Quick and dirty analysis of a Twitter social network
I use R to retrieve some data from Twitter, do some exploratory data analysis and visualisation and examine a network of followers.
Motivation
I use Twitter to get live updates of what my follow scientists are up to, to communicate about my students’ awesome work and to share material that I hope is useful to some people1.
Recently, I reached 5,000 followers and I thought I’d spend some time trying to know better who they/you are. To do so, I use R to retrieve some data from Twitter using
rtweet, do some data exploration and visualisation using the
tidyverse and examine my network of followers with
tidygraph,
ggraph and
igraph. Data and codes are available from
https://github.com/oliviergimenez/sna-twitter.
To reproduce the analyses below, you will need to access a Twitter API (application programming interface) to retrieve the information about your followers. In brief, an API is an intermediary application that allows applications to talk to each other. To access the Twitter APIs, you need a developer account for which you may apply at https://developer.twitter.com/en. There is a short form to fill in, and it takes less than a day to get an answer.
Below I rely heavily on the code shared by Joe Cristian through the Algoritma Technical Blog at https://algotech.netlify.app/blog/social-network-analysis-in-r/. Kuddos and credits to him.
Data retrieving
We load the rtweet package to work with Twitter from R.
#devtools::install_github("ropensci/rtweet")
library(rtweet)
We also load the tidyverse for data manipulation and visualisation.
library(tidyverse)
theme_set(theme_light(base_size = 14))
First we need to get credentials. The usual rtweet sequence with rtweet_app, auth_save and auth_as is supposed to work (see
here), but the Twitter API kept failing (error 401) for me. Tried a few things, in vain. I will use the deprecated create_token function instead. You might need to change the defaults of your Twitter app from “Read only” to “Read, write and access direct messages”.
Enter my API keys and access tokens (not shown), then authenticate.
token <- create_token(
app = "sna-twitter-network-5k",
consumer_key = api_key,
consumer_secret = api_secret_key,
access_token = access_token,
access_secret = access_token_secret)
Get my Twitter info with my description, number of followers, number of likes, etc.
og <- lookup_users("oaggimenez")
str(og, max = 2)
## 'data.frame': 1 obs. of 21 variables:
## $ id : num 7.51e+17
## $ id_str : chr "750892662224453632"
## $ name : chr "Olivier Gimenez \U0001f596"
## $ screen_name : chr "oaggimenez"
## $ location : chr "Montpellier, France"
## $ description : chr "Scientist (he/him) @CNRS @cefemontpellier @twitthair1 • Grown statistician • Improvised ecologist • Sociologist"| __truncated__
## $ url : chr "https://t.co/l7NImYeGdY"
## $ protected : logi FALSE
## $ followers_count : int 5042
## $ friends_count : int 2191
## $ listed_count : int 68
## $ created_at : chr "Thu Jul 07 03:22:16 +0000 2016"
## $ favourites_count : int 18066
## $ verified : logi FALSE
## $ statuses_count : int 5545
## $ profile_image_url_https: chr "https://pbs.twimg.com/profile_images/1330619806396067845/mIPmR-x4_normal.jpg"
## $ profile_banner_url : chr "https://pbs.twimg.com/profile_banners/750892662224453632/1602417664"
## $ default_profile : logi FALSE
## $ default_profile_image : logi FALSE
## $ withheld_in_countries :List of 1
## ..$ : list()
## $ entities :List of 1
## ..$ :List of 2
## - attr(*, "tweets")='data.frame': 1 obs. of 36 variables:
## ..$ created_at : chr "Thu Jul 29 16:15:52 +0000 2021"
## ..$ id : num 1.42e+18
## ..$ id_str : chr "1420780122949529601"
## ..$ text : chr "Looking forward to digging into this paper \U0001f929\U0001f9ee\U0001f60d https://t.co/ddCMraM3kj"
## ..$ truncated : logi FALSE
## ..$ entities :List of 1
## ..$ source : chr "<a href=\"http://twitter.com/download/iphone\" rel=\"nofollow\">Twitter for iPhone</a>"
## ..$ in_reply_to_status_id : logi NA
## ..$ in_reply_to_status_id_str : logi NA
## ..$ in_reply_to_user_id : logi NA
## ..$ in_reply_to_user_id_str : logi NA
## ..$ in_reply_to_screen_name : logi NA
## ..$ geo : logi NA
## ..$ coordinates :List of 1
## ..$ place :List of 1
## ..$ contributors : logi NA
## ..$ is_quote_status : logi TRUE
## ..$ quoted_status_id : num 1.42e+18
## ..$ quoted_status_id_str : chr "1420768369620312065"
## ..$ retweet_count : int 0
## ..$ favorite_count : int 8
## ..$ favorited : logi FALSE
## ..$ retweeted : logi FALSE
## ..$ possibly_sensitive : logi FALSE
## ..$ lang : chr "en"
## ..$ quoted_status :List of 1
## ..$ display_text_width : int 70
## ..$ user :List of 1
## ..$ full_text : logi NA
## ..$ favorited_by : logi NA
## ..$ display_text_range : logi NA
## ..$ retweeted_status : logi NA
## ..$ quoted_status_permalink : logi NA
## ..$ metadata : logi NA
## ..$ query : logi NA
## ..$ possibly_sensitive_appealable: logi NA
Now I obtain the id of my followers using get_followers.
followers <- get_followers(user = "oaggimenez",
n = og$followers_count,
retryonratelimit = T)
From their id, I can get the same details I got on my account using lookup_users. This function is not vectorized, therefore I use a loop. Takes some time so I saved the results and load them.
details_followers <- NULL
for (i in 1:length(followers$user_id)){
tmp <- try(lookup_users(followers$user_id[i], retryonratelimit = TRUE), silent = TRUE)
if (length(tmp) == 1){
next
} else {
tmp$listed_count <- NULL # get rid of this column which raised some format issues, we do not it anyway
details_followers <- bind_rows(details_followers, tmp)
}
}
save(details_followers, file = "details_followers.RData")
What info do we have?
