This question already has answers here:
How can I match fuzzy match strings from two datasets?
(7 answers)
Closed 3 years ago.
I have 2 dataframes and I want to join by name, but names are not written exactly the same:
Df1:
ID Name Age
1 Jose 13
2 M. Jose 12
3 Laura 8
4 Karol P 32
Df2:
Name Surname
José Hall
María José Perez
Laura Alza
Karol Smith
I need to join and get this:
ID Name Age Surname
1 Jose 13 Hall
2 M. Jose 12 Perez
3 Laura 8 Alza
4 Karol P 32 Smith
How to consider that the names are not exactly the same before to join?
You can get close to your result using stringdist_left_join from fuzzyjoin
library(fuzzyjoin)
stringdist_left_join(df1, df2, by = "Name")
# ID Name.x Age Name.y Surname
#1 1 Jose 13 José Hall
#2 2 M. Jose 12 <NA> <NA>
#3 3 Laura 8 Laura Alza
#4 4 Karol P 32 Karol Smith
For the example shared it does not work for 1 entry since it is difficult to match Maria with M.. You can get the result for it by adjusting the max_dist argument to a higher value (default is 2) however, this will screw up other results and would give unwanted matches. If you have minimal NA entries (like the example shared) after this join you could just match them by "hand".
I would clean the database before (for example deleting those ´, in excel is easy doing those replace) and then use
new_df <- merge(df1, df2, by="name")
or you could try to assign an ID for df2 that coincide with df2 if it is possible.
Related
I'm trying to use visNetwork to create a node diagram. However, my data is not in the correct format and I haven't been able to find any help on this on the internet.
My current data frame looks similar to this:
name town car color age school
John Bringham Swift Red 22 Brighton
Sarah Bringham Corolla Red 33 Rustal
Beth Burb Swift Blue 43 Brighton
Joe Spring Polo Black 18 Riding
I'm wanting to change use this to create nodes and edges lists that can be used to create a vis network.
I know that the "nodes" list will be made from the unique values in the "name" column but I'm not sure how I would use the rest of the data to create the "edges" list?
I was thinking that it may be possible to group by each column and then read back the matches from this function but I am not sure how to implement this. The idea that I thought of is to weight the edges based on how many matches they detect in the various group by functions. I'm not sure how to actually implement this yet.
For example, Joe will not match with anyone because he shares no common columns with any of the others. John and Sarah will have a weight of 2 because they share two common columns.
Also open to solutions in python!
One option is to compar row by row, in order to calculate the number of commun values.
For instance for John (first row) and Sarah (second row):
sum(df[1,] == df[2,])
# 2
Then you use the function combn() from library utils to know in advance the number of pair-combinaison you have to calculate:
nodes <- matrix(combn(df$name, 2), ncol = 2, byrow = T) %>% as.data.frame()
nodes$V1 <- as.character(nodes$V1)
nodes$V2 <- as.character(nodes$V2)
nodes$weight <- NA
(nodes)
# V1 V2 weight
#1 John Sarah NA
#2 John Beth NA
#3 John Joe NA
#4 Sarah Beth NA
#5 Sarah Joe NA
#6 Beth Joe NA
Finally a loop to calculate weight for each node.
for(n in 1:nrow(nodes)){
name1 <- df[df$name == nodes$V1[n],]
name2 <- df[df$name == nodes$V2[n],]
nodes$weight[n] <- sum(name1 == name2)
}
# V1 V2 weight
#1 John Sarah 2
#2 John Beth 2
#3 John Joe 0
#4 Sarah Beth 0
#5 Sarah Joe 0
#6 Beth Joe 0
I think node will be the kind of dataframe that you can use in the function visNetwork().
One data table (let's call is A) contains the ID numbers:
ID
3
5
12
8
...
and another table (let's call it B) contains the lower bound and the upper bound and the name for that ID.
ID_lower ID_upper Name
1 4 James
5 7 Arthur
8 11 Jacob
12 13 Sarah
so based on table B, given the ID from table A, we can find the matching name by finding the name on the row in table B such that
ID_lower <= ID <= ID upper
and I wanna create a table of ID and Name, so in the above example, it would be
ID Name
3 James
5 Arthur
12 Sarah
8 Jacob
... ...
I used for loop, so that for each row of A, I look for the row in B such that ID is between the ID_lower and ID_upper for that row and joined the name from there.
However, this method was a bit slow. Is there a fast way of doing it in R?
