model-building capabilities and show how easy it is to set up your first electric motor analysis. You'll see the basic steps of geometry…
Practical hybrid solver workflows in Simcenter Feko
Join us for a Community Conversation where we'll discuss practical approaches to analyzing complex electromagnetic models using hybrid solver techniques. This session will provide a general overview of how different computational methods can be combined within a single simulation to address models containing varied…

View all community events