Today, our generation lives in a time where tv shows are being streamed on Netflix and news is breaking online. With the rapid development of computer science and information technology, social media have become prominent parts of life for many people. Through social media, people can connect with each other, share their opinions, and access information they need easily and immediately. Social media has also changed the fashion industry, as well as every other industry completely. It gives companies the ability to directly reach their consumers and enables customers to share their opinions and reviews on fashion brands and trends.
In this project, I will web scrape data in fashion industry from Twitter and make it ready for analysis through data cleaning and data management process. Text mining will be conducted to analyze tweets discussion of clothing and fashion, and provide us insights on people’s opinions on fashion brands and trends.
The data in this project was collected from tweets regarding fashion brands and clothing. To analyze people's opinions on fashion brands and trends, Twitter Search API was used to collect recent past tweets regarding the top 15 clothing brands for late twenties and early thirties, including the following: ASOS, Ann Taylor, ModCloth, J.Crew, UNIQLO, Madewell, Forever 21, COS, Rag & Bone, Tory Burch, Rebecca Minkoff, Kate Spade, Michael Kors, Calvin Klein, and Club Monaco. In addition, real-time tweets regarding fashion and clothing was also collected by using Twitter Streaming API for the purpose of exploratory analysis.
To collect data from Twitter, we'll first need to set up an application at https://apps.twitter.com. This will allow us to connect R to the Twitter stream. You’ll need a name, description, and website. I’ve used my Github address as a website, though I’d imagine that anything works. Once created, click on "modify app permissions" and allow the application to read, write and access direct messages. Lastly, click on the API Keys tab and scroll to the bottom of the page. Under token actions, click on "Create my access token". Four API credentials will be created, including API Key, API Secret, Access Token, and Access Token Secret. We’ll need these access tokens when we fire up R.
To use the Twitter Search API, Streaming API and conduct data analysis and text mining, we need the following packages.
library(twitteR)
library(streamR)
library(RCurl)
library(RJSONIO)
library(stringr)
library(ROAuth)
library(tm)
library(SnowballC)
library(textstem)
library(plyr)
library(dplyr)
library(ggplot2)
library(wordcloud)
library(grid)Once these packages are installed and loaded, we can set up our connection to the Twitter Search API and collect recent posted tweets regarding the top 15 clothing brands for late twenties and early thirties. A function called getTweets was created to get tweets in data.frame format. In addition, the ldply function in package plyr was used to get tweets for all brands in parallel.
api_key <- "Your Twitter API Key"
api_secret <- "Your Twitter API Secret"
access_token <- "Your Twitter Access Token"
access_token_secret <- "Your Twitter Access Token Secret"
setup_twitter_oauth(api_key, api_secret, access_token, access_token_secret)## [1] "Using direct authentication"
#Collect Tweets
brands <- list("ASOS", "AnnTaylor", "ModCloth", "jcrew", "Uniqlo", "madewell", "Forever21", "cosstores", "rag_bone", "toryburch", "RebeccaMinkoff", "katespadeny", "MichaelKors", "CalvinKlein", "ClubMonaco")
getTweets <- function(brand, n = 10000) {
search_string <- paste0("@", brand, " OR ", "#", brand, " OR ", brand)
# get tweets
tweets <- searchTwitter(search_string, since = "2017-05-01", n = n)
# strip retweets
tweets <- strip_retweets(tweets, strip_manual = TRUE, strip_mt = TRUE)
# convert to data.frame
tweetdf <- twListToDF(tweets)
# add brand and return
out <- cbind(brand, tweetdf)
return(out)
}
tweets_by_brand <- ldply(brands, function(brand) getTweets(brand, n = 10000))## Warning in doRppAPICall("search/tweets", n, params = params,
## retryOnRateLimit = retryOnRateLimit, : 10000 tweets were requested but the
## API can only return 520
## Warning in doRppAPICall("search/tweets", n, params = params,
## retryOnRateLimit = retryOnRateLimit, : 10000 tweets were requested but the
## API can only return 1293
## Warning in doRppAPICall("search/tweets", n, params = params,
## retryOnRateLimit = retryOnRateLimit, : 10000 tweets were requested but the
## API can only return 2607
## [1] "Rate limited .... blocking for a minute and retrying up to 119 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 118 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 117 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 116 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 115 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 114 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 113 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 112 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 111 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 110 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 109 times ..."
## Warning in doRppAPICall("search/tweets", n, params = params,
## retryOnRateLimit = retryOnRateLimit, : 10000 tweets were requested but the
## API can only return 1727
## [1] "Rate limited .... blocking for a minute and retrying up to 119 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 118 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 117 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 116 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 115 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 114 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 113 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 112 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 111 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 110 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 109 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 108 times ..."
