Hello everyone,
It is interesting to mine patterns from image datasets in such a way the problem is how to represent an image as a transaction. Any proposed algorithm such as FP-growth can be used to mine the frequent patterns from the resulting dataset.
Does it exist a way to convert a given image into a transaction so that FIM methods can be used?
I appreciate your feedback.
Regards,
Saleh
Pattern Mining from image dataset
Re: Pattern Mining from image dataset
Hi,
Just curious, what is your goal of mining patterns in images? I mean, if you find itemsets in pictures, how would you then use these patterns? Or what would you like to discover? If you have a clear goal about why you want to mine patterns, then it could help to determine how to prepare your data. I mean that different goals, maybe would require to prepare the data in different ways. I am no expert on image, so that is just my opinion.
Just curious, what is your goal of mining patterns in images? I mean, if you find itemsets in pictures, how would you then use these patterns? Or what would you like to discover? If you have a clear goal about why you want to mine patterns, then it could help to determine how to prepare your data. I mean that different goals, maybe would require to prepare the data in different ways. I am no expert on image, so that is just my opinion.
-
- Posts: 5
- Joined: Thu Apr 07, 2022 1:09 pm
Re: Pattern Mining from image dataset
Thanks for your reply,
My knowledge of images is limited, but I am curious about the possibility of changing an image into a transaction.
Yes, you are correct; we are looking at how we can use an ARM to detect hidden knowledge from image data. This way, the results can be interpreted. As an example, consider an image dataset that contains ultrasound images of breast cancer. This dataset can be converted into a transactional dataset where each transaction represents the main features of a given image. So, after we convert the images, we can examine the rules generated from them. We may be able to detect cancer early if we know what symptoms lead to it.
Kind regards,
Saleh
My knowledge of images is limited, but I am curious about the possibility of changing an image into a transaction.
Yes, you are correct; we are looking at how we can use an ARM to detect hidden knowledge from image data. This way, the results can be interpreted. As an example, consider an image dataset that contains ultrasound images of breast cancer. This dataset can be converted into a transactional dataset where each transaction represents the main features of a given image. So, after we convert the images, we can examine the rules generated from them. We may be able to detect cancer early if we know what symptoms lead to it.
Kind regards,
Saleh