I've rcently submitted a paper on Closed High Utility Freqent Itemsets Mining in static database. Now, I want to implement this idea on stream data by making a high ranked journal paper as a base. Although, I have done literature a lot but could not find the high-rank journal paper relevant to my work. Could anyone please support me in the following aspects:-
- recommend the high ranked journal paper relevant to my work
- suggest any idea which may be applicable to stream data mining
- or else any research guideline in the Stream Data Mining

Hi Muhammad,

Nice to receive your message. ESWA is a good journal. Wish your paper will have success there.

HUIM is a popular topic. There are several papers about this topic in journals likes Knowledge-based systems, Information Sciences, Applied Intelligence, etc. About stream mining, I participated to a recent paper on this topic for high utility itemset mining:

Duong, H., Ramampiaro, H., Norvag, K., Fournier-Viger, P., Dam, T.-L. (2018). High Utility Drift Detection in Quantitative Data Streams. Knowledge-Based Systems (KBS), Elsevier, 157 (1): 34-51.
DOI: 10.1016/j.knosys.2018.05.014

You coulde look at it. Also I think that Prof. Unil Yun from South Korea has a few papers related to stream mining for high utility itemsets.

Where to start? I think you should first read a bit about the algorithms for frequent itemset mining in streams. There are many papers about this. High utility mining is an extenstion of frequent itemset mining, so you can get a lot of ideas by looking at the papers from frequent itemsets mining from streams. Then you can think about the concepts in these papers and how to apply them to high utility itemset mining in stream.

But of course if you want to make a paper for a good journal, it is better that you also add some new idea to your paper. For example, you can redefine the problem of high utility itemset mining to add some new constraints and then design a new algorithm to deal for this new problem. That would be more interesting than just doing a faster algorithm.

Hope this helps

Thanks a lot for your valued and prompt respnose.
Also, The work on FIM over data stream has been started from near about 2005. Can you think we can get the noval ideas by reading them?

Yes, there is still a lot of thing can be done.

Between the frequent itemset mining and high utility mining there are also several differences, so some strategies for frequent itemset mining in stream may not work directly for high utility itemset mining and may need to be adapted. This can lead to some new contributions.

Besides, a good idea is also to make some new problems by combining some ideas with some other ideas. For example:

High utility episode mining + stream = high utility episode mining in a stream

or

Peak high utility itemset mining + stream = peak high utility itemset mining in a stream.

Then if you combine two topics like that, maybe that some new challenges are emerging because the problem is more complex than traditional frequent itemset mining in a stream, so you can improve the data structures, strategies etc. to solve the more difficult problem. In other words, the papers about frequent itemset mining from 2005 will give you some inspiration about how to solve the problem but you need to make some changes to make it work for HUIM.

There are many many possible research topics in pattern mining.... You need to think to find one that is challenging and that you can also convince people that it is useful in real life.



Edited 1 time(s). Last edit at 10/31/2020 10:12PM by webmasterphilfv.

Got your ideas. Thanks a lot again.



Edited 1 time(s). Last edit at 10/31/2020 10:30PM by i201606.