## Warning in doRppAPICall("search/tweets", n, params = params,
## retryOnRateLimit = retryOnRateLimit, : 10000 tweets were requested but the
## API can only return 158
## Warning in doRppAPICall("search/tweets", n, params = params,
## retryOnRateLimit = retryOnRateLimit, : 10000 tweets were requested but the
## API can only return 5819
## Warning in doRppAPICall("search/tweets", n, params = params,
## retryOnRateLimit = retryOnRateLimit, : 10000 tweets were requested but the
## API can only return 965
## Warning in doRppAPICall("search/tweets", n, params = params,
## retryOnRateLimit = retryOnRateLimit, : 10000 tweets were requested but the
## API can only return 560
## Warning in doRppAPICall("search/tweets", n, params = params,
## retryOnRateLimit = retryOnRateLimit, : 10000 tweets were requested but the
## API can only return 1090
## Warning in doRppAPICall("search/tweets", n, params = params,
## retryOnRateLimit = retryOnRateLimit, : 10000 tweets were requested but the
## API can only return 5146
## [1] "Rate limited .... blocking for a minute and retrying up to 119 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 118 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 117 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 116 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 115 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 114 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 113 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 112 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 111 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 110 times ..."
## [1] "Rate limited .... blocking for a minute and retrying up to 109 times ..."
## Warning in doRppAPICall("search/tweets", n, params = params,
## retryOnRateLimit = retryOnRateLimit, : 10000 tweets were requested but the
## API can only return 4858
## Warning in doRppAPICall("search/tweets", n, params = params,
## retryOnRateLimit = retryOnRateLimit, : 10000 tweets were requested but the
## API can only return 158
The output below shows the dimension of the data frame created (tweets_by_brand) and the first 5 observations in the data frame. we can see that more than 30,000 tweets are collected and there are 17 columns in the data frame.
dim(tweets_by_brand)## [1] 31475 17
head(tweets_by_brand, 5)## brand
## 1 ASOS
## 2 ASOS
## 3 ASOS
## 4 ASOS
## 5 ASOS
## text
## 1 @ASOS_HeretoHelp are you replying to dms, also sent a message on Facebook over 12 hours ago with no reply
## 2 @GaryM16 keep an eye out as is should be sorted soon, thanks for bearing with us
## 3 @missyelliott69 Thanks for the feedback, we'll definitely get this passed over and perhaps this will be brought in in the near future.
## 4 @GaryM16 there's been a slight delay with order confirmations but we're working hard to get this resolved
## 5 Asos: We don’t know if social media is a threat or an enabler https://t.co/H2aMXBg8RQ
## favorited favoriteCount replyToSN created truncated
## 1 FALSE 0 ASOS_HeretoHelp 2017-07-15 00:04:28 FALSE
## 2 FALSE 0 GaryM16 2017-07-15 00:03:27 FALSE
## 3 FALSE 0 missyelliott69 2017-07-15 00:03:24 FALSE
## 4 FALSE 0 GaryM16 2017-07-15 00:03:21 FALSE
## 5 FALSE 0 <NA> 2017-07-15 00:03:13 FALSE
## replyToSID id replyToUID
## 1 <NA> 886013560190586880 29420835
## 2 885962243158028289 886013305382465537 259291327
## 3 885963597440049152 886013289414656000 48795965
## 4 885962243158028289 886013279415488512 259291327
## 5 <NA> 886013245542322177 <NA>
## statusSource
## 1 <a href="http://twitter.com/download/iphone" rel="nofollow">Twitter for iPhone</a>
## 2 <a href="http://www.gnatta.com" rel="nofollow">gnatta.com</a>
## 3 <a href="http://www.gnatta.com" rel="nofollow">gnatta.com</a>
## 4 <a href="http://www.gnatta.com" rel="nofollow">gnatta.com</a>
## 5 <a href="http://www.hootsuite.com" rel="nofollow">Hootsuite</a>
## screenName retweetCount isRetweet retweeted longitude latitude
## 1 adzalexander89 0 FALSE FALSE <NA> <NA>
## 2 ASOS_HeretoHelp 0 FALSE FALSE <NA> <NA>
## 3 ASOS_HeretoHelp 0 FALSE FALSE <NA> <NA>
## 4 ASOS_HeretoHelp 0 FALSE FALSE <NA> <NA>
## 5 profileglue 0 FALSE FALSE <NA> <NA>
We now have a data frame with more than 30,000 observations and 17 variables, which is ready for exploratory data analysis and text mining. While most of the variables are self-explanatory, the following are likely the most important ones:
-
Brand: the clothing brand that the tweet mentioned about
-
Text: the text of the actual tweet
-
Created: the date and timestamp of creation
-
Longitude and Latitude: if available, the longitude and latitude of the tweet.
