Distance calculation optimization in R

点点圈 提交于 2020-06-15 05:30:58

问题


I would like to know if there is any way to optimize the distance calculation process below. I left a small example below, however I am working with a spreadsheet with more than 6000 rows, and it takes considerable time to calculate the variable d. It would be possible to somehow adjust this to have the same results, but in an optimized way.

library(rdist)
library(tictoc)
library(geosphere)

time<-tic()

df<-structure(list(Industries=c(1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19), Latitude = c(-23.8, -23.8, -23.9, -23.9, -23.9,  -23.9, -23.9, -23.9, -23.9, -23.9, -23.9, -23.9, -23.9, -23.9, 
+ + -23.9, -23.9, -23.9, -23.9, -23.9), Longitude = c(-49.6, -49.6, -49.6, -49.6, -49.6, -49.6, -49.6, -49.6, -49.6, -49.6, -49.7, 
+ + -49.7, -49.7, -49.7, -49.7, -49.6, -49.6, -49.6, -49.6)), class = "data.frame", row.names = c(NA, -19L))

k=3 
#clusters
coordinates<-df[c("Latitude","Longitude")]
d<-as.dist(distm(coordinates[,2:1]))
fit.average<-hclust(d,method="average") 
clusters<-cutree(fit.average, k) 
nclusters<-matrix(table(clusters))  
df$cluster <- clusters 

time<-toc()

1.54 sec elapsed

d
          1        2        3        4        5        6        7        8
2      0.00                                                               
3  11075.61 11075.61                                                      
4  11075.61 11075.61     0.00                                             
5  11075.61 11075.61     0.00     0.00                                    
6  11075.61 11075.61     0.00     0.00     0.00                           
7  11075.61 11075.61     0.00     0.00     0.00     0.00                  
8  11075.61 11075.61     0.00     0.00     0.00     0.00     0.00         
9  11075.61 11075.61     0.00     0.00     0.00     0.00     0.00     0.00
10 11075.61 11075.61     0.00     0.00     0.00     0.00     0.00     0.00
11 15048.01 15048.01 10183.02 10183.02 10183.02 10183.02 10183.02 10183.02
12 15048.01 15048.01 10183.02 10183.02 10183.02 10183.02 10183.02 10183.02
13 15048.01 15048.01 10183.02 10183.02 10183.02 10183.02 10183.02 10183.02
14 15048.01 15048.01 10183.02 10183.02 10183.02 10183.02 10183.02 10183.02
15 15048.01 15048.01 10183.02 10183.02 10183.02 10183.02 10183.02 10183.02
16 11075.61 11075.61     0.00     0.00     0.00     0.00     0.00     0.00
17 11075.61 11075.61     0.00     0.00     0.00     0.00     0.00     0.00
18 11075.61 11075.61     0.00     0.00     0.00     0.00     0.00     0.00
19 11075.61 11075.61     0.00     0.00     0.00     0.00     0.00     0.00
          9       10       11       12       13       14       15       16
2                                                                         
3                                                                         
4                                                                         
5                                                                         
6                                                                         
7                                                                         
8                                                                         
9                                                                         
10     0.00                                                               
11 10183.02 10183.02                                                      
12 10183.02 10183.02     0.00                                             
13 10183.02 10183.02     0.00     0.00                                    
14 10183.02 10183.02     0.00     0.00     0.00                           
15 10183.02 10183.02     0.00     0.00     0.00     0.00                  
16     0.00     0.00 10183.02 10183.02 10183.02 10183.02 10183.02         
17     0.00     0.00 10183.02 10183.02 10183.02 10183.02 10183.02     0.00
18     0.00     0.00 10183.02 10183.02 10183.02 10183.02 10183.02     0.00
19     0.00     0.00 10183.02 10183.02 10183.02 10183.02 10183.02     0.00
         17       18
2                   
3                   
4                   
5                   
6                   
7                   
8                   
9                   
10                  
11                  
12                  
13                  
14                  
15                  
16                  
17                  
18     0.00         
19     0.00     0.00

