PS5-Xbox sentiment analysis
Amos Anzele
12/3/2020
Boasting 2 billion monthly active users, YouTube us the second highest visited sire with 2 billion active monthly users. At 335 million monthly active users, Twitter is also one of the most used social networking platforms in the world. With the release of the Playstation 5 and Xbox series X, I wanted to leverage the interactions from both these platforms to understand the reception each console is getting. From my the people I interact with, the Playstation console always seems to have the most favorable reviews compared the Xbox.
Link to my Github for datasets
Required packages:
Data collected from Twitter and Youtube (YouTube video comparing both products from popular YouTuber):
xbox_tw<- read_csv("xboxTweets.csv")## Warning: Missing column names filled in: 'X1' [1]## Parsed with column specification:
## cols(
## X1 = col_double(),
## text = col_character(),
## favorited = col_logical(),
## favoriteCount = col_double(),
## replyToSN = col_character(),
## created = col_datetime(format = ""),
## truncated = col_logical(),
## replyToSID = col_double(),
## id = col_double(),
## replyToUID = col_double(),
## statusSource = col_character(),
## screenName = col_character(),
## retweetCount = col_double(),
## isRetweet = col_logical(),
## retweeted = col_logical(),
## longitude = col_logical(),
## latitude = col_logical()
## )ps_tw<- read_csv("ps5Tweets.csv")## Warning: Missing column names filled in: 'X1' [1]## Parsed with column specification:
## cols(
## X1 = col_double(),
## text = col_character(),
## favorited = col_logical(),
## favoriteCount = col_double(),
## replyToSN = col_character(),
## created = col_datetime(format = ""),
## truncated = col_logical(),
## replyToSID = col_double(),
## id = col_double(),
## replyToUID = col_double(),
## statusSource = col_character(),
## screenName = col_character(),
## retweetCount = col_double(),
## isRetweet = col_logical(),
## retweeted = col_logical(),
## longitude = col_logical(),
## latitude = col_logical()
## )yt <- read_csv("ytcomments.csv")## Warning: Missing column names filled in: 'X1' [1]## Parsed with column specification:
## cols(
## X1 = col_double(),
## videoId = col_character(),
## textDisplay = col_character(),
## textOriginal = col_character(),
## authorDisplayName = col_character(),
## authorProfileImageUrl = col_character(),
## authorChannelUrl = col_character(),
## authorChannelId.value = col_character(),
## canRate = col_logical(),
## viewerRating = col_character(),
## likeCount = col_double(),
## publishedAt = col_datetime(format = ""),
## updatedAt = col_datetime(format = ""),
## id = col_character(),
## parentId = col_character(),
## moderationStatus = col_logical()
## )Exploring the data: - Both Twitter datasets have 17 variables, while the YouTube data has 16, however the YouTube comments on this video is about three times the size.
head(xbox_tw)## # A tibble: 6 x 17