A clothing brand popularity plot can be created as below. According to the plot, we can find that the number of tweets mentioned ASOS, Forever 21, and Rag & Bone are much higher than the number of tweets regarding other brands. This somewhat suggests that ASOS, Forever 21, and Rag & Bone are the most popular clothing brands among all the 15 clothing brands we discussed in this project.
tweets_count <- data.frame(tweets_by_brand %>% group_by(brand) %>% summarise(tweets_count=n()) %>% arrange(desc(tweets_count)))
library(ggplot2)
ggplot(tweets_count, aes(brand, tweets_count)) + geom_bar(stat="identity", fill="blue") + xlab("Brand") + ylab("Tweets Count") + ggtitle("Clothing Brand Popularity") + theme(axis.text.x=element_text(angle=45, hjust=1), plot.title = element_text(hjust = 0.5))
Before conducting the text mining, we will first need to clean the text which serves to remove extraneous characters, unwanted words, and excessive whitespace, to ensure that words are tallied correctly. For example, we may want to remove URLs, emojis, and special symbols in the tweets. In addition, the tm package in R includes numerous transformations to make the process of text cleaning quick and efficient. The following codes will, respectively, remove URLs, remove special symbol "w/" and "&", remove emojis, create a corpus (each tweet text is a distinct document), lowercase all our terms, remove common stop words in English, remove numbers, remove punctuation, strip extraneous whitespace, and lemmatize terms.
tweets_text <- data.frame(text = tweets_by_brand$text, stringsAsFactors = F)
tweets_text$text <- gsub("http\\S+\\s*", "", tweets_text$text)
tweets_text$text <- gsub("w/", "", tweets_text$text)
tweets_text$text <- gsub("&", "", tweets_text$text)
tweets_text$text <- iconv(tweets_text$text, "latin1", "ASCII", sub="")
text_corpus <- Corpus(DataframeSource(tweets_text))
text_corpus <- tm_map(text_corpus, content_transformer(tolower))
my_stopwords <- c(stopwords(kind = "en"), "im", "youre", "hes", "shes", "its", "were", "theyre", "ive", "youve", "weve", "theyve", "id", "youd", "hed", "shed", "wed", "theyd", "ill", "youll", "hell", "shell", "well", "theyll", "isnt", "arent", "wasnt", "werent", "hasnt", "havent", "hadnt", "doesnt", "dont", "didnt", "wont", "wouldnt", "shant", "shouldnt", "can", "cant", "couldnt", "mustnt", "lets", "thats", "whos", "whats", "heres", "theres", "whens", "wheres", "whys", "hows", "RT")
text_corpus <- tm_map(text_corpus, removeWords, my_stopwords)
text_corpus <- tm_map(text_corpus, removeNumbers)
text_corpus <- tm_map(text_corpus, removePunctuation, preserve_intra_word_dashes = T)
text_corpus <- tm_map(text_corpus, stripWhitespace)
text_corpus <- tm_map(text_corpus, content_transformer(lemmatize_strings))From our cleaned corpus we can then create the document term matrix and view its summary information. Here we can see that the DocumentTermMatrix function has created a matrix for us, stored in the dtm variable, with over 30,000 tweets text as rows and about 29,000 terms as columns.
dtm <- DocumentTermMatrix(text_corpus)
dtm## <<DocumentTermMatrix (documents: 31475, terms: 29561)>>
## Non-/sparse entries: 223811/930208664
## Sparsity : 100%
## Maximal term length: 86
## Weighting : term frequency (tf)
dim(dtm)## [1] 31475 29561
Now that we have created the document term matrix, there are a number of built-in functions to assist in understanding our corpus. First, term frequency tends to be a useful starting point in understanding what terms dominate our corpus. We can create a full list of term frequencies and use the head function to view a number of the top-most terms. Additionally, the findFreqTerms function can be used to assess a particular document term matrix, with the lower bound frequency we desire. Furthermore, word frequency barplot and wordclouds can also be useful visual tool for representing frequencies at a glance.
