I write a script that helps to verify structure and do basic descriptive statistics of a very large data.tables. I need to have a fast way to tell if a given variable has more than - say - 50 unique levels.
I might do it by
function(DT, colnr)
{
b<-DT[, list(var=colnames(DT)[[colnr]],.N), by = list(level=DT[[colnr]])]
if(nrow(b)<50)
#less than 50 levels, probably we need a dictionary.
setorder(b, level)
b
else
#more than 50 levels. Just return the number of unique values.
data.table(level=NA, var=colnames(DT)[[colnr]], N=-nrow(b))
}
but DT[, list(var=colnames(DT)[[colnr]],.N), by = list(level=DT[[colnr]])] takes very long for variables with as many unique values as there are cases (I have over 4 million of cases in DT)
Is there a way to improve the performance? I don't need to count the number of unique cases if there is more then 50 of them. Returning "more than 50" would suffice.
Related
I have a r code question that has kept me from completing several tasks for the last year, but I am relatively new to r. I am trying to loop over a list to create two variables with a specified correlation structure. I have been able to "cobble" this together with a "for" loop. To further complicate matters, I need to be able to put the correlation number into a data frame two times.
For my ultimate usage, I am concerned about speed, efficiency, and long-term effectiveness of my code.
library(mvtnorm)
n=100
d = NULL
col = c(0, .3, .5)
for (j in 1:length(col)){
X.corr = matrix(c(1, col[j], col[j], 1), nrow=2, ncol=2)
x=rmvnorm(n, mean=c(0,0), sigma=X.corr)
x1=x[,1]
x2=x[,2]
}
d = rbind(d, c(j))
Let me describe my code, so my logic is clear. This is part of a larger simulation. I am trying to draw 2 correlated variables from the mvtnorm function with 3 different correlation levels per pass using 100 observations [toy data to get the coding correct]. d is a empty data frame. The 3 correlation levels will occur in the following way pass 1 uses correlation 0 then create the variables, and yes other code will occur; pass 2 uses correlation .3 to create 2 new variables, and then other code will occur; pass 3 uses correlation .5 to create 2 new variables, and then other code will occur. Within my larger code, the for-loop gets the job done. The last line puts the number of the correlation into the data frame. I realize as presented here it will only put 1 number into this data frame, but when it is incorporated into my larger code it works as desired by putting 3 different numbers in a single column (1=0, 2=.3, and 3=.5). To reiterate, the for-loop gets the job done, but I believe there is a better way--perhaps something in the apply family. I do not know how to construct this and still access which correlation is being used. Would someone help me develop this little piece of code? Thank you.
I'm trying to clean up our customer database by identifying customer data that is similar enough to consider them the same customer (thus, give them the same customer id). I've concatenated relevant customerdata into one column named customerdata. I've found the R package stringdist and I'm using the following code to calculate the distance between every single record:
output <- df$id
for(i in 1:(length(df$customerdata)-1) ){
for(j in (i+1):length(df$customerdata)){
if(abs(df$customerdataLEN[i]-df$customerdataLEN[j]) < 10){
if( stringdist(df$customerdata[i],df$customerdata[j])<10){
output[j] <- df$id[i]
}
}
}
}
df$newcustomerid <- output
So here, I first initialize a vector named output with customerid data. Then, I loop through all customers. I have a column called customerdatalength. To reduce calculation time, I first check if there is large (>10) difference in length between columns. If that is the case, I don't bother calculating the stringdist. Otherwise, if the distance between the two customers is < 10, I consider them to be the same customer, and I give that customer the same id.
I'm looking to speed up the process however. At 2000 rows, this loop takes 2 minutes. At 7400 rows, this loop takes 32 minutes. I'm looking to run this on around 1 000 000 rows. Does anyone have any idea on how to improve the speed of this loop?
I have a big dataset (around 100k rows) with 2 columns referencing a device_id and a date and the rest of the columns being attributes (e.g. device_repaired, device_replaced).
I'm building a ML algorithm to predict when a device will have to be maintained. To do so, I want to calculate certain features (e.g. device_reparations_on_last_3days, device_replacements_on_last_5days).
I have a function that subsets my dataset and returns a calculation:
For the specified device,
That happened before the day in question,
As long as there's enough data (e.g. if I want last 3 days, but only 2 records exist this returns NA).
