I am trying to cluster gas wells based on their production behavior based on Gas/Water ratio vs cumulative water produced plots.
Would appreciate suggestions to solve this problem.
$\begingroup$
$\endgroup$
3
-
$\begingroup$ What do those plots look like? Are these scatter plots with gas/water on one axis and water produced on the other? $\endgroup$– KirkD_COCommented Nov 22, 2017 at 14:02
-
$\begingroup$ Hi ! I have attached a sample plot for reference. On X-axis we have cumulative water produced and on Y-axis we have Gas/Water ratio There are 200+ plots like these which i want to cluster $\endgroup$– Anoop Kumar MangarajCommented Nov 24, 2017 at 9:31
-
3$\begingroup$ What would the purpose of this clustering be? Since there are so many different valid ways in which such data could be clustered, you need to be more specific about what you're trying to achieve. $\endgroup$– whuber ♦Commented Nov 24, 2017 at 13:30
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
You need to come up with a metric to compare two graphs. If all your graphs had points with the same X-axis values, you could align them by the X-axes and calculate a Euclidean distance between Y-axis values, for instance. This would allow to compare all pairs for any typical clustering algorithm like K-means, hierarchical clustering, or something more interesting like self organizing maps.
I'm guessing that not all graphs share the same values on the X-axis, however. You could interpolate between points to accomplish that. The graph you show looks to be high enough resolution that interpolation between points wouldn't cause too much problem or introduce too much noise. This would give you all the points you need to align your graphs.