freq <- sort(colSums(as.matrix(dtm)), decreasing = T)
head(freq, 100)## forever man asos ragnbone uniqlo
## 4515 4016 3765 3041 2272
## michaelkors calvinklein human jcrew via
## 2218 1847 1775 1601 1579
## video asosheretohelp get check youtube
## 1538 1409 1388 1348 1294
## order dress bone now bag
## 1282 1271 1270 1264 1223
## just ebay rag sale madewell
## 1216 1215 1214 1181 1130
## look job new like size
## 1068 1044 1033 1012 968
## love style win official skin
## 896 842 809 808 787
## please shop email hire enter
## 786 777 765 736 711
## thank top hot see retail
## 700 679 673 643 633
## day work katespadeny one good
## 624 615 606 603 588
## fashion summer modcloth numb back
## 585 568 567 563 557
## want add address black take
## 552 526 525 523 521
## leather today shirt careerarc buy
## 511 510 506 481 477
## womens help sorry jeans will
## 476 462 450 444 444
## make brand onsale send wear
## 438 435 435 424 423
## need utme toryburch late hello
## 416 416 413 412 409
## nowplaying return come que clothe
## 407 394 393 392 382
## still know white nwt time
## 366 362 356 352 352
## cute shoe item open hear
## 349 344 343 342 337
## great michael find store anntaylor
## 333 333 330 323 320
findFreqTerms(dtm, lowfreq = 350)## [1] "add" "address" "asos" "asosheretohelp"
## [5] "back" "bag" "black" "bone"
## [9] "brand" "buy" "calvinklein" "careerarc"
## [13] "check" "clothe" "come" "day"
## [17] "dress" "ebay" "email" "enter"
## [21] "fashion" "forever" "get" "good"
## [25] "hello" "help" "hire" "hot"
## [29] "human" "jcrew" "jeans" "job"
## [33] "just" "katespadeny" "know" "late"
## [37] "leather" "like" "look" "love"
## [41] "madewell" "make" "man" "michaelkors"
## [45] "modcloth" "need" "new" "now"
## [49] "nowplaying" "numb" "nwt" "official"
## [53] "one" "onsale" "order" "please"
## [57] "que" "rag" "ragnbone" "retail"
## [61] "return" "sale" "see" "send"
## [65] "shirt" "shop" "size" "skin"
## [69] "sorry" "still" "style" "summer"
## [73] "take" "thank" "time" "today"
## [77] "top" "toryburch" "uniqlo" "utme"
## [81] "via" "video" "want" "wear"
## [85] "white" "will" "win" "womens"
## [89] "work" "youtube"
word_freq = data.frame(Term=names(freq), Count=freq)
ggplot(subset(word_freq, freq > 800), aes(Term, Count)) + geom_bar(stat="identity") + xlab("Term") + ylab("Frequency") + ggtitle("Word Frequency") + theme(axis.text.x=element_text(angle=45, hjust=1), plot.title = element_text(hjust = 0.5))
set.seed(1234)
wordcloud(names(freq), freq, min.freq = 420, random.order = FALSE, colors=brewer.pal(8, "Accent"))
From the results and plots above, we can find that the most frequet words in our collected tweets include clothing brand names and some other interesting words such as man, bag, black, leather, shirt, jeans, shoe, etc. These frequent words can give us a sense of what people are talking about regarding fashion and clothing recently.
Now that we have the most frequently occurring terms in hand, we can check for correlations between some of these and other terms that occur in the corpus. In this context, correlation is a quantitative measure of the co-occurrence of words in tweets we collected. The tm package provides the findAssocs function to do this. In addition, it can be useful to visually represent term correlations.
findAssocs(dtm, term = "asos", 0.10)## $asos
## onsale curve midi dress report commande maternity
## 0.25 0.15 0.15 0.14 0.14 0.12 0.12
## mph sur climate knot precip threat
## 0.12 0.12 0.10 0.10 0.10 0.10
plot(dtm, terms = names(findAssocs(dtm,term="asos",0.1)[["asos"]]), corThreshold = 0.1, attrs=list(node=list(label="foo", fillcolor="lightgreen", fontsize="15", shape="ellipse")), edge=list(color="cyan"))
The plot above shows the words that are closely related to ASOS and their correlation with each other. It seems most people are talking about ASOS Curve (women's plus size clothing with ASOS), ASOS midi dress, and ASOS maternity dress.
To collect streaming data from Twitter, we need to set up an OAuth handshake. This can be accomplished using the ROAuth package and codes below. You will only need to do this one time, so long as you save it to an .Rdata file in disk for future sessions.
requestURL <- "https://api.twitter.com/oauth/request_token"
accessURL <- "https://api.twitter.com/oauth/access_token"
authURL <- "https://api.twitter.com/oauth/authorize"
consumerKey <- "t9CMDimaZVL9jVzWZPfpwHxoo"