about NA values and featurebased complexity

[isilay123]

Hi,
Thank you very much for this useful work. I have two questions.

1. Is it not possible to calculate feature-based or other complexity measures for each data? Sometimes, I see this error:Error in match.arg(measures, ls.correlation(), TRUE) :
   'arg' should be one of “C1”, “C2”, “C3”, “C4”

2. I couldn't understand the NA output reason. For example, when I recognize the NA value in N3, can I say it doesn't have a standard deviation? It calculates both mean and standard deviation, but do these metrics have standard deviation values because they calculate more than one mean?

Kind regards,
Isill

[LesleyWheat]

Not the developer. For point 2, N3 should not return NA for sd as that would indicate the leave one out error is only one value (sd of a one item vector is NA in R). Have you checked that your input values are formatted correctly? Also, there is always only one mean, just multiple values for leave one out error. Ex. an error rate of 56 out of 495 points is mean 0.113 and sd of 0.317.

[isilay123]

Not the developer. For point 2, N3 should not return NA for sd as that would indicate the leave one out error is only one value (sd of a one item vector is NA in R). Have you checked that your input values are formatted correctly? Also, there is always only one mean, just multiple values for leave one out error. Ex. an error rate of 56 out of 495 points is mean 0.113 and sd of 0.317.

Thank you for your answer. I will check again, but I couldn't understand your example, for example, I calculated
featurebased.F1.mean = 2.775642e-01 , featurebased.F1.sd 2.612623e-01 but I couldn't understand.

[LesleyWheat]

Not the developer. For point 2, N3 should not return NA for sd as that would indicate the leave one out error is only one value (sd of a one item vector is NA in R). Have you checked that your input values are formatted correctly? Also, there is always only one mean, just multiple values for leave one out error. Ex. an error rate of 56 out of 495 points is mean 0.113 and sd of 0.317.

Thank you for your answer. I will check again, but I couldn't understand your example, for example, I calculated featurebased.F1.mean = 2.775642e-01 , featurebased.F1.sd 2.612623e-01 but I couldn't understand.

I'm not sure exactly what the sd value for F1 is, I work primarily with the neighbourhood measures. I was talking about N3, the leave one out error rate can be represented as a vector of classified correct/incorrectly as 0 and 1, from that the mean and sd can be taken. I don't personally use that sd, but it seems strange to be NA.

[isilay123]

Thank you for answer

[aclorena]

The original F1 measure takes only the maximum discriminant value among all features. For obtaining this result, you should choose max as a summarization function of the measure. The mean and std are output when considering the discriminant value of all features. *Sent with Right Inbox <https://www.rightinbox.com/?utm_source=signature>* Em qua., 29 de mar. de 2023 às 17:54, isilay123 ***@***.***> escreveu:

…

Thank you for answer — Reply to this email directly, view it on GitHub <#64 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AFIURWTOHGQD2X42GVTTXHLW6SORPANCNFSM6AAAAAAWKSTJYE> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

-- Prof Ana Carolina Lorena Divisão de Ciência da Computação (IEC) Instituto Tecnológico de Aeronáutica (ITA)

[isilay123]

The original F1 measure takes only the maximum discriminant value among all features. For obtaining this result, you should choose max as a summarization function of the measure. The mean and std are output when considering the discriminant value of all features. Sent with Right Inbox https://www.rightinbox.com/?utm_source=signature Em qua., 29 de mar. de 2023 às 17:54, isilay123 @.> escreveu:
…
Thank you for answer — Reply to this email directly, view it on GitHub <#64 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFIURWTOHGQD2X42GVTTXHLW6SORPANCNFSM6AAAAAAWKSTJYE . You are receiving this because you are subscribed to this thread.Message ID: @.>
-- Prof Ana Carolina Lorena Divisão de Ciência da Computação (IEC) Instituto Tecnológico de Aeronáutica (ITA)

Thank you very much .