## X1 text favorited favoriteCount replyToSN created truncated
## <dbl> <chr> <lgl> <dbl> <chr> <dttm> <lgl>
## 1 1 "RT ~ FALSE 0 <NA> 2020-12-03 05:30:13 FALSE
## 2 2 "It'~ FALSE 0 <NA> 2020-12-03 05:30:13 FALSE
## 3 3 "RT ~ FALSE 0 <NA> 2020-12-03 05:30:12 FALSE
## 4 4 "RT ~ FALSE 0 <NA> 2020-12-03 05:30:09 FALSE
## 5 5 "RT ~ FALSE 0 <NA> 2020-12-03 05:30:09 FALSE
## 6 6 "I g~ FALSE 0 <NA> 2020-12-03 05:30:07 FALSE
## # ... with 10 more variables: replyToSID <dbl>, id <dbl>, replyToUID <dbl>,
## # statusSource <chr>, screenName <chr>, retweetCount <dbl>, isRetweet <lgl>,
## # retweeted <lgl>, longitude <lgl>, latitude <lgl>head(ps_tw)## # A tibble: 6 x 17
## X1 text favorited favoriteCount replyToSN created truncated
## <dbl> <chr> <lgl> <dbl> <chr> <dttm> <lgl>
## 1 1 "Whe~ FALSE 0 <NA> 2020-12-03 05:36:42 FALSE
## 2 2 "RT ~ FALSE 0 <NA> 2020-12-03 05:36:41 FALSE
## 3 3 "RT ~ FALSE 0 <NA> 2020-12-03 05:36:40 FALSE
## 4 4 "@AB~ FALSE 0 ABC 2020-12-03 05:36:37 TRUE
## 5 5 "RT ~ FALSE 0 <NA> 2020-12-03 05:36:37 FALSE
## 6 6 "Jap~ FALSE 0 <NA> 2020-12-03 05:36:37 FALSE
## # ... with 10 more variables: replyToSID <dbl>, id <dbl>, replyToUID <dbl>,
## # statusSource <chr>, screenName <chr>, retweetCount <dbl>, isRetweet <lgl>,
## # retweeted <lgl>, longitude <lgl>, latitude <lgl>head(yt)## # A tibble: 6 x 16
## X1 videoId textDisplay textOriginal authorDisplayNa~ authorProfileIm~
## <dbl> <chr> <chr> <chr> <chr> <chr>
## 1 1 Jq-ODz~ "Which one~ "Which one ~ Unbox Therapy https://yt3.ggp~
## 2 2 Jq-ODz~ "Way to mu~ "Way to muc~ Damian Woodward https://yt3.ggp~
## 3 3 Jq-ODz~ "POV: You ~ "POV: You w~ sjayel518 https://yt3.ggp~
## 4 4 Jq-ODz~ "This is s~ "This is so~ Mojiece https://yt3.ggp~
## 5 5 Jq-ODz~ "Until the~ "Until the ~ Mike Tran https://yt3.ggp~
## 6 6 Jq-ODz~ "( No fans~ "( No fans ~ July Amar https://yt3.ggp~
## # ... with 10 more variables: authorChannelUrl <chr>,
## # authorChannelId.value <chr>, canRate <lgl>, viewerRating <chr>,
## # likeCount <dbl>, publishedAt <dttm>, updatedAt <dttm>, id <chr>,
## # parentId <chr>, moderationStatus <lgl>dim(xbox_tw)## [1] 5000 17dim(ps_tw)## [1] 5000 17dim(yt)## [1] 15468 16names(xbox_tw)## [1] "X1" "text" "favorited" "favoriteCount"
## [5] "replyToSN" "created" "truncated" "replyToSID"
## [9] "id" "replyToUID" "statusSource" "screenName"
## [13] "retweetCount" "isRetweet" "retweeted" "longitude"
## [17] "latitude"names(ps_tw)## [1] "X1" "text" "favorited" "favoriteCount"
## [5] "replyToSN" "created" "truncated" "replyToSID"
## [9] "id" "replyToUID" "statusSource" "screenName"
## [13] "retweetCount" "isRetweet" "retweeted" "longitude"
## [17] "latitude"names(yt)## [1] "X1" "videoId" "textDisplay"
## [4] "textOriginal" "authorDisplayName" "authorProfileImageUrl"
## [7] "authorChannelUrl" "authorChannelId.value" "canRate"
## [10] "viewerRating" "likeCount" "publishedAt"
## [13] "updatedAt" "id" "parentId"
## [16] "moderationStatus"5 descriptives from the dataset
#Cleaning & seeing top words for xbox tweets
'%!in%' <- function(x,y)!('%in%'(x,y))
#Top 20 words for Xbox tweets
xbox_tw %>%
select(text,created) %>%
unnest_tokens(word,text) %>%
anti_join(stop_words) %>%
filter(word %!in% c("https","t.co","rt","4","12","500","rangermj","person","tweet")) %>%
count(word,sort=TRUE) %>%
mutate(word=fct_reorder(word,n)) %>%
slice(1:20) %>%
ggplot(aes(y=word,x=n))+
geom_col()+
labs(title="Top words in Xbox tweets",
y="",
x="")+
theme_light()## Joining, by = "word"
#Cleaning & seeing top words for PS5 tweets
#Top 20 words for Xbox tweets
ps_tw %>%