Using the new non-equi joins feature in the current development version of data.table, this is straightforward:
require(data.table) # v1.9.7+
dt2[dt1, .(ID, Name), on=.(ID_lower <= ID, ID_upper >= ID)]
See the installation instructions for devel version here.
where,
dt1=fread('ID
3
5
12
8')
dt2 = fread('ID_lower ID_upper Name
1 4 James
5 7 Arthur
8 11 Jacob
12 13 Sarah')
You can make a look-up table with your second data.frame (B):
lu <- do.call(rbind,
apply(B,1,function(x)
data.frame(ID=c(x[1]:x[2]),Name=x[3], row.names = NULL)))
then you query it with your first data.frame (A):
A$Name <- lu[A$ID,"Name"]
You can try this data.table solution:
data.table::setDT(B)[, .(Name, ID = Map(`:`, ID_lower, ID_upper))]
[, .(ID = unlist(ID)), .(Name)][ID %in% A$ID]
Name ID
1: James 3
2: Arthur 5
3: Sarah 12
4: Jacob 8
I believe findInterval() on ID_lower might be the ideal approach here:
A[,Name:=B[findInterval(ID,ID_lower),Name]];
A;
## ID Name
## 1: 3 James
## 2: 5 Arthur
## 3: 12 Sarah
## 4: 8 Jacob
This will only be correct if (1) B is sorted by ID_lower and (2) all values in A$ID are covered by the ranges in B.
I have a df where each row represents an individual and each column a characteristic of these individuals. One of the columns is TeamName, which is the name of the Team that individual belongs to. Multiple individuals belong to a Team.
I'd like a function in R that creates a new column with the number of team members for each Team.
So, for example I have:
df
Name Surname TeamName
John Smith Champions
Mary Osborne Socceroos
Mark Johnson Champions
Rory Bradon Champions
Jane Bryant Socceroos
Bruce Harper
I'd like to have
df1
Name Surname TeamName TeamNo
John Smith Champions 3
Mary Osborne Socceroos 2
Mark Johnson Champions 3
Rory Bradon Champions 3
Jane Bryant Socceroos 2
Bruce Harper 0
So as you can see the counting includes that individual too, and if someone (e.g. Bruce Harper) has no Team name, then he gets a 0.
How can I do that? Thanks!
This is a solution based on using data.table which perhaps is too much for what you need, but here it goes:
library(data.table)
dt=data.table(df)
# First, let's convert the factors of TeamName, to characters
dt[,TeamName:=as.character(TeamName)]
# Now, let find all the team numbers
dt[,TeamNo:=.N, by='TeamName']
# Let's exclude the special cases
dt[is.na(TeamName),TeamNo:=NA]
dt[TeamName=="",TeamNo:=NA]
It is clearly not the best solution, but I hope this helps
If you need to know the number of unique members in the first two columns based on the 'TeamName' column, one option is n_distinct from dplyr
library(dplyr)
library(tidyr)
df %>%
unite(Var, Name, Surname) %>% #paste the columns together
group_by(TeamName) %>% #group by TeamName
mutate(TeamNo= n_distinct(Var)) %>% #create the TeamNo column
separate(Var, into=c('Name', 'Surname')) #split the 'Var' column
Or if it just the number of rows per 'TeamName', we can group by 'TeamName', get the number of rows per group with n(), create the 'TeamNo' column with mutate based on that n(), and if needed an ifelse condition can be used to give NA for 'TeamName' that are '' or NA.
df %>%
group_by(TeamName) %>%
mutate(TeamNo = ifelse(is.na(TeamName)|TeamName=='', NA_integer_, n()))
# Name Surname TeamName TeamNo
#1 John Smith Champions 3
#2 Mary Osborne Socceroos 2
#3 Mark Johnson Champions 3
#4 Rory Bradon Champions 3
#5 Jane Bryant Socceroos 2
#6 Bruce Harper NA
Or you can use ave from base R. Suppose if there are '' and NA, I would first convert the '' to NA and then use ave to get the length of 'TeamNo' grouped by that column. It will give NA for `NA' values. For example.
v1 <- c(df$TeamName, NA)# appending an NA with the example to show the case
is.na(v1) <- v1=='' #convert the `'' to `NA`
as.numeric(ave(v1, v1, FUN=length))
#[1] 3 2 3 3 2 NA NA
Using sqldf:
library(sqldf)
sqldf("SELECT Name, Surname, TeamName, n
FROM df
LEFT JOIN
(SELECT TeamName, COUNT(Name) AS n
FROM df
WHERE NOT TeamName IS '' GROUP BY TeamName)
USING (TeamName)")
Output:
Name Surname TeamName n
1 John Smith Champions 3
2 Mary Osborne Socceroos 2
3 Mark Johnson Champions 3
4 Rory Bradon Champions 3
5 Jane Bryant Socceroos 2
6 Bruce Harper NA
I have a data.frame with 1,000 rows and 3 columns. It contains a large number of duplicates and I've used plyr to combine the duplicate rows and add a count for each combination as explained in this thread.
