I have a situation where a bargraph is given with average CPU utilization data. The bar graph shows average CPU utilization per hour. The utilization is categorized to be normal, over utilized and under utilized (threshold for categorization are pre-specified). Now I am supposed to give a summary of this graph, like: the utilization on Monday's is high, utilization on Sunday's is below average etc. Kindly suggest how to approach this problem. Is this problem belongs to association rule mining. Are there few Machine Learning algorithms that could be used in such a scenario?
Hi Rajeev,
Why not make a simple decision tree on a daily basis, which writes its output to an ascii file. The output can be a text expression ike the one you used as examples in your problem description.
Advantage, a decision tree is fast, and easy to develop when you want to categorize daily measurements whether from CPU loadings or other variables.
Disadvantage: Its deadly simple ;-) Some people don't like that.
Cheers,
Frank
Here's a link to books that deal with this research area: http://books.google.com/books/about/Diagrammatic_Representation_and_Reasonin.html?id=lbZOcVtDcIEC
Hi Rajeev,
Why not make a simple decision tree on a daily basis, which writes its output to an ascii file. The output can be a text expression ike the one you used as examples in your problem description.
Advantage, a decision tree is fast, and easy to develop when you want to categorize daily measurements whether from CPU loadings or other variables.
Disadvantage: Its deadly simple ;-) Some people don't like that.
Cheers,
Frank
Thanks everyone for your comments. i'll try to incorporate these methods in my task :)
Hi Rajeev,
For generating text from the visual graph or any numeric data, there are well developed researches in NLG (Natural language generation) community which are responsible to textualise the summarise of data. You can easily use SimpleNLG java library to define the components of your messages (sentences) based on the numeric information of your graph. For sure, you have to first mine your raw data and organize the abstracted knowledge, then map them to linguistic vocabs. With NLG you can make Natural and human-like summary text.
Hope it helps you.
For more information:
http://en.wikipedia.org/wiki/Natural_language_generation
and these papers:
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3168319/pdf/amiajnl-2011-000193.pdf
https://www.researchgate.net/publication/257941854_A_framework_for_automatic_text_generation_of_trends_in_physiological_time_series_data
Conference Paper A Framework for Automatic Text Generation of Trends in Physi...
If i understood your question well, i think that this a direct and simple problem, and there is a simple and direct answer in image processing/analysis , let us make it simpler then you do the rest, suppose one image with one bar.
Input image with a bar of different color from the background
output: high , average, low
1- do color segmentation with a specific threshold that represent the color of the bar
2- convert the image to a binary on this bases, white color (1) is represent the bar, black (0) represent the background
3- find the minimum y of the white points call it miny
4- get the height of the image call it h
5- if miny h print "low"
else print "average"
if you have more than one bar, do step 1-2 mentioned above
2- use bucket-fill algorithm to group the white areas which represent each day
3- if you have 7 bars starting from sun to sat, do steps 3-5 above for each group (column )
do not forget to print the day with the result
enjoy it.
- Badges
- Science topic
- Similar topics
- Computer Science
- Data Mining