Motif and discord discovery

[rajvardhan21]

Hello,

In stemgen dataset,m=640 was selected on the basis of domain knowledge,but why 643 & 8724 was selected for motif ,how did you deduce that?

[rajvardhan21]

Hello,
In stemgen dataset,m=640 was selected on the basis of domain knowledge,but why 643 & 8724 was selected for motif ,how did you deduce that?

Also, How can we print matrix profile

[rajvardhan21]

Hi Sean,

thanks for your reply, can you help me with this error. this error shows up while finding out discord

[seanlaw]

@rajvardhan21 It looks like you are missing a closing parenthesis:

print(
    np.argsort(mp[:, 0])[-1]
)

Also, notice that the tutorial uses [-1] and not [:-1] (note that there is no colon).

[rajvardhan21]

Thanks sean,

1. can you help me visualize the mp array data in matrix form like you have shown in the presentation?
2. do you have any used case for finding the best value of m

[seanlaw]

1. can you help me visualize the mp array data in matrix form like you have shown in the presentation?

As I had mentioned above, it is not possible to output the distance matrix (only the matrix profile) in STUMPY. This is simply not feasible for longer time series as you would run out of computer memory.

2. do you have any used case for finding the best value of m

For finding m, this is unique/different for every time series and so there is no easy answer aside from trial and error. I recommend taking a look at this paper for ideas.

[rajvardhan21]

Thanks sean,

can we save mp array data and use mp.describe function?

also, can we plot next closest values (3-4 values) of mp on graph, if yes how?