select(text,created) %>%
unnest_tokens(word,text) %>%
anti_join(stop_words) %>%
filter(word %!in% c("https","t.co","rt","4","12","500","leanandcuisine","glock","pay5n7yiup","1000","fe0f","aint")) %>%
count(word,sort=TRUE) %>%
mutate(word=fct_reorder(word,n)) %>%
slice(1:20) %>%
ggplot(aes(y=word,x=n))+
geom_col()+
labs(title="Top words in PS5 tweets",
y="",
x="")+
theme_light()## Joining, by = "word"
#Cleaning & seeing top words for Youtube comments
#Top 20 words from youtube tweets
yt %>%
select(textOriginal) %>%
unnest_tokens(word,textOriginal) %>%
anti_join(stop_words) %>%
filter(word %!in% c("https","t.co","rt","4","12","500","2")) %>%
count(word,sort=TRUE) %>%
mutate(word=fct_reorder(word,n)) %>%
slice(1:20) %>%
ggplot(aes(y=word,x=n))+
geom_col()+
labs(title="Top words in Xbox tweets",
y="",
x="")+
theme_light()## Joining, by = "word"
Tweets by time for Xbox
xbox_tw %>%
mutate(created=floor_date(created, unit="2 minute")) %>%
count(created) %>%
ggplot(aes(x=created,y=n))+
geom_line()+
labs(title="Tweet trend of Xbox tweets on twitter",
y="Count of tweets",
x="Mention Time")+
theme_light()
Tweets by time for PS
ps_tw %>%
mutate(created=floor_date(created, unit="2 minute")) %>%
count(created) %>%
ggplot(aes(x=created,y=n))+
geom_line()+
labs(title="Tweet trend of PS tweets on twitter",
y="Count of tweets",
x="Mention Time")+
theme_light()
Comparisons
#Creating sentiment analysis for xbox & PS tweets & Youtube comments
p1 <- xbox_tw %>%
select(text,created) %>%
unnest_tokens(word,text) %>%
anti_join(stop_words) %>%
inner_join(get_sentiments("nrc")) %>%
count(sentiment) %>%
mutate(sentiment=fct_reorder(sentiment,n)) %>%
ggplot(aes(x=n,y=sentiment,fill=sentiment))+
geom_col(show.legend = FALSE)+
theme_light()+
labs(title="Sentiments from Xbox tweets",
y="Sentiments",
x="Frequency")## Joining, by = "word"
## Joining, by = "word"p2 <- ps_tw %>%
select(text,created) %>%
unnest_tokens(word,text) %>%
anti_join(stop_words) %>%
inner_join(get_sentiments("nrc")) %>%
count(sentiment) %>%
mutate(sentiment=fct_reorder(sentiment,n)) %>%
ggplot(aes(x=n,y=sentiment,fill=sentiment))+
geom_col(show.legend = FALSE)+
theme_light()+
labs(title="Sentiments from PS tweets",
y="Sentiments",
x="Frequency")## Joining, by = "word"
## Joining, by = "word"p3 <- yt %>%
select(textOriginal) %>%
unnest_tokens(word,textOriginal) %>%
anti_join(stop_words) %>%
inner_join(get_sentiments("nrc")) %>%
count(sentiment) %>%
mutate(sentiment=fct_reorder(sentiment,n)) %>%
ggplot(aes(x=n,y=sentiment,fill=sentiment))+
geom_col(show.legend = FALSE)+
theme_light()+
labs(title="Sentiments from Youtube comments",
y="Sentiments",
x="Frequency")## Joining, by = "word"
## Joining, by = "word"grid.arrange(p1,p2,p3)
What are the top positive words from Xbox, PS & YT tweets
p4 <- xbox_tw %>%
select(text,created) %>%
unnest_tokens(word,text) %>%
anti_join(stop_words) %>%
inner_join(get_sentiments("bing")) %>%
count(word,sentiment,sort=TRUE) %>%
filter(sentiment=="positive") %>%
mutate(word=fct_reorder(word,n)) %>%
slice(1:20) %>%
ggplot(aes(x=n,y=word,fill=word))+
geom_col(show.legend = FALSE)+
theme_light()+
labs(title="Top positive words - Xbox",
y="",
x="Frequency")## Joining, by = "word"
## Joining, by = "word"p5 <- ps_tw %>%
select(text,created) %>%
