Getting a Youtube Channel Data in R

Getting Youtube channel id’s

Before we move on, we’ll need every channel id: this can be obtained using external pages like this, just copy and paste the channel link and it will print the channel id right away. After finding all channel id’s we should have a dataframe like this:

Defining methods to extract a channel videos and stats.

We coded the following functions which will help us to extract all information on a channel playlist. We’ll come back later to explain a little bit deeper on this.

API_key="PUT YOUR API KEY HERE"
getstats_video<-function(video_id,API_key){
 
  url=paste0("https://www.googleapis.com/youtube/v3/videos?part=snippet,statistics&id=",video_id,"&key=",API_key)
  result <- fromJSON(txt=url)  
  salida=list()
  return(data.frame(name=result$items$snippet$channelTitle, result$items$statistics,title=result$items$snippet$title,date=result$items$snippet$publishedAt,descrip=result$items$snippet$description))
}

get_playlist_canal<-function(id,API_key,topn=15){
  
  url=paste0('https://www.googleapis.com/youtube/v3/playlistItems?part=contentDetails&playlistId=',id,'&key=',API_key,'&maxResults=',topn)
  
  result=fromJSON(txt=url)
  return(data.frame(result$items$contentDetails))
}

getstats_canal<-function(id,API_key){

url=paste0('https://www.googleapis.com/youtube/v3/channels?part=snippet%2CcontentDetails%2Cstatistics&id=',id,'&key=',API_key)
  
result <- fromJSON(txt=url)   
return(data.frame(name=result$items$snippet$title,result$items$statistics,pl_list_id=result$items$contentDetails$relatedPlaylists))
  
}

getall_channels<-function(ids,API_key,topn=5){
  
  videos=lapply(ids,FUN=get_playlist_canal,API_key=API_key,topn=topn) %>% bind_rows()
  
  stats=lapply(videos[,1],FUN=getstats_video,API_key=API_key)
  stats=bind_rows(stats)
  stats$vid_id=videos[,1]
  return(stats)
}

Statistics per Channel

By using getstats_canal function, we can obtain a channel basic statistics such as total views, comments (if available), count of videos published and so on.

can_st=lapply(comp_data$cha_id,FUN = getstats_canal,API_key=API_key)
can_st=bind_rows(can_st)
can_st$viewCount=as.numeric(can_st$viewCount)
can_st[,1:6] %>% kable() %>%kable_styling()

can_st$viewCount=round(as.numeric(can_st$viewCount)/1000000,2)
p1=can_st %>% ggplot(aes(x=reorder(name,viewCount),y=viewCount,fill=name))+
geom_bar(stat="sum")+guides(size=F)+coord_flip()+scale_fill_manual(values = c("red", "darkblue", "yellow2"))+
geom_text(inherit.aes = T,aes(label=paste(viewCount,"M")),nudge_y =0,angle = 90)+
labs(x="Total Visualizations(Millions)",y="Visualizations",fill="")+
theme(legend.position = "top")+mytheme3
  
  ggplotly(p1,tooltip=c("name","viewCount"))  %>% 
    layout(legend = list(orientation = "h",x = 0.01, y = -0.1,autosize=F))  

Information on Individual Videos per Channel

After using getall_channels on a channel id, it will retrieve all videos from that channel and will call getstats_video to get all information out of every video id. Here we plot dislikeCount but you could try favoriteCount or commentCount. Another parameter one should change is topn, this is the number of videos the function will get from each channel. It’s set to 5 by default.

var_to_see="dislikeCount" #favoriteCount or commentCount

info=getall_channels(ids = can_st$pl_list_id.uploads,API_key = API_key,topn =20)


 datacond=melt(info[,c(1:6,8)],id.vars = c("name","date"))
 datacond$date=as.Date(datacond$date)
 datacond$value=as.numeric(datacond$value)
    
    
ggplot(filter(datacond,variable==var_to_see),aes(x=as.Date(date),y=value,fill=name))+
geom_bar(stat="sum")+labs(x=var_to_see,y="",fill="")+guides(size=FALSE)+scale_fill_manual(values = c("red", "darkblue", "yellow2"))+theme(legend.position = "top")+
  scale_x_date(limits =as.Date(c(as.Date(min(datacond$date)),as.Date(Sys.time()))),date_breaks  ="month",date_labels="%b %y")+theme(legend.position = "top")+mytheme2 

Seems like Banco de Bogotá stopped having dislikes but now Bancolombia is leading this blacklist.

tubeR or Raw Query?

Using tubeR package could be another way of getting information from youtube within R. However if we want to append this analysis to a Shiny/Dash app or an automated script, using the raw query is the best way to go (speed based-statement).