I'm a bit of an R novice and have been trying to experiment a bit using the agrep function in R. I have a large data base of customers (1.5 million rows) of which I'm sure there are many duplicates. Many of the duplicates though are not revealed using the table() to get the frequency of repeated exact names. Just eyeballing some of the rows, I have noticed many duplicates that are "unique" because there was a minor miss-key in the spelling of the name.
So far, to find all of the duplicates in my data set, I have been using agrep() to accomplish the fuzzy name matching. I have been playing around with the max.distance argument in agrep() to return different approximate matches. I think I have found a happy medium between returning false positives and missing out on true matches. As the agrep() is limited to matching a single pattern at a time, I was able to find an entry on stack overflow to help me write a sapply code that would allow me to match the data set against numerous patterns. Here is the code I am using to loop over numerous patterns as it combs through my data sets for "duplicates".
dups4<-data.frame(unlist(sapply(unique$name,agrep,value=T,max.distance=.154,vf$name)))
unique$name= this is the unique index I developed that has all of the "patterns" I wish to hunt for in my data set.
vf$name= is the column in my data frame that contains all of my customer names.
This coding works well on a small scale of a sample of 600 or so customers and the agrep works fine. My problem is when I attempt to use a unique index of 250K+ names and agrep it against my 1.5 million customers. As I type out this question, the code is still running in R and has not yet stopped (we are going on 20 minutes at this point).
Does anyone have any suggestions to speed this up or improve the code that I have used? I have not yet tried anything out of the plyr package. Perhaps this might be faster... I am a little unfamiliar though with using the ddply or llply functions.
Any suggestions would be greatly appreciated.
I'm so sorry, I missed this last request to post a solution. Here is how I solved my agrep, multiple pattern problem, and then sped things up using parallel processing.
What I am essentially doing is taking a a whole vector of character strings and then fuzzy matching them against themselves to find out if there are any fuzzy matched duplicate records in the vector.
Here I create clusters (twenty of them) that I wish to use in a parallel process created by parSapply
cl<-makeCluster(20)
So let's start with the innermost nesting of the code parSapply. This is what allows me to run the agrep() in a paralleled process. The first argument is "cl", which is the number of clusters I have specified to parallel process across ,as specified above.
The 2nd argument is the specific vector of patterns I wish to match against. The third argument is the actual function I wish to use to do the matching (in this case agrep). The next subsequent arguments are all arguments related to the agrep() that I am using. I have specified that I want the actual character strings returned (not the position of the strings) using value=T. I have also specified my max.distance I am willing to accept in a fuzzy match... in this case a cost of 2. The last argument is the full list of patterns I wish to be matched against the first list of patterns (argument 2). As it so happens, I am looking to identify duplicates, hence I match the vector against itself. The final output is a list, so I use unlist() and then data frame it to basically get a table of matches. From there, I can easily run a frequency table of the table I just created to find out, what fuzzy matched character strings have a frequency greater than 1, ultimately telling me that such a pattern match against itself and one other pattern in the vector.
truedupevf<-data.frame(unlist(parSapply(cl,
s4dupe$fuzzydob,agrep,value=T,
max.distance=2,s4dupe$fuzzydob)))
I hope this helps.
Related
A function in a package gives me a character, where the original strings are merged together. I need to separate them, in other words I have to find the original elements. Here is an example and what I have tried:
orig<-c("answer1","answer2","answer3")
result<-"answer3answer2"
What I need as an outcome is:
c("answer2","answer3")
I have tried to split() result, but there is no string to base it on, especially that I have no former knowledge of what the answers will be.
I have tried to match() the result to the orig, but I would need to do that with all substrings.
There has to be an easy solution, but I haven't found it.
index <- gregexpr(paste(orig,collapse='|'),result)[[1]]
starts <- as.numeric(index)
stops <- starts + attributes(index)$match.length - 1 )
substring(result, starts, stops)
This should work for well defined and reversible inputs. Alternatively, is it possible to append some strings to the input of the function, such that it can be easily separated afterwards?
What you're describing seems to be exactly string matching, and for your strings, grepl seems to be just the thing, in particular:
FindSubstrings <- function(orig, result){
orig[sapply(orig, grepl, result)]
}
In more detail: grepl takes a pattern argument and looks whether it occurs in your string (result in our case), and returns a TRUE/FALSE value. We subset the original values by the logical vector - does the value occur in the string?
Possible improvements:
fixed=TRUE may be a bright idea, because you don't need the full regex power for simple strings matching
some match patterns may contain others, for example, "answer10" contains "answer1"
stringi may be faster for such tasks (just rumors floating around, haven't rigorously tested), so you may want to look into it if you do this a lot.