unnest_tokens(word,text) %>%
anti_join(stop_words) %>%
inner_join(get_sentiments("bing")) %>%
count(word,sentiment,sort=TRUE) %>%
filter(sentiment=="positive") %>%
mutate(word=fct_reorder(word,n)) %>%
slice(1:20) %>%
ggplot(aes(x=n,y=word,fill=word))+
geom_col(show.legend = FALSE)+
theme_light()+
labs(title="Top positive words - PS",
y="",
x="Frequency")## Joining, by = "word"
## Joining, by = "word"p6 <- yt %>%
select(textOriginal) %>%
unnest_tokens(word,textOriginal) %>%
anti_join(stop_words) %>%
inner_join(get_sentiments("bing")) %>%
count(word,sentiment,sort=TRUE) %>%
filter(sentiment=="positive") %>%
mutate(word=fct_reorder(word,n)) %>%
slice(1:20) %>%
ggplot(aes(x=n,y=word,fill=word))+
geom_col(show.legend = FALSE)+
theme_light()+
labs(title="Top positive words - Youtube",
y="",
x="Frequency")## Joining, by = "word"
## Joining, by = "word"grid.arrange(p4,p5,p6,nrow=1)
Top negative words from Xbox, PS5 & youtube comments
What are the top positive words from Xbox, PS & YT tweets
p7 <- xbox_tw %>%
select(text,created) %>%
unnest_tokens(word,text) %>%
anti_join(stop_words) %>%
inner_join(get_sentiments("bing")) %>%
count(word,sentiment,sort=TRUE) %>%
filter(sentiment=="negative") %>%
mutate(word=fct_reorder(word,n)) %>%
slice(1:20) %>%
ggplot(aes(x=n,y=word,fill=word))+
geom_col(show.legend = FALSE)+
theme_light()+
labs(title="Top negative words - Xbox",
y="",
x="Frequency")## Joining, by = "word"
## Joining, by = "word"p8 <- ps_tw %>%
select(text,created) %>%
unnest_tokens(word,text) %>%
anti_join(stop_words) %>%
inner_join(get_sentiments("bing")) %>%
count(word,sentiment,sort=TRUE) %>%
filter(sentiment=="negative") %>%
mutate(word=fct_reorder(word,n)) %>%
slice(1:20) %>%
ggplot(aes(x=n,y=word,fill=word))+
geom_col(show.legend = FALSE)+
theme_light()+
labs(title="Top negative words - PS",
y="",
x="Frequency")## Joining, by = "word"
## Joining, by = "word"p9 <- yt %>%
select(textOriginal) %>%
unnest_tokens(word,textOriginal) %>%
anti_join(stop_words) %>%
inner_join(get_sentiments("bing")) %>%
count(word,sentiment,sort=TRUE) %>%
filter(sentiment=="negative") %>%
mutate(word=fct_reorder(word,n)) %>%
slice(1:20) %>%
ggplot(aes(x=n,y=word,fill=word))+
geom_col(show.legend = FALSE)+
theme_light()+
labs(title="Top negative words- youtube",
y="",
x="Frequency")## Joining, by = "word"
## Joining, by = "word"grid.arrange(p7,p8,p9,nrow=1)
Creating wordcloud of words from Xbox, Ps & Youtube data
xbox_topwords <- xbox_tw %>%
select(text,created) %>%
unnest_tokens(word,text) %>%
anti_join(stop_words) %>%
count(word)## Joining, by = "word"ps_topwords <- ps_tw %>%
select(text,created) %>%
unnest_tokens(word,text) %>%
anti_join(stop_words) %>%
count(word)## Joining, by = "word"yt_topwords<- yt %>%
select(textOriginal) %>%
unnest_tokens(word,textOriginal) %>%
anti_join(stop_words) %>%
count(word)## Joining, by = "word"pal <- brewer.pal(8, "Dark2")Word cloud for Xbox
xbox_topwords %>%
filter(word %!in% c("https","t.co","rt","4","12","500","rangermj","person")) %>%
with(wordcloud(word, n, random.order = FALSE, max.words = 50, colors=pal))
Word cloud for PS
ps_topwords %>%
filter(word %!in% c("https","t.co","rt","4","12","500","leanandcuisine","glock","pay5n7yiup","1000","fe0f","aint")) %>%