Here's an example of what I have now (I still also have the original data.frame with all of the duplicates if I need to start from there):
name1 name2 name3 total
1 Bob Fred Sam 30
2 Bob Joe Frank 20
3 Frank Sam Tom 25
4 Sam Tom Frank 10
5 Fred Bob Sam 15
However, column order doesn't matter. I just want to know how many rows have the same three entries, in any order. How can I combine the rows that contain the same entries, ignoring order? In this example I would want to combine rows 1 and 5, and rows 3 and 4.
Define another column that's a "sorted paste" of the names, which would have the same value of "Bob~Fred~Sam" for rows 1 and 5. Then aggregate based on that.
Brief code snippet (assumes original data frame is dd): it's all really intuitive. We create a lookup column (take a look and should be self explanatory), get the sums of the total column for each combination, and then filter down to the unique combinations...
dd$lookup=apply(dd[,c("name1","name2","name3")],1,
function(x){paste(sort(x),collapse="~")})
tab1=tapply(dd$total,dd$lookup,sum)
ee=dd[match(unique(dd$lookup),dd$lookup),]
ee$newtotal=as.numeric(tab1)[match(ee$lookup,names(tab1))]
You now have in ee a set of unique rows and their corresponding total counts. Easy - and no external packages needed. And crucially, you can see at every stage of the process what is going on!
(Minor update to help OP:) And if you want a cleaned-up version of the final answer:
outdf = with(ee,data.frame(name1,name2,name3,
total=newtotal,stringsAsFactors=FALSE))
This gives you a neat data frame with the three all-important name columns, and with the aggregated totals in a column called total rather than newtotal.
Sort the index columns, then use ddply to aggregate and sum:
Define the data:
dat <- " name1 name2 name3 total
1 Bob Fred Sam 30
2 Bob Joe Frank 20
3 Frank Sam Tom 25
4 Sam Tom Frank 10
5 Fred Bob Sam 15"
x <- read.table(text=dat, header=TRUE)
Create a copy:
xx <- x
Use apply to sort the columns, then aggregate:
xx[, -4] <- t(apply(xx[, -4], 1, sort))
library(plyr)
ddply(xx, .(name1, name2, name3), numcolwise(sum))
name1 name2 name3 total
1 Bob Frank Joe 20
2 Bob Fred Sam 45
3 Frank Sam Tom 35
I keep reading about the importance of vectorized functionality so hopefully someone can help me out here.
Say I have a data frame with two columns: name and ID. Now I also have another data frame with name and birthplace, but this data frame is much larger than the first, and contains some but not all of the names from the first data frame. How can I add a third column to the the first table that is populated with birthplaces looked up using the second table.
What I have is now is:
corresponding.birthplaces <- sapply(table1$Name,
function(name){return(table2$Birthplace[table2$Name==name])})
This seems inefficient. Thoughts? Does anyone know of a good book/resource for using R 'properly'..I get the feeling that I generally do think in the least computationally effective manner conceivable.
Thanks :)
See ?merge which will perform a database link merge or join.
Here is an example:
set.seed(2)
d1 <- data.frame(ID = 1:5, Name = c("Bill","Bob","Jessica","Jennifer","Robyn"))
d2 <- data.frame(Name = c("Bill", "Gavin", "Bob", "Joris", "Jessica", "Andrie",
"Jennifer","Joshua","Robyn","Iterator"),
Birthplace = sample(c("London","New York",
"San Francisco", "Berlin",
"Tokyo", "Paris"), 10, rep = TRUE))
which gives:
> d1
ID Name
1 1 Bill
2 2 Bob
3 3 Jessica
4 4 Jennifer
5 5 Robyn
> d2
Name Birthplace
1 Bill New York
2 Gavin Tokyo
3 Bob Berlin
4 Joris New York
5 Jessica Paris
6 Andrie Paris
7 Jennifer London
8 Joshua Paris
9 Robyn San Francisco
10 Iterator Berlin
Then we use merge() to do the join:
> merge(d1, d2)
Name ID Birthplace
1 Bill 1 New York
2 Bob 2 Berlin
3 Jennifer 4 London
4 Jessica 3 Paris
5 Robyn 5 San Francisco