load("dat/details_followers.RData")
names(details_followers)
## [1] "id" "id_str"
## [3] "name" "screen_name"
## [5] "location" "description"
## [7] "url" "protected"
## [9] "followers_count" "friends_count"
## [11] "created_at" "favourites_count"
## [13] "verified" "statuses_count"
## [15] "profile_image_url_https" "profile_banner_url"
## [17] "default_profile" "default_profile_image"
## [19] "withheld_in_countries" "entities"
In more details.
str(details_followers, max = 1)
## 'data.frame': 5025 obs. of 20 variables:
## $ id : num 1.03e+18 1.42e+18 1.32e+09 1.42e+18 1.36e+18 ...
## $ id_str : chr "1026053171024736258" "1419513064697712640" "1317282258" "1419222504367927296" ...
## $ name : chr "Dr. Zoe Nhleko" "akash Shelke" "Rilwan Ugbedeojo" "Thaana Van Dessel" ...
## $ screen_name : chr "ZoeNhleko" "akashSh88945929" "abuhrilwan" "ThaanaD" ...
## $ location : chr "United States" "" "Lagos, Nigeria" "France" ...
## $ description : chr "Wildlife ecologist with experience on African large mammals. PhD from @UF. Looking for job opportunities in eco"| __truncated__ "" "MSc, Plant Ecology | Seeking PhD Position in Community Ecology | Civic Leader | YALI RLC & African Presidential"| __truncated__ "MSc student majoring in Animal Ecology @WURanimal. \nFocused on Conservation Behavior & Human-Wildlife Coexistence." ...
## $ url : chr "https://t.co/uaFY2L8BnL" NA NA NA ...
## $ protected : logi FALSE FALSE FALSE FALSE FALSE FALSE ...
## $ followers_count : int 2878 1 244 0 2856 4 112 6 69 476 ...
## $ friends_count : int 1325 223 2030 14 2575 99 344 32 73 1671 ...
## $ created_at : chr "Sun Aug 05 10:31:53 +0000 2018" "Mon Jul 26 04:21:18 +0000 2021" "Sat Mar 30 22:35:27 +0000 2013" "Sun Jul 25 09:06:36 +0000 2021" ...
## $ favourites_count : int 8003 0 1843 1 12360 0 484 0 83 713 ...
## $ verified : logi FALSE FALSE FALSE FALSE FALSE FALSE ...
## $ statuses_count : int 7261 0 670 1 2112 0 141 0 30 926 ...
## $ profile_image_url_https: chr "https://pbs.twimg.com/profile_images/1306353853571280896/rk9qAxoa_normal.jpg" "https://pbs.twimg.com/profile_images/1419513134142738433/erR2XRwf_normal.png" "https://pbs.twimg.com/profile_images/1413368052012433413/EyzbYfWB_normal.jpg" "https://pbs.twimg.com/profile_images/1419222962566217731/TQT96npR_normal.jpg" ...
## $ profile_banner_url : chr "https://pbs.twimg.com/profile_banners/1026053171024736258/1533465589" NA "https://pbs.twimg.com/profile_banners/1317282258/1624183687" "https://pbs.twimg.com/profile_banners/1419222504367927296/1627205254" ...
## $ default_profile : logi TRUE TRUE FALSE TRUE TRUE TRUE ...
## $ default_profile_image : logi FALSE FALSE FALSE FALSE FALSE FALSE ...
## $ withheld_in_countries :List of 5025
## $ entities :List of 5025
## - attr(*, "tweets")='data.frame': 1 obs. of 36 variables:
Data exploration and visualisation
Let’s display the 100 bigger accounts that follow me. First thing I learned. It is humbling to be followed by influential individuals I admire like @MicrobiomDigest, @nathanpsmad, @FrancoisTaddei, @freakonometrics, @allison_horst, @HugePossum, @apreshill and scientific journals, institutions and societies.
details_followers %>%
arrange(-followers_count) %>%
select(screen_name,
followers_count,
friends_count,
favourites_count) %>%
head(n = 100)
## screen_name followers_count friends_count favourites_count
## 1 dianefrancis1 230560 132635 70
## 2 simongerman600 218444 209489 56976
## 3 WildlifeMag 191613 23995 14697
## 4 cagrimbakirci 170164 9401 21334
## 5 BernardMarr 130900 29532 15764
## 6 ZotovMax 119625 133610 894
## 7 _atanas_ 116717 74266 102899
## 8 MicrobiomDigest 105825 33779 98199
## 9 Eliances 97488 94488 15164
## 10 PLOSBiology 89263 6646 4278
## 11 IPBES 78086 11797 53957
## 12 7wData 76131 84155 74609
## 13 biconnections 63802 52918 91970
## 14 aydintufekci43 60212 22627 99427
## 15 AliceVachet 56381 2320 182435
## 16 Pr_Logos 52482 50520 19852
## 17 DrJoeNyangon 51296 20858 1413
## 18 derekantoncich 48823 51449 347
## 19 Afro_Herper 45215 3392 117693
## 20 KumarAGarg 43461 47795 4052
## 21 GatelyMark 41500 37944 1153
## 22 overleaf 41306 7830 35289
## 23 figshare 40610 38361 10395
## 24 mlamons1 40415 29364 2498
## 25 coraman 40398 2169 16548
## 26 BritishEcolSoc 37394 1143 11685
## 27 nathanpsmad 36760 5312 43729
## 28 WildlifeRescue 34843 7976 61284
## 29 rudyagovic 34063 28147 3708
## 30 grp_resilience 34011 7998 3559
## 31 JNakev 33427 28532 29442
## 32 FrancoisTaddei 32631 17903 23894
## 33 RunEducator 31248 19848 17522
## 34 LarryLLapin 30782 34265 501
## 35 _Alex_Iaco_ 30775 26720 2390
## 36 owasow 30562 12932 36909
## 37 JEcology 30534 689 3312
## 38 valmasdel 28972 4512 4334
## 39 freakonometrics 28038 15112 31106
## 40 MethodsEcolEvol 26207 10347 2083
## 41 mmw_lmw 26014 12157 1068
## 42 drmichellelarue 25424 6281 56120
## 43 nicolasberrod 25258 6042 12990
## 44 DD_NaNa_ 24253 11051 39916
## 45 callin_bull 24106 2831 10572
## 46 INEE_CNRS 23483 1873 5089
## 47 Datascience__ 22793 10695 216
## 48 VisualPersist 22779 5599 79872
## 49 coywolfassoc 22203 19491 9470
## 50 RBGSydney 20259 16049 13435
## 51 allison_horst 19289 2865 7725
## 52 rhskraus 18429 20029 3503
## 53 fred_sallee 18144 18995 3423
## 54 thembauk 17748 2066 3260
## 55 parolesdhist 17257 7136 7497
## 56 CREAF_ecologia 16654 3201 11899
## 57 ChamboncelLea 16609 4059 9797
## 58 HugePossum 16461 12732 7806
## 59 ScientistFemale 16379 9039 2079
## 60 sophie_e_hill 16292 5811 10315
## 61 mohamma64508589 16130 14550 24584
## 62 SamuelHayat 15999 3068 9767
## 63 MathieuAvanzi 15890 1604 8526
## 64 Erky321 15859 16284 4841
## 65 ConLetters 15482 