with(wordcloud(word, n, random.order = FALSE, max.words = 50, colors=pal))
Word cloud for Youtube
yt_topwords %>%
filter(word %!in% c("https","t.co","rt","4","12","500","2")) %>%
with(wordcloud(word, n, random.order = FALSE, max.words = 50, colors=pal))
Topic Modeling in R
xbox_words <- xbox_tw %>%
select(text) %>%
unnest_tokens(word,text) %>%
anti_join(stop_words)## Joining, by = "word"ps_words <- ps_tw %>%
select(text) %>%
unnest_tokens(word,text) %>%
anti_join(stop_words)## Joining, by = "word"yt_words <- yt %>%
select(textOriginal) %>%
unnest_tokens(word,textOriginal) %>%
anti_join(stop_words) ## Joining, by = "word"myCorpus <- Corpus(VectorSource(xbox_words$word))
dtm <- DocumentTermMatrix(myCorpus)
dtm_new <- removeSparseTerms(dtm,sparse = 0.999)
rowTotals <- apply(dtm_new , 1, sum) #Find the sum of words in each Document
dtm.new <- dtm_new[rowTotals> 0, ]
topic_xbox <- topicmodels::LDA(dtm.new, k = 3, control = list(seed = 1234))
x_topic <- tidy(topic_xbox)
x_top_terms <- x_topic %>%
group_by(topic) %>%
top_n(10, beta) %>%
ungroup() %>%
arrange(topic, -beta)
x <- x_top_terms %>%
mutate(term = reorder_within(term, beta, topic)) %>%
ggplot(aes(beta, term, fill = factor(topic))) +
geom_col(show.legend = FALSE) +
facet_wrap(~ topic, scales = "free") +
scale_y_reordered()+
theme_light()+
labs(title="Topics from Xbox tweets")
myCorpus_1 <- Corpus(VectorSource(ps_words$word))
dtm_1 <- DocumentTermMatrix(myCorpus_1)
dtm_new_1 <- removeSparseTerms(dtm_1,sparse = 0.999)
rowTotals <- apply(dtm_new_1 , 1, sum) #Find the sum of words in each Document
dtm.new_1 <- dtm_new_1[rowTotals> 0, ]
topic_ps <- topicmodels::LDA(dtm.new_1, k = 3, control = list(seed = 1234))
ps_topic <- tidy(topic_ps)
ps_top_terms <- ps_topic %>%
group_by(topic) %>%
top_n(10, beta) %>%
ungroup() %>%
arrange(topic, -beta)
ps <- ps_top_terms %>%
mutate(term = reorder_within(term, beta, topic)) %>%
ggplot(aes(beta, term, fill = factor(topic))) +
geom_col(show.legend = FALSE) +
facet_wrap(~ topic, scales = "free") +
scale_y_reordered()+
theme_light()+
labs(title="Topics from PS tweets")
myCorpus_2 <- Corpus(VectorSource(yt_words$word))
dtm_2 <- DocumentTermMatrix(myCorpus_2)
dtm_new_2 <- removeSparseTerms(dtm_2,sparse = 0.999)
rowTotals <- apply(dtm_new_2 , 1, sum) #Find the sum of words in each Document
dtm.new_2 <- dtm_new_2[rowTotals> 0, ]
topic_yt <- topicmodels::LDA(dtm.new_2, k = 3, control = list(seed = 1234))
yt_topic <- tidy(topic_yt)
yt_top_terms <- yt_topic %>%
group_by(topic) %>%
top_n(10, beta) %>%
ungroup() %>%
arrange(topic, -beta)
yt <- yt_top_terms %>%
mutate(term = reorder_within(term, beta, topic)) %>%
ggplot(aes(beta, term, fill = factor(topic))) +
geom_col(show.legend = FALSE) +
facet_wrap(~ topic, scales = "free") +
scale_y_reordered()+
theme_light()+
labs(title="Topics from YT comments")
grid.arrange(x,ps,yt)
Conclusion: My original intent was to understand how these two products were being received, and which one garnered the most positive acclaim. As stated above, I believed the Playstation would come out the clear favorite, but after this analysis, I come out needing to expand my datasets to get a clearer picture. I also need to collect more data from the YouTube comments as there seems to be more engagement.