Comparation

> df$cluster <- clusters 
> df
   Industries Latitude Longitude cluster
1           1    -23.8     -49.6       1
2           2    -23.8     -49.6       1
3           3    -23.9     -49.6       2
4           4    -23.9     -49.6       2
5           5    -23.9     -49.6       2
6           6    -23.9     -49.6       2
7           7    -23.9     -49.6       2
8           8    -23.9     -49.6       2
9           9    -23.9     -49.6       2
10         10    -23.9     -49.6       2
11         11    -23.9     -49.7       3
12         12    -23.9     -49.7       3
13         13    -23.9     -49.7       3
14         14    -23.9     -49.7       3
15         15    -23.9     -49.7       3
16         16    -23.9     -49.6       2
17         17    -23.9     -49.6       2
18         18    -23.9     -49.6       2
19         19    -23.9     -49.6       2

> clustered_df
   Industries Latitude Longitude cluster     Dist Cluster
1          11    -23.9     -49.7       3     0.00       1
2          12    -23.9     -49.7       3     0.00       1
3          13    -23.9     -49.7       3     0.00       1
4          14    -23.9     -49.7       3     0.00       1
5          15    -23.9     -49.7       3     0.00       1
6           3    -23.9     -49.6       2 10183.02       2
7           4    -23.9     -49.6       2     0.00       2
8           5    -23.9     -49.6       2     0.00       2
9           6    -23.9     -49.6       2     0.00       2
10          7    -23.9     -49.6       2     0.00       2
11          8    -23.9     -49.6       2     0.00       2
12          9    -23.9     -49.6       2     0.00       2
13         10    -23.9     -49.6       2     0.00       2
14         16    -23.9     -49.6       2     0.00       2
15         17    -23.9     -49.6       2     0.00       2
16         18    -23.9     -49.6       2     0.00       2
17         19    -23.9     -49.6       2     0.00       2
18          1    -23.8     -49.6       1 11075.61       3
19          2    -23.8     -49.6       1     0.00       3

回答1:


@Jose Perhaps not as sound mathematically (in terms of the clustering) but (generally) a better measure of great circle distances (Vincenty's formulae). And ~8 times faster to achieve (what I think is your desired result) - (just using your sample data).

# Order the dataframe by Lon and Lat: ordered_df => data.frame
ordered_df <- 
  df %>% 
  arrange(., Longitude, Latitude)  

# Scalar valued at how many clusters we are expecting => integer vector
k = 3

# Matrix of co-ordinates: coordinates => matrix
coordinates <-   
  ordered_df %>% 
  select(Longitude, Latitude) %>% 
  as.matrix()

# Generate great circle distances between points and Long-Lat Matrix: d => data.frame
d <- data.frame(Dist = c(0, distVincentyEllipsoid(coordinates)))

# Segment the distances into groups: cluster => factor 
d$Cluster <- factor(cumsum(d$Dist > (quantile(d$Dist, 1/k))) + 1)

# Merge with base data: clustered_df => data.frame
clustered_df <- cbind(ordered_df, d)

Libraries and sample data:

library(geosphere)
library(dplyr)

df <- structure(list(Industries=c(1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19), 
Latitude = c(-23.8, -23.8, -23.9, -23.9, -23.9,  -23.9, -23.9, -23.9, -23.9, -23.9, -23.9, -23.9, -23.9, -23.9, -23.9, -23.9, -23.9, -23.9, -23.9),
Longitude = c(-49.6, -49.6, -49.6, -49.6, -49.6, -49.6, -49.6, -49.6, -49.6, -49.6, -49.7,-49.7, -49.7, -49.7, -49.7, -49.6, -49.6, -49.6, -49.6)),
class = "data.frame", row.names = c(NA, -19L))
start_time <- Sys.time()


来源:https://stackoverflow.com/questions/62055976/distance-calculation-optimization-in-r

易学教程内所有资源均来自网络或用户发布的内容,如有违反法律规定的内容欢迎反馈
该文章没有解决你所遇到的问题?点击提问,说说你的问题,让更多的人一起探讨吧!