5547 1489
## 66 oceansresearch 14969 1759 7599
## 67 WCSNewsroom 14926 11224 7277
## 68 CMastication 14687 8091 31769
## 69 apreshill 14106 1607 32806
## 70 SumanMalla10 13872 11977 10946
## 71 umontpellier 13667 1360 8745
## 72 WriteNThrive 13615 13523 7631
## 73 abdirashidmd 13586 2389 18417
## 74 JackKinross 13392 11263 774
## 75 PaulREhrlich 13383 14208 1
## 76 teppofelin 13330 12410 264
## 77 ConservOptimism 13249 4306 15648
## 78 jepeteso 13137 13595 3420
## 79 gwintrob 12484 8751 43479
## 80 SallyR_ISLHE 12411 12477 28924
## 81 ActuIAFr 12321 8856 2495
## 82 Booksmag 12282 1785 332
## 83 jusoneoftoomany 12212 11049 2296
## 84 galaxyproject 12208 8757 813
## 85 thepostdoctoral 12186 12802 814
## 86 FrontMarineSci 11498 5733 5832
## 87 jobRxiv 11442 872 786
## 88 EcographyJourna 11381 1465 4938
## 89 davidghamilton1 11242 2231 16244
## 90 5amStartups 11170 12370 147062
## 91 RosmarieKatrin 11079 11159 4060
## 92 ScienceAndMaps 11043 8679 3596
## 93 joelgombin 11012 10335 10868
## 94 SteveLenhert 10990 8328 669
## 95 Naturevolve 10973 11349 7500
## 96 cookinforpeace 10655 2861 17519
## 97 DrPaulEvans1 10537 7927 3131
## 98 Ibycter 10517 4450 47359
## 99 Sciencecomptoir 10436 1268 15800
## 100 SteveAtLogikk 10421 8864 13123
Where do my followers come from? Second thing I learnt. Followers are folks from France (my home), USA, UK, Canada, Australia, and other European countries. I was happy to realize that I have followers from other parts of the world too, in India, South Africa, Brazil, Peru, Mexico, Nepal, Argentina, Kenya, Bangladesh, etc. and from Greece (my other home).
details_followers %>%
mutate(location = str_replace(location, ".*France.*", "France"),
location = str_replace(location, ".*England", "United Kingdom"),
location = str_replace(location, ".*United Kingdom", "United Kingdom"),
location = str_replace(location, ".*London", "United Kingdom"),
location = str_replace(location, ".*Scotland", "United Kingdom"),
location = str_replace(location, ".*Montpellier", "France"),
location = str_replace(location, ".*Canada, BC.*", "Canada"),
location = str_replace(location, ".*New Caledonia", "France"),
location = str_replace(location, ".*Marseille", "France"),
location = str_replace(location, ".*Grenoble", "France"),
location = str_replace(location, ".*Perú", "Peru"),
location = str_replace(location, ".*Martinique", "France"),
location = str_replace(location, ".*La Rochelle", "France"),
location = str_replace(location, ".*Grenoble, Rhône-Alpes", "France"),
location = str_replace(location, ".*Avignon", "France"),
location = str_replace(location, ".*Paris", "France"),
location = str_replace(location, ".*paris", "France"),
location = str_replace(location, ".*france", "France"),
location = str_replace(location, ".*Bordeaux", "France"),
location = str_replace(location, ".*Mytilene, Lesvos, Greece", "Greece"),
location = str_replace(location, ".*Mytiline, Greece", "Greece"),
location = str_replace(location, ".*Germany.*", "Germany"),
location = str_replace(location, ".*Vienna, Austria", "Austria"),
location = str_replace(location, ".*Arusha, Tanzania", "Tanzania"),
location = str_replace(location, ".*Athens", "Greece"),
location = str_replace(location, ".*Bangor, ME", "United Kingdom"),
location = str_replace(location, ".*Bangor, Wales", "United Kingdom"),
location = str_replace(location, ".*Baton Rouge, LA", "United States of America"),
location = str_replace(location, ".*Lisboa, Portugal", "Portugal"),
location = str_replace(location, ".*Oslo", "Norway"),
location = str_replace(location, ".*Islamic Republic of Iran", "Iran"),
location = str_replace(location, ".*Madrid, España", "Spain"),
location = str_replace(location, ".*Madrid", "Spain"),
location = str_replace(location, ".*Madrid, Spain", "Spain"),
location = str_replace(location, ".*Rome, Italy", "Italy"),
location = str_replace(location, ".*Santiago, Chile", "Chile"),
location = str_replace(location, ".*Belém, Brazil", "Brazil"),
location = str_replace(location, ".*Belo Horizonte, Brazil", "Brazil"),
location = str_replace(location, ".*Berlin", "Germany"),
location = str_replace(location, ".*Mexico, ME", "Mexico"),
location = str_replace(location, ".*Mexico City", "Mexico"),
location = str_replace(location, ".*Santiago, Chile", "Chile"),
location = str_replace(location, ".*Ontario", "Canada"),
location = str_replace(location, ".*Aberdeen", "United Kingdom"),
location = str_replace(location, ".*Cambridge", "United Kingdom"),
location = str_replace(location, ".*Bologna, Emilia Romagna", "Italy"),
location = str_replace(location, ".*Bogotá, D.C., Colombia", "Colombia"),
location = str_replace(location, ".*Medellín, Colombia", "Colombia"),
location = str_replace(location, ".*Bruxelles, Belgique", "Belgium"),
location = str_replace(location, ".*Brasil", "Brazil"),
location = str_replace(location, ".*Budweis, Czech Republic", "Czech Republic"),
location = str_replace(location, ".*Calgary, Alberta", "Canada"),
location = str_replace(location, ".*Canada, BC", "Canada"),
location = str_replace(location, ".*Cardiff, Wales", "United Kingdom"),
location = str_replace(location, ".*Brisbane", "Australia"),
location = str_replace(location, ".*Sydney", "Australia"),
location = str_replace(location, ".*Queensland", "Australia"),
location = str_replace(location, ".*Australia", "Australia"),
location = str_replace(location, ".*Germany", "Germany"),
location = str_replace(location, ".*Vancouver", "Canada"),
location = str_replace(location, ".*Ottawa, Ontario", "Canada"),
location = str_replace(location, ".*Québec, Canada", "Canada"),