Here's a sample of the data and the function outlined above:
data = data.frame(device_id=c(rep(1,5),rep(2,10))
,day=c(1:5,1:10)
,device_repaired=sample(0:1,15,replace=TRUE)
,device_replaced=sample(0:1,15,replace=TRUE))
# Exaxmple: How many times the device 1 was repaired over the last 2 days before day 3
# => getCalculation(3,1,data,"device_repaired",2)
getCalculation <- function(fday,fdeviceid,fdata,fattribute,fpreviousdays){
# Subset dataset
df = subset(fdata,day<fday & day>(fday-fpreviousdays-1) & device_id==fdeviceid)
# Make sure there's enough data; if so, make calculation
if(nrow(df)<fpreviousdays){
calculation = NA
} else {
calculation = sum(df[,fattribute])
}
return(calculation)
}
My problem is that the amount of attributes available (e.g. device_repaired) and the features to calculate (e.g. device_reparations_on_last_3days) has grown exponentially and my script takes around 4 hours to execute, since I need to loop over each row and calculate all these features.
I'd like to vectorize this logic using some apply approach which would also allow me to parallelize its execution, but I don't know if/how it's possible to add these arguments to a lapply function.
Lets say I have this data. My objective is to extraxt combinations of sequences.
I have one constraint, the time between two events may not be more than 5, lets call this maxGap.
User <- c(rep(1,3)) # One users
Event <- c("C","B","C") # Say this is random events could be anything from LETTERS[1:4]
Time <- c(c(1,12,13)) # This is a timeline
df <- data.frame(User=User,
Event=Event,
Time=Time)
If want to use these sequences as binary explanatory variables for analysis.
Given this dataframe the result should be like this.
res.df <- data.frame(User=1,
C=1,
B=1,
CB=0,
BC=1,
CBC=0)
(CB) and (CBC) will be 0 since the maxGap > 5.
I was trying to write a function for this using many for-loops, but it becomes very complex if the sequence becomes larger and the different number of evets also becomes larger. And also if the number of different User grows to 100 000.
Is it possible of doing this in TraMineR with the help of seqeconstraint?
Here is how you would do that with TraMineR
df.seqe <- seqecreate(id=df$User, timestamp=df$Time, event=df$Event)
constr <- seqeconstraint(maxGap=5)
subseq <- seqefsub(df.seqe, minSupport=0, constraint=constr)
(presence <- seqeapplysub(subseq, method="presence"))
which gives
(B) (B)-(C) (C)
1-(C)-11-(B)-1-(C) 1 1 1
presence is a table with a column for each subsequence that occurs at least once in the data set. So, if you have several individuals (event sequences), the table will have one row per individual and the columns will be the binary variable you are looking for. (See also TraMineR: Can I get the complete sequence if I give an event sub sequence? )
However, be aware that TraMineR works fine only with subsequences of length up to about 4 or 5. We suggest to set maxK=3 or 4 in seqefsub. The number of individuals should not be a problem, nor should the number of different possible events (the alphabet) as long as you restrict the maximal subsequence length you are looking for.
Hope this helps
I am trying to find all of the unique groupings of a vector/list of items, length 39. Below is the code I have:
x <- c("Dominion","progress","scarolina","tampa","tva","TminKTYS",
"TmaxKTYS","TminKBNA","TmaxKBNA","TminKMEM","TmaxKMEM",
"TminKCRW","TmaxKCRW","TminKROA","TmaxKROA","TminKCLT",
"TmaxKCLT","TminKCHS","TmaxKCHS","TminKATL","TmaxKATL",
"TminKCMH","TmaxKCMH","TminKJAX","TmaxKJAX","TminKLTH",
"TmaxKLTH","TminKMCO","TmaxKMCO","TminKMIA","TmaxKMIA",
"TminKPTA","TmaxKTPA","TminKPNS","TmaxKPNS","TminKLEX",
"TmaxKLEX","TminKSDF","TmaxKSDF")
# Generate a list with the combinations
zz <- sapply(seq_along(x), function(y) combn(x,y))
# Filter out all the duplicates
sapply(zz, function(z) t(unique(t(z))))
However, the code causes my computer to run out of memory. Is there a better way to do this? I realize I have a large list. thanks.
To calculate all unique subsets, you are simply creating all binary vectors with the same length as the cardinality of the original set of items. If there are 39 items, then you are looking at all binary vectors of length 39. Each element of each vector identifies, yes or no, whether or not the item is in the corresponding subset.
As there are 39 items, and each can either be in or not-in a given subset, then there are 2^39 possible subsets. Excluding the empty set, i.e. the all-0 vector, you have 2^39 - 1 possible subsets.
That is, as #joran said, about 549B vectors. Given that the binary vectors are most compactly representing the data (i.e. without strings), then you will need 549B * 39 bits to return all of the subsets. I don't think you want to store this: that's about 2.68E12 bytes. If you insist on using the characters, you're likely to be in the many tens of terabytes.
It's certainly feasible to buy a system that can support this, but not very cost-effective.
At a meta-level, it is very likely, as #JD said, that this is not the path you really need to go. I recommend posting a new question and maybe it can be refined here or on the statistics-related SE site.
You might try using expand.grid.
Create a data frame from all combinations of the supplied vectors or
factors. See the description of the return value for precise details
of the way this is done.