I have one dataframe with less than 5000 rows (csv file). I have plenty of columns, one of them is the company name.
However, there are many duplicates with different names, for example, one company can be called: HH 785 EN
And his duplicate could be called : HH 785EN or HH784 EN
Every duplicates have like 1 or 2 differents characters from the original company.
I'm looking for an algorithm that could potentially detect these duplicates.
Most of the fuzzy match problems I have seen have 2 datasets involved which isn't my case.
I have seen many algorithm which takes one word and a list as entry, but I want to check my whole column of companies names with itself.
Thanks for your help.
I think you are looking for agrep function that does Levenshtein distance. You can combine agrep with sapply to find the fuzzy match.
sapply(df$company_name,agrep,df$company_name)
I'm attempting to clean up a database that, over the years, had acquired many duplicate records, with slightly different names. For example, in the companies table, there are names like "Some Company Limited" and "SOME COMPANY LTD!".
My plan was to export the offending tables into R, convert names to lower case, replace common synonyms (like "limited" -> "ltd"), strip out non-alphabetic characters and then use agrep to see what looks similar.
My first problem is that agrep only accepts a single pattern to match, and looping over every company name to match against the others is slow. (Some tables to be cleaned will have tens, possibly hundreds of thousands of names to check.)
I've very briefly looked at the tm package (JSS article), and it seems very powerful but geared towards analysing big chunks of text, rather than just names.
I have a few related questions:
Is the tm package appropriate for this sort of task?
Is there a faster alternative to agrep? (Said function uses the
Levenshtein edit distance which is anecdotally slow.)
Are there other suitable tools in R, apart from agrep and tm?
Should I even be doing this in R, or should this sort of thing be
done directly in the database? (It's an Access database, so I'd
rather avoid touching it if possible.)
If you're just doing small batches that are relatively well-formed, then the compare.linkage() or compare.dedup() functions in the RecordLinkage package should be a great starting point. But if you have big batches, then you might have to do some more tinkering.
I use the functions jarowinkler(), levenshteinSim(), and soundex() in RecordLinkage to write my own function that use my own weighting scheme (also, as it is, you can't use soundex() for big data sets with RecordLinkage).
If I have two lists of names that I want to match ("record link"), then I typically convert both to lower case and remove all punctuation. To take care of "Limited" versus "LTD" I typically create another vector of the first word from each list, which allows extra weighting on the first word. If I think that one list may contain acronyms (maybe ATT or IBM) then I'll acronym-ize the other list. For each list I end up with a data frame of strings that I would like to compare that I write as separate tables in a MySQL database.
So that I don't end up with too many candidates, I LEFT OUTER JOIN these two tables on something that has to match between the two lists (maybe that's the first three letters in each list or the first three letters and the first three letters in the acronym). Then I calculate match scores using the above functions.
You still have to do a lot of manual inspection, but you can sort on the score to quickly rule out non-matches.
Maybe google refine could help. It looks maybe more fitted if you have lots of exceptions and you don't know them all yet.
What you're doing is called record linkage, and it's been a huge field of research over many decades already. Luckily for you, there's a whole bunch of tools out there that are ready-made for this sort of thing. Basically, you can point them at your database, set up some cleaning and comparators (like Levenshtein or Jaro-Winkler or ...), and they'll go off and do the job for you.
These tools generally have features in place to solve the performance issues, so that even though Levenshtein is slow they can run fast because most record pairs never get compared at all.
The Wikipedia link above has links to a number of record linkage tools you can use. I've personally written one called Duke in Java, which I've used successfully for exactly this. If you want something big and expensive you can buy a Master Data Management tool.
In your case probably something like edit-distance calculation would work, but if you need to find near duplicates in larger text based documents, you can try
http://www.softcorporation.com/products/neardup/
I'm attempting to clean up a database that, over the years, had acquired many duplicate records, with slightly different names. For example, in the companies table, there are names like "Some Company Limited" and "SOME COMPANY LTD!".
My plan was to export the offending tables into R, convert names to lower case, replace common synonyms (like "limited" -> "ltd"), strip out non-alphabetic characters and then use agrep to see what looks similar.
My first problem is that agrep only accepts a single pattern to match, and looping over every company name to match against the others is slow. (Some tables to be cleaned will have tens, possibly hundreds of thousands of names to check.)
I've very briefly looked at the tm package (JSS article), and it seems very powerful but geared towards analysing big chunks of text, rather than just names.