location = str_replace(location, ".*Winnipeg, Manitoba", "Canada"),
location = str_replace(location, ".*New South Wales", "Canada"),
location = str_replace(location, ".*Victoria", "Canada"),
location = str_replace(location, ".*British Columbia", "Canada"),
location = str_replace(location, ".*Norway", "Norway"),
location = str_replace(location, ".*Finland", "Finland"),
location = str_replace(location, ".*South Africa", "South Africa"),
location = str_replace(location, ".*Switzerland", "Switzerland"),
location = str_replace(location, ".*CO", "United States of America"),
location = str_replace(location, ".*OK", "United States of America"),
location = str_replace(location, ".*KS", "United States of America"),
location = str_replace(location, ".*MS", "United States of America"),
location = str_replace(location, ".*CO", "United States of America"),
location = str_replace(location, ".*CO", "United States of America"),
location = str_replace(location, ".*CO", "United States of America"),
location = str_replace(location, ".*CO", "United States of America"),
location = str_replace(location, ".*WA", "United States of America"),
location = str_replace(location, ".*MD", "United States of America"),
location = str_replace(location, ".*Colorado", "United States of America"),
location = str_replace(location, ".*Community of Valencia, Spain", "Spain"),
location = str_replace(location, ".*Dunedin City, New Zealand", "New Zealand"),
location = str_replace(location, ".*Rio Claro, Brasil", "Brazil"),
location = str_replace(location, ".*Saskatoon, Saskatchewan", "Canada"),
location = str_replace(location, ".*Sherbrooke, Québec", "Canada"),
location = str_replace(location, ".*United Kingdom.*", "United Kingdom"),
location = str_replace(location, ".*University of Oxford", "United Kingdom"),
location = str_replace(location, ".*University of St Andrews", "United Kingdom"),
location = str_replace(location, ".*ID", "United States of America"),
location = str_replace(location, ".*NE", "United States of America"),
location = str_replace(location, ".*United States of America, USA", "United States of America"),
location = str_replace(location, ".*Spain, Spain", "Spain"),
location = str_replace(location, ".*USA", "United States of America"),
location = str_replace(location, ".*Wisconsin, USA", "United States of America"),
location = str_replace(location, ".*Florida, USA", "United States of America"),
location = str_replace(location, ".*Liege, Belgium", "Belgium"),
location = str_replace(location, ".*Ghent, Belgium", "Belgium"),
location = str_replace(location, ".*Pune, India", "India"),
location = str_replace(location, ".*Hyderabad, India", "India"),
location = str_replace(location, ".*Prague, Czech Republic", "Czech Republic"),
location = str_replace(location, ".*Canada, BC, Canada", "Canada"),
location = str_replace(location, ".*CA", "United States of America"),
location = str_replace(location, ".*United States.*", "United States of America"),
location = str_replace(location, ".*DC", "United States of America"),
location = str_replace(location, ".*FL", "United States of America"),
location = str_replace(location, ".*GA", "United States of America"),
location = str_replace(location, ".*HI", "United States of America"),
location = str_replace(location, ".*ME", "United States of America"),
location = str_replace(location, ".*MA", "United States of America"),
location = str_replace(location, ".*MI", "United States of America"),
location = str_replace(location, ".*PA", "United States of America"),
location = str_replace(location, ".*NC", "United States of America"),
location = str_replace(location, ".*MO", "United States of America"),
location = str_replace(location, ".*NY", "United States of America"),
location = str_replace(location, ".*NH", "United States of America"),
location = str_replace(location, ".*IL", "United States of America"),
location = str_replace(location, ".*NM", "United States of America"),
location = str_replace(location, ".*MT", "United States of America"),
location = str_replace(location, ".*OR", "United States of America"),
location = str_replace(location, ".*WY", "United States of America"),
location = str_replace(location, ".*WI", "United States of America"),
location = str_replace(location, ".*MN", "United States of America"),
location = str_replace(location, ".*CT", "United States of America"),
location = str_replace(location, ".*TX", "United States of America"),
location = str_replace(location, ".*VA", "United States of America"),
location = str_replace(location, ".*OH", "United States of America"),
location = str_replace(location, ".*Massachusetts, USA", "United States of America"),
location = str_replace(location, ".*California, USA", "United States of America"),
location = str_replace(location, ".*Montréal, Québec", "Canada"),
location = str_replace(location, ".*Edmonton, Alberta", "Canada"),
location = str_replace(location, ".*Toronto, Ontario", "Canada"),
location = str_replace(location, ".*Canada, Canada", "Canada"),
location = str_replace(location, ".*Montreal", "Canada"),
location = str_replace(location, ".*Lisbon, Portugal", "Portugal"),
location = str_replace(location, ".*Coimbra, Portugal", "Portugal"),
location = str_replace(location, ".*Cork, Ireland", "Ireland"),
location = str_replace(location, ".*Dublin City, Ireland", "Ireland"),
location = str_replace(location, ".*Barcelona, Spain", "Spain"),
location = str_replace(location, ".*Barcelona", "Spain"),
location = str_replace(location, ".*Leipzig", "Germany"),
location = str_replace(location, ".*Seville, Spain", "Spain"),
location = str_replace(location, ".*Seville, Spain", "Spain"),
location = str_replace(location, ".*Buenos Aires, Argentina", "Argentina"),
location = str_replace(location, ".*Rio de Janeiro, Brazil", "Brazil"),
location = str_replace(location, ".*Canberra", "Australia"),
location = str_remove(location, "Global"),
location = str_remove(location, "Earth"),
location = str_remove(location, "Worldwide"),
location = str_remove(location, "Europe"),
location = str_remove(location, " "),
location = str_replace(location, ".*Dhaka, Bangladesh", "Bangladesh"),
location = str_replace(location, ".*Copenhagen, Denmark", "Denmark"),
location = str_replace(location, ".*Amsterdam, The Netherlands", "The Netherlands"),
location = str_replace(location, ".*Groningen, Nederland", "The Netherlands"),
location = str_replace(location, ".*Wageningen, Nederland", "The Netherlands"),
location = str_replace(location, ".*Aarhus, Denmark", "Denmark"),
location = str_replace(location, ".*Antwerp, Belgium", "Belgium"),
location = str_replace(location, ".*Aveiro, Portugal", "Portugal"),
location = str_replace(location, ".*Australia, AUS", "Australia"),
location = str_replace(location, ".*Australian National University", "Australia"),
location = str_replace(location, ".*Auckland, New Zealand", "New Zealand"),
location = str_replace(location, ".*Belfast, Northern Ireland", "United Kingdom"),
location = str_replace(location, ".*Ireland", "United Kingdom"),
location = str_replace(location, ".*Hobart, Tasmania", "Australia"),
location = str_replace(location, ".*Dhaka, Bangladesh", "Bangladesh"),
location = str_replace(location, ".*Nairobi, Kenya", "Kenya"),
location = str_replace(location, ".*Dhaka, Bangladesh", "Bangladesh"),
location = str_replace(location, ".*Berlin, Deutschland", "Germany"),
location = str_replace(location, ".*Munich, Bavaria", "Germany"),
location = str_replace(location, ".*Dehradun, India", "India"),
location = str_replace(location, ".*Bengaluru, India", "India"),
location = str_replace(location, ".*Berlin, Deutschland", "Germany"),
location = str_replace(location, ".*Deutschland", "Germany"),
location = str_replace(location, ".*Edinburgh", "United Kingdom"),
location = str_replace(location, ".*Glasgow", "United Kingdom"),
location = str_replace(location, ".*New York, USA", "United States of America"),
location = str_replace(location, ".*Washington, USA", "United States of America"),
location = str_replace(location, ".*California", "United States of America"),
location = str_replace(location, ".*California", "United States of America"),
location = str_replace(location, ".*Christchurch City, New Zealand", "New Zealand"),
location = str_replace(location, ".*Harare, Zimbabwe", "Zimbabwe"),
location = str_replace(location, ".*Islamabad, Pakistan", "Pakistan"),
location = str_replace(location, ".*Kolkata, India", "India"),
location = str_replace(location, ".*Lagos, Nigeria", "Nigeria"),
location = str_replace(location, ".*Lima, Peru", "Peru"),
location = str_replace(location, ".*Valparaíso, Chile", "Chile"),
location = str_replace(location, ".*Uppsala, Sweden", "Sweden"),
location = str_replace(location, ".*Uppsala, Sverige", "Sweden"),
location = str_replace(location, ".*Stockholm, Sweden", "Sweden"),
location = str_replace(location, ".*Stockholm", "Sweden"),
location = str_replace(location, ".*Turin, Piedmont", "Italy"),
location = str_replace(location, ".*University of Iceland", "Iceland"),
location = str_replace(location, ".*University of Helsinki", "Finland"),
location = str_replace(location, ".*Tucumán, Argentina", "Argentina"),
location = str_replace(location, ".*The Hague, The Netherlands", "The Netherlands"),
location = str_replace(location, ".*UK", "United Kingdom")
) %>%
count(location, sort = TRUE) %>%
#slice(-40) %>%
head(n = 45) -> followers_by_country
followers_by_country
## location n
## 1 1273
## 2 United States of America 606
## 3 France 571
## 4 United Kingdom 473
## 5 Canada 166
## 6 Germany 130
## 7 Australia 112
## 8 Norway 50
## 9 Switzerland 43
## 10 Spain 34
## 11 India 33
## 12 South Africa 32
## 13 Brazil 30
## 14 Portugal 27
## 15 Finland 26
## 16 Sweden 18
## 17 Belgium 17
## 18 Italy 16
## 19 The Netherlands 14
## 20 New Zealand 13
## 21 Peru 13
## 22 Denmark 11
## 23 Mexico 11
## 24 Nepal 10
## 25 Argentina 9
## 26 Kenya 9
## 27 Bangladesh 8
## 28 Chile 8
## 29 Greece 8
## 30 Colombia 6
## 31 Czech Republic 6
## 32 Singapore 6
## 33 Iceland 5
## 34 Senegal 5
## 35 Iran 4
## 36 Nigeria 4
## 37 Pakistan 4
## 38 Tanzania 4
## 39 Zimbabwe 4
## 40 3
## 41 Austria 3
## 42 Ecuador 3
## 43 Guatemala 3
## 44 Hong Kong 3
## 45 Luxembourg 3
Let’s have a look to the same info on a quick and dirty map. White countries do not appear in the 45 countries with most followers, or do not have any followers.
library(rworldmap)
library(classInt)
spdf <- joinCountryData2Map(followers_by_country,
joinCode="NAME",
nameJoinColumn="location",
verbose=TRUE)
## 43 codes from your data successfully matched countries in the map
## 2 codes from your data failed to match with a country code in the map
## failedCodes failedCountries
## [1,] "" ""
## [2,] "" " "
## 200 codes from the map weren't represented in your data
classInt <- classIntervals(spdf$n,
n=9,
style = "jenks")
catMethod <- classInt[["brks"]]
library(RColorBrewer)
colourPalette <- brewer.pal(9,'RdPu')
mapParams <- mapCountryData(spdf,
nameColumnToPlot="n",
addLegend=FALSE,
catMethod = catMethod,
colourPalette=colourPalette,
mapTitle="Number of followers per country")
do.call(addMapLegend,
c(mapParams,
legendLabels="all",
legendWidth=0.5,
legendIntervals="data",
legendMar = 2))
How many followers are verified?
details_followers %>%
count(verified)
## verified n
## 1 FALSE 5009
## 2 TRUE 16
Who are the verified users? Scientific entities, but also some scientists.
details_followers %>%
filter(verified==TRUE) %>%
select(name, screen_name, location, followers_count) %>%
arrange(-followers_count)
## name screen_name
## 1 Çağrı Mert Bakırcı cagrimbakirci
## 2 Elisabeth Bik MicrobiomDigest
## 3 ipbes IPBES
## 4 Dr. Earyn McGee, Lizard lassoer \U0001f98e Afro_Herper
## 5 figshare figshare
## 6 Nathan Peiffer-Smadja nathanpsmad
## 7 Wildlife Alliance WildlifeRescue
## 8 Arthur Charpentier freakonometrics
## 9 L'écologie au CNRS INEE_CNRS
## 10 The Royal Botanic Garden Sydney RBGSydney
## 11 CREAF CREAF_ecologia
## 12 Jean-Paul Carrera JeanCarrPaul
## 13 Deckset decksetapp
## 14 𝙳𝚞𝚌-𝚀𝚞𝚊𝚗𝚐 𝙽𝚐𝚞𝚢𝚎𝚗 duc_qn
## 15 Christophe Josset \U0001f4dd CJosset
## 16 Casey Fung TheCaseyFung
## location followers_count
## 1 Texas, USA 170164
## 2 San Francisco, CA 105825
## 3 Bonn, Germany 78086
## 4 United States 45215
## 5 London 40610
## 6 Paris - London 36760
## 7 Phnom Penh, Cambodia 34843
## 8 Montréal, Québec 28038
## 9 23483
## 10 Sydney, New South Wales 20259
## 11 Bellaterra (Barcelona) 16654
## 12 Oxford, England 5744
## 13 Your Mac 3884
## 14 Lausanne, Switzerland 2808
## 15 Paris, France 2002
## 16 Bundjalung Country | Mullum 1612
How old are my followers, where age is defined as the time elapsed since they created their Twitter account. Twitter was created in 2006. I created my account in July 2016 but started using it only two or three years ago I’d say. This dataviz is not too much informative I guess.
details_followers %>%
mutate(month = created_at %>% str_split(" ") %>% map_chr(2),
day = created_at %>% str_split(" ") %>% map_chr(3),
year = created_at %>% str_split(" ") %>% map_chr(6),
created_at = paste0(month,"-",year),
created_at = lubridate::my(created_at)) %>%
group_by(created_at) %>%
summarise(counts = n()) %>%
ggplot(aes(x = created_at, y = counts)) +
geom_line() +
labs(x = "date of creation", y = "how many accounts were created this month-year")
What is the distribution of the number of followers and friends. On average, 400 followers and 400 friends.
foll <- details_followers %>%
ggplot() +
aes(x = log(followers_count)) +
geom_histogram(fill="#69b3a2", color="#e9ecef", alpha=0.9) +
labs(x = "log number of followers", y = "counts", title = "followers")
friends <- details_followers %>%
ggplot() +
aes(x = log(friends_count)) +
geom_histogram(fill="#69b3a2", color="#e9ecef", alpha=0.9) +
labs(x = "log number of friends", y = "", title = "friends")
library(patchwork)
foll + friends
Retrieving social links
To investigate the social network made of my followers, we need to gather who they follow, and their followers. Unfortunately, the Twitter API imposes some time and quantity constraints in retrieving data.
Here, for the sake of illustration, I will focus on some accounts only. To do so, I consider accounts with between 1000 and 1500 followers, and more than 30000 tweets users have liked in their account’s lifetime (which denotes some activity). The more followers we consider, the more time it would take to gather info on them: it is approx one minute per follower on average, so one hour for sixty followers, and more than eighty hours for all my followers! With the filters I’m using, I end up with 35 followers here, this is very few to say infer meaningful about a network, but it’ll do the job for now.
some_followers <- details_followers %>%
filter((followers_count > 1000 & followers_count < 1500),
favourites_count > 30000) # %>% nrow()
Create empty list and name it after their screen name.
foler <- vector(mode = 'list', length = length(some_followers$screen_name))
names(foler) <- some_followers$screen_name
Get followers of these selected followers. Takes ages, so save and load later.
for (i in 1:length(some_followers$screen_name)) {
message("Getting followers for user #", i, "/", nrow(some_followers))
foler[[i]] <- get_followers(some_followers$screen_name[i],
n = some_followers$followers_count[i],
retryonratelimit = TRUE)
if(i %% 5 == 0){
message("sleep for 5 minutes")
Sys.sleep(5*60)
}
}
save(foler, file = "foler.RData")
Format the data.
load("dat/foler.RData")
folerx <- bind_rows(foler, .id = "screen_name")
active_fol_x <- some_followers %>% select(id_str, screen_name)
foler_join <- left_join(folerx, some_followers, by = "screen_name")
algo_follower <- foler_join %>%
select(id_str, screen_name) %>%
setNames(c("follower", "active_user")) %>%
na.omit()
Get friends of my followers. Takes ages. Again I save and load later.
friend <- data.frame()
for (i in seq_along(some_followers$screen_name)) {
message("Getting following for user #", i ,"/",nrow(some_followers))
kk <- get_friends(some_followers$screen_name[i],
n = some_followers$friends_count[i],
retryonratelimit = TRUE)
friend <- rbind(friend, kk)
if(i %% 15 == 0){
message("sleep for 15 minutes")
Sys.sleep(15*60+1)
}
}
save(friend, file = "friend.RData")
Format the data.
load("dat/friend.RData")
all_friend <- friend %>% setNames(c("screen_name", "user_id"))
all_friendx <- left_join(all_friend, active_fol_x, by="screen_name")
algo_friend <- all_friendx %>% select(user_id, screen_name) %>%
setNames(c("following","active_user"))
Now that we have all info on followers and friends, we’re gonna build the network of people who follow each other.
un_active <- unique(algo_friend$active_user) %>%
data.frame(stringsAsFactors = F) %>%
setNames("active_user")
algo_mutual <- data.frame()
for (i in seq_along(un_active$active_user)){
aa <- algo_friend %>%
filter(active_user == un_active$active_user[i])
bb <- aa %>% filter(aa$following %in% algo_follower$follower) %>%
setNames(c("mutual","active_user"))
algo_mutual <- rbind(algo_mutual,bb)
}
Instead of ids, we use screen names.
detail_friend <- lookup_users(algo_mutual$mutual)
algo_mutual <- algo_mutual %>%
left_join(detail_friend, by = c("mutual" = "id_str")) %>%
na.omit() %>%
select(mutual, active_user, screen_name)
algo_mutual
## # A tibble: 18 × 3
## mutual active_user screen_name
## <chr> <chr> <chr>
## 1 824265743575306243 g33k5p34k robbie_emmet
## 2 761735501707567104 g33k5p34k LauraBlissEco
## 3 1439272374 alexnicolharper LeafyEricScott
## 4 824265743575306243 alexnicolharper robbie_emmet
## 5 21467726 alexnicolharper g33k5p34k
## 6 4181949437 LegalizeBrain EvpokPadding
## 7 824265743575306243 Souzam7139 robbie_emmet
## 8 888890201673580548 Souzam7139 mellenmartin
## 9 824265743575306243 mellenmartin robbie_emmet
## 10 21467726 LauraBlissEco g33k5p34k
## 11 831398755920445440 JosiahParry RoelandtN42
## 12 4181949437 gau EvpokPadding
## 13 21467726 robbie_emmet g33k5p34k
## 14 4159201575 robbie_emmet alexnicolharper
## 15 888890201673580548 robbie_emmet mellenmartin
## 16 29538964 EvpokPadding gau
## 17 1439272374 lifedispersing LeafyEricScott
## 18 301687349 RoelandtN42 JosiahParry
Add my account to the network.
un_active <- un_active %>%
mutate(mutual = rep("oaggimenez"))
un_active <- un_active[,c(2,1)]
un_active <- un_active %>%
setNames(c("active_user","screen_name"))
algo_mutual <- bind_rows(algo_mutual %>% select(-mutual), un_active)
For what follows, we will need packages to work with networks.
library(igraph)
library(tidygraph)
library(ggraph)
Now we create the edges, nodes and build the network.
nodes <- data.frame(V = unique(c(algo_mutual$screen_name,algo_mutual$active_user)),
stringsAsFactors = F)
edges <- algo_mutual %>%
setNames(c("from","to"))
network_ego1 <- graph_from_data_frame(d = edges, vertices = nodes, directed = F) %>%
as_tbl_graph()
Network metrics
Create communities using group_louvain() algorithm, and calculate standard metrics using tidygraph.
set.seed(123)
network_ego1 <- network_ego1 %>%
activate(nodes) %>%
mutate(community = as.factor(group_louvain())) %>%
mutate(degree_c = centrality_degree()) %>%
mutate(betweenness_c = centrality_betweenness(directed = F,normalized = T)) %>%
mutate(closeness_c = centrality_closeness(normalized = T)) %>%
mutate(eigen = centrality_eigen(directed = F))
network_ego_df <- as.data.frame(network_ego1)
Identify key nodes with respect to network metrics.
network_ego_df
## name community degree_c betweenness_c closeness_c eigen
## 1 robbie_emmet 2 8 0.0049299720 0.5384615 0.5238676
## 2 LauraBlissEco 2 3 0.0000000000 0.5147059 0.2717415
## 3 LeafyEricScott 5 3 0.0008403361 0.5223881 0.2287709
## 4 g33k5p34k 2 6 0.0024649860 0.5303030 0.4342056
## 5 EvpokPadding 3 4 0.0011204482 0.5223881 0.2336052
## 6 mellenmartin 2 4 0.0000000000 0.5223881 0.3371889
## 7 RoelandtN42 4 3 0.0000000000 0.5147059 0.2050991
## 8 alexnicolharper 2 5 0.0019607843 0.5303030 0.3942456
## 9 gau 3 3 0.0000000000 0.5147059 0.2133909
## 10 JosiahParry 4 3 0.0000000000 0.5147059 0.2050991
## 11 genius_c137 1 1 0.0000000000 0.5072464 0.1454399
## 12 Zjbb 1 1 0.0000000000 0.5072464 0.1454399
## 13 SJRAfloat 1 1 0.0000000000 0.5072464 0.1454399
## 14 BMPARMA17622540 1 1 0.0000000000 0.5072464 0.1454399
## 15 maryam_adeli 1 1 0.0000000000 0.5072464 0.1454399
## 16 waywardaf 1 1 0.0000000000 0.5072464 0.1454399
## 17 SusyVF6 1 1 0.0000000000 0.5072464 0.1454399
## 18 Sinalo_NM 1 1 0.0000000000 0.5072464 0.1454399
## 19 Leila_Lula 1 1 0.0000000000 0.5072464 0.1454399
## 20 CARThorpe 1 1 0.0000000000 0.5072464 0.1454399
## 21 j_wilson_white 1 1 0.0000000000 0.5072464 0.1454399
## 22 LegalizeBrain 3 2 0.0000000000 0.5147059 0.1794154
## 23 JaishriJuice 1 1 0.0000000000 0.5072464 0.1454399
## 24 jaguaretepy 1 1 0.0000000000 0.5072464 0.1454399
## 25 Souzam7139 2 3 0.0000000000 0.5223881 0.2706719
## 26 LaLince_ 1 1 0.0000000000 0.5072464 0.1454399
## 27 Nina_Ella_ 1 1 0.0000000000 0.5072464 0.1454399
## 28 samcox 1 1 0.0000000000 0.5072464 0.1454399
## 29 TAdamsBio42 1 1 0.0000000000 0.5072464 0.1454399
## 30 brubakerl 1 1 0.0000000000 0.5072464 0.1454399
## 31 robanhk 1 1 0.0000000000 0.5072464 0.1454399
## 32 InesCCarv 1 1 0.0000000000 0.5072464 0.1454399
## 33 groundhog0202 1 1 0.0000000000 0.5072464 0.1454399
## 34 zenmart 1 1 0.0000000000 0.5072464 0.1454399
## 35 lifedispersing 5 2 0.0000000000 0.5147059 0.1787123
## 36 oaggimenez 1 35 0.9685154062 1.0000000 1.0000000
Get users with highest values of each metric.
kp_ego <- data.frame(
network_ego_df %>% arrange(-degree_c) %>% select(name),
network_ego_df %>% arrange(-betweenness_c) %>% select(name),
network_ego_df %>% arrange(-closeness_c) %>% select(name),
network_ego_df %>% arrange(-eigen) %>% select(name)) %>%
setNames(c("degree","betweenness","closeness","eigen"))
kp_ego[-1,]
## degree betweenness closeness eigen
## 2 robbie_emmet robbie_emmet robbie_emmet robbie_emmet
## 3 g33k5p34k g33k5p34k g33k5p34k g33k5p34k
## 4 alexnicolharper alexnicolharper alexnicolharper alexnicolharper
## 5 EvpokPadding EvpokPadding LeafyEricScott mellenmartin
## 6 mellenmartin LeafyEricScott EvpokPadding LauraBlissEco
## 7 LauraBlissEco LauraBlissEco mellenmartin Souzam7139
## 8 LeafyEricScott mellenmartin Souzam7139 EvpokPadding
## 9 RoelandtN42 RoelandtN42 LauraBlissEco LeafyEricScott
## 10 gau gau RoelandtN42 gau
## 11 JosiahParry JosiahParry gau RoelandtN42
## 12 Souzam7139 genius_c137 JosiahParry JosiahParry
## 13 LegalizeBrain Zjbb LegalizeBrain LegalizeBrain
## 14 lifedispersing SJRAfloat lifedispersing lifedispersing
## 15 genius_c137 BMPARMA17622540 genius_c137 robanhk
## 16 Zjbb maryam_adeli Zjbb Sinalo_NM
## 17 SJRAfloat waywardaf SJRAfloat samcox
## 18 BMPARMA17622540 SusyVF6 BMPARMA17622540 genius_c137
## 19 maryam_adeli Sinalo_NM maryam_adeli Zjbb
## 20 waywardaf Leila_Lula waywardaf SJRAfloat
## 21 SusyVF6 CARThorpe SusyVF6 BMPARMA17622540
## 22 Sinalo_NM j_wilson_white Sinalo_NM waywardaf
## 23 Leila_Lula LegalizeBrain Leila_Lula SusyVF6
## 24 CARThorpe JaishriJuice CARThorpe Leila_Lula
## 25 j_wilson_white jaguaretepy j_wilson_white j_wilson_white
## 26 JaishriJuice Souzam7139 JaishriJuice JaishriJuice
## 27 jaguaretepy LaLince_ jaguaretepy jaguaretepy
## 28 LaLince_ Nina_Ella_ LaLince_ LaLince_
## 29 Nina_Ella_ samcox Nina_Ella_ Nina_Ella_
## 30 samcox TAdamsBio42 samcox brubakerl
## 31 TAdamsBio42 brubakerl TAdamsBio42 groundhog0202
## 32 brubakerl robanhk brubakerl zenmart
## 33 robanhk InesCCarv robanhk maryam_adeli
## 34 InesCCarv groundhog0202 InesCCarv CARThorpe
## 35 groundhog0202 zenmart groundhog0202 InesCCarv
## 36 zenmart lifedispersing zenmart TAdamsBio42
Robbie Emmet has the highest degree, betweenness, closeness centrality and eigenvector centrality. He has the most relations with the other nodes in the network. He also can spread information the further away and faster than anyone, and is also surrounded by important persons in the network. This is within the network we’ve just built.
By the way, Robbie has just passed his dissertation defense, congrats! Follow him on Twitter at @robbie_emmet, he is awesome!
Visualize Network
We visualize the network using clusters or communities (first three only).
options(ggrepel.max.overlaps = Inf)
network_viz <- network_ego1 %>%
filter(community %in% 1:3) %>%
mutate(node_size = ifelse(degree_c >= 50,degree_c,0)) %>%
mutate(node_label = ifelse(betweenness_c >= 0.01,name,NA))
plot_ego <- network_viz %>%
ggraph(layout = "stress") +
geom_edge_fan(alpha = 0.05) +
geom_node_point(aes(color = as.factor(community),size = node_size)) +
geom_node_label(aes(label = node_label),nudge_y = 0.1,
show.legend = F, fontface = "bold", fill = "#ffffff66") +
theme_graph() +
theme(legend.position = "none") +
labs(title = "Top 3 communities")
plot_ego
Let’s use an alternative dataviz to identify the nodes.
network_viz %>%
mutate(community = group_spinglass()) %>%
ggraph(layout = "nicely") +
geom_edge_fan(alpha = 0.25) +
geom_node_point(aes(color = factor(community)),size = 5, show.legend = F) +
geom_node_text(aes(label = name),repel = T) +
theme_graph() + theme(legend.position = "none") +
labs(title = "Network with named nodes")
Concluding words
Not sure what the communities are about. Due to limitation with the Twitter API I did not attempt to retrieve the info on all the 5,000 followers, but I used only 35 of them, which explains why there is not much to say about this network I guess. I might take some time to try and retrieve all the info (will take days), or not. I might also use some information I already have, like how followers describe themselves and do some topic modelling to identify common themes of interest.
Here I have simply followed the steps that Joe Cristian illustrated so brillantly in his post at https://algotech.netlify.app/blog/social-network-analysis-in-r/. Make sure you read his post for more details on the code and theory. In particular, he illustrates how information spreads through the network.
For more about social network analyses and Twitter, see this thread https://twitter.com/Mehdi_Moussaid/status/1389174715990876160?s=20 and this video https://www.youtube.com/watch?v=UX7YQ6m2r_o in French, this is what inspired me to do the analysis. See also this https://github.com/eleurent/twitter-graph for a much better analysis than what I could do (with Python for data retrieving and manipulation, and Gephi for network visualisation).
-
This post was also published on https://www.r-bloggers.com. ↩︎
