Network Mining Plugin: Some comments of a beginner user

 

Hello,

 

I am Visone user and I have begun to use “Network Mining” plugin as well. First of all, the connection (viz input and output) is working great with Visone!

 

I am learning the workflows that are available at http://tech.knime.org/node/36105. So far I could try the “Social Network Analysis” example. And I have realized how much the analysis can be improved using Knime features together with Visone. Congratulations for this plugin!

 

Follow some general comments:  

 

- As I come to Knime because of Visone, I am not familiarized about Knime interface and functioning yet. At beginning I had difficult to locate/download the example workflows which are stored in the “public KNIME server”. I would suggest to make available the following link, in which this step is more detailed (http://www.knime.org/example-workflows).

 

- Other point to beginners in Knime is to deal with “file reader” and “network reader” nodes. In the downloaded workflows, the data appears as stored in “C:\DATA\Demo\network\...”. I think this is the original source. I had to find out the data inside the workspace where I downloaded the workflows. Maybe it seems obvious, but a little explanation about it could be useful to beginners as well.

 

- Some workflows inputs demands other workflows outputs by using “network writer” node in the first workflow and “network reader” node in the second. For instance, this happens in “Network Creation Workflow (01202001_networkCreation)” which provides data to “Phase of Life Prediction (01202002_phaseOfLifePrediction)”. I had to change “network writer” node configuration. Specifically, I had to check in “memory policy” tab the option “write tables to disc”. Anyway, if I save or close the workflow the “.knet” table which is generated by “network writer” just disappear. In other words, the table “.knet” isn’t stored definitively. So, always I want to use “Phase of Life Prediction” I have to run the workflow “Network Creation Workflow” first. Is there any way to prevent the missing of this table?

 

 

All the best in this new year!

 

Cadu

Hello Cadu,
thanks for the useful comments. We have added the suggested link to the example homepage.
We have also updated the example flows to use internal "drop" directories for the reader nodes. When the dialog is opened, it should show a hint on the bottom of the dialog that the fileName parameter is controlled by a variable, which points to this internal directory. Therefore, you should be able to execute/reset and/re-execute the node without defining the location of the file to read. The previous workflows should use this feature as well but it seems to produce some errors like the problem you encounter with the "Phase of Life Prediction" workflow. The network nodes ignore the „memory policy“ and creates always a persistent file. The memory option applies only to nodes that create standard KNIME data tables (nodes that have the triangular out port).


Bye,
Tobias
 

 

Hi Tobias,

 

Some additional comments related to “Network Filtering (01202003_networkFiltering)”:

 

- The “feature value filter” is set to “No of participants” (which is a graph attribute). Considering the node description, the feature of this node seems to be “age”. The output of this node shows “No of nodes: 73 /  No of edges: 240”. But there was a filter; then, the No of edges would be lower. Using “viz output connector” and reading the network through Visone, the edges are 113, not 240 as Knime is showing. The same happens to other node filters (it seems to be related to networks in which there are alone nodes, with missing edges). After “edge degree filter” every output is matching with Visone.

 

Ps.: I have opened a thread (link bellow) in which I need to know if Knime Network Plugin is able to extract/obtain the link list format ("id id") from a “viz input connector” (attached to Visone). Do you know if it is possible and how to make?

 

http://tech.knime.org/forum/knime-users/asking-for-advice-about-knime-network-plugin-applied-to-the-analysis-of-a-network

 

Best,

Cadu