I have a few related questions:
Is the tm package appropriate for this sort of task?
Is there a faster alternative to agrep? (Said function uses the
Levenshtein edit distance which is anecdotally slow.)
Are there other suitable tools in R, apart from agrep and tm?
Should I even be doing this in R, or should this sort of thing be
done directly in the database? (It's an Access database, so I'd
rather avoid touching it if possible.)
If you're just doing small batches that are relatively well-formed, then the compare.linkage() or compare.dedup() functions in the RecordLinkage package should be a great starting point. But if you have big batches, then you might have to do some more tinkering.
I use the functions jarowinkler(), levenshteinSim(), and soundex() in RecordLinkage to write my own function that use my own weighting scheme (also, as it is, you can't use soundex() for big data sets with RecordLinkage).
If I have two lists of names that I want to match ("record link"), then I typically convert both to lower case and remove all punctuation. To take care of "Limited" versus "LTD" I typically create another vector of the first word from each list, which allows extra weighting on the first word. If I think that one list may contain acronyms (maybe ATT or IBM) then I'll acronym-ize the other list. For each list I end up with a data frame of strings that I would like to compare that I write as separate tables in a MySQL database.
So that I don't end up with too many candidates, I LEFT OUTER JOIN these two tables on something that has to match between the two lists (maybe that's the first three letters in each list or the first three letters and the first three letters in the acronym). Then I calculate match scores using the above functions.
You still have to do a lot of manual inspection, but you can sort on the score to quickly rule out non-matches.
Maybe google refine could help. It looks maybe more fitted if you have lots of exceptions and you don't know them all yet.
What you're doing is called record linkage, and it's been a huge field of research over many decades already. Luckily for you, there's a whole bunch of tools out there that are ready-made for this sort of thing. Basically, you can point them at your database, set up some cleaning and comparators (like Levenshtein or Jaro-Winkler or ...), and they'll go off and do the job for you.
These tools generally have features in place to solve the performance issues, so that even though Levenshtein is slow they can run fast because most record pairs never get compared at all.
The Wikipedia link above has links to a number of record linkage tools you can use. I've personally written one called Duke in Java, which I've used successfully for exactly this. If you want something big and expensive you can buy a Master Data Management tool.
In your case probably something like edit-distance calculation would work, but if you need to find near duplicates in larger text based documents, you can try
http://www.softcorporation.com/products/neardup/
First off, this may be the wrong Forum for this question, as it's pretty darn R+Bioconductor specific. Here's what I have:
library('GEOquery')
GDS = getGEO('GDS785')
cd4T = GDS2eSet(GDS)
cd4T <- cd4T[!fData(cd4T)$symbol == "",]
Now cd4T is an ExpressionSet object which wraps a big matrix with 19794 rows (probesets) and 15 columns (samples). The final line gets rid of all probesets that do not have corresponding gene symbols. Now the trouble is that most genes in this set are assigned to more than one probeset. You can see this by doing
gene_symbols = factor(fData(cd4T)$Gene.symbol)
length(gene_symbols)-length(levels(gene_symbols))
[1] 6897
So only 6897 of my 19794 probesets have unique probeset -> gene mappings. I'd like to somehow combine the expression levels of each probeset associated with each gene. I don't care much about the actual probe id for each probe. I'd like very much to end up with an ExpressionSet containing the merged information as all of my downstream analysis is designed to work with this class.
I think I can write some code that will do this by hand, and make a new expression set from scratch. However, I'm assuming this can't be a new problem and that code exists to do it, using a statistically sound method to combine the gene expression levels. I'm guessing there's a proper name for this also but my googles aren't showing up much of use. Can anyone help?
I'm not an expert, but from what I've seen over the years everyone has their own favorite way of combining probesets. The two methods that I've seen used the most on a large scale has been using only the probeset which has the largest variance across the expression matrix and the other being to take the mean of the probesets and creating a meta-probeset out of it. For smaller blocks of probesets I've seen people use more intensive methods involving looking at per-probeset plots to get a feel for what's going on ... generally what happens is that one probeset turns out to be the 'good' one and the rest aren't very good.
I haven't seen generalized code to do this - as an example we recently realized in my lab that a few of us have our own private functions to do this same thing.
The word you are looking for is 'nsFilter' in R genefilter package. This function assign two major things, it looks for only entrez gene ids, rest of the probesets will be filtered out. When an entrez id has multiple probesets, then the largest value will be retained and the others removed. Now you have unique entrez gene id mapped matrix. Hope this helps.