Sentiment Analysis & Brand Perception of “ONEPLUS” by using Twitter data

An attempt to leverage the strength of data analytics tool like R and Tableau and understanding the brand perception on live twitter data.

Need of Study

• OnePlus 5 was unveiled in India on June 22, and the sales for the smartphone began June 27. 
• Newly Launched phone.
• OnePlus 5 was the bestselling smartphone during the Amazon ‘Prime day’, said the company in a press statement. OnePlus did not give out the number of units sold but claimed that hourly sales of OnePlus 5 saw six times jump between 6 PM on July 10 to midnight of the July 11, compared to average hourly sales on the platform.
• To understand the sentiments of the cstomers on ONEPLUS which help to understand the brand perfromance.

Introduction- ONEPLUS

• OnePlus is a Shenzhen based Chinese smartphone manufacturer founded by Pete Lau (CEO) and Carl Pei in December 2013. It is headquartered in Guangdong. The company serves 42 countries and regions around the world as of March 2016. They have released 6 phones, amongst other products.
• OnePlus was founded on 16 December 2013 by former Oppo vice-president Pete Lauand Carl Pei .
•  According to the Chinese government's documentation, the only institutional stockholder in OnePlus is Oppo. Although Lau initially denied that OnePlus was a wholly owned subsidiary of Oppo, upon release of the regulatory filings they admitted that they are owned by Oppo and are "in talks with other investors" (although nothing has been announced to date).
• The company's main goal was to design a smartphone that would balance high-end quality with a lower price than other phones in its class, believing that users would "Never Settle"  for the lower-quality devices produced by other companies.

Following process is followed to analyse the twitter data:

1. API twitter Login and data grabbing

• Twittter API setup 
• setup_twitter_oauth(api_key, api_secret, access_token, access_token_secret)
• set.seed(123)

• Grab latest tweets-  2500 tweets are extracted, duplicate tweets were deleted at raw level
• tweets_oneplus <- searchTwitter('oneplus', n=2500, lang='en')
• tweets.df <- twListToDF(tweets_oneplus)
• dim(tweets.df)
• tweets.df <- subset(tweets.df, !duplicated(tweets.df$text))

2. Data Mining using Tm package

Cleaning the text data by removing links, tags and delimiters. Build a Corpus, and specify the location to be the character Vectors.

• Remove character string between < >
• tweets.text <- genX(tweets.df$text, " <", ">")
• Create document corpus with tweets text
• myCorpus<- Corpus(VectorSource(tweets.text)) 
• Convert to Lowercase  
• Remove the retweets  
• Remove @people
• Remove the links (URLs)  
• Remove anything except the english language and space  
• Remove Stopwords  
• Remove Single letter words          
• Remove Extra Whitespaces  
• Extracted Unique tweets

3. Cleaned Tweets

 [980] im getting nowhere ticket guys keep blaming issue clear wa                                          

 [981] wish buy                                                                                            

 [982] guess never owned iphone plus                                                                                                   

 [988] oh joy samsung crappy ui get cleaner android maybe nexus oneplus al                             

 [989] thanks now oneplus runs oxygenos however stay tuned us social                                       

 [990] another iphone wannabe launch device cant even design                                                        [995] bro sick oneplus lg review price better youtube                                                     

 [996] honor oneplus best stuff                                                                            

 [997] oneplus international test                                                                          

 [998] dear will fix kannada rendering oneplust phone display language without errors            [999] change icon pack oneplus spice interface tech android ios windows

4.Most Frequent tweets- Using TermDocument Matrix

• After cleaning, no. of tweets reduces to 1071
• <<TermDocumentMatrix (terms: 1960, documents: 1071)>>
• Non-/sparse entries: 7302/2091858
• Sparsity           : 100%
• Maximal term length: 28
• Weighting          : term frequency (tf)
• Code for minimum frequency of words appearing=30
• dtm1 <- TermDocumentMatrix(myCorpus)  

• (freq.terms <- findFreqTerms(dtm1, lowfreq = 30))
• term.freq <- rowSums(as.matrix(dtm1))
• term.freq <- subset(term.freq, term.freq > 30)
• df <- data.frame(term = names(term.freq), freq= term.freq)   

                                                           Figure: Term Frequency Chart

We can understand from the above plot that words like  "oneplus" ,"issu" ,“thank" ,"can" ,"use" ,"best", "flashoneplus, "camera"  are the most tweeted words. Most of them are positive words and people are discussing about features and comparing with brands like oppo, samsung, apple. 

5.Word Cloud-ONEPLUS

Figure: Word Cloud

• Words revolved around our area of interest- oneplus.
• Words like thank, camera, new, best, realli, prime awesome etc are postive words which are tweeted by the people.
• We can also see issue, dont, problem, are negative words.
• One Plus mobile is compareed with samsung, apple,honor.
• Mobile features like GB, Camera, speed, RAM, device,are being discussed.
• Most of them are still looking and asking for feedback, suggestions about the phone

6.Word Correlations- Graphical Analysis

• We can understand from the above plot that BEST is highly correlated with xperia,premium,sony, fan,blackberry, inquire etc which implies ONEPLUS is a competitive brand and people are inquiring about the product.

• Whereas, camera feedback is very good. it is highly correlated with satisfactory and brilliant

7.Sentiment Analysis

• 5000 tweets are pulled
• tweets_oneplus1 <- searchTwitter('oneplus', n=5000, lang='en')
• noof.tweets = c(length(tweets.txt))
• tweets.text = laply(tweets_oneplus1,function(t)t$getText())

• we also need are our lists with the positive and the negative words to differentiate sentiments.We can find them here: https://github.com/mjhea0/twitter-sentiment-analysis/tree/master/wordbanks
• pos = scan('E:/Web analytics/positive words.txt', what='character', comment.char=';')
• neg = scan('E:/Web analytics/negative words.txt', what='character',comment.char=';')
• Now we have to insert a small algoriytm written by Jeffrey Breen analyzing our words and to generate sentiment score
• scores = score.sentiment(tweets.txt, pos, neg)
• scores$positive <- as.numeric(scores$score >0)
• scores$negative <- as.numeric(scores$score <0)
• scores$neutral <- as.numeric(scores$score==0)

• The positive values stand for positive tweets and the negative values for negative tweets.

Sentiment Analysis-Graphical using Tableau

Above chart indicate that the perception of one plus brand is mostly towards the positive and neutral side. We can understand that newly launched one plus is being loved by the people. People are liking the brand and product. 1325: positive, 3261: neutral, 385: negative.

Conclusion

• OnePlus brand perception is on positive side after the launch of OnePlus5 and sale on Amazon
• Features like camera,memory, processors are mostly talked about.
• People are comparing with with brands like Apple, Samsung, Honor.
• Sentiment Analysis helps us to understand what consumers are thinking about the product and services. 
• Moreover, the insights gathered can be implemented to devise marketing strategy.
• Customer feedback can be collected and hence improving the product quality and customer satisfaction.

For any feedback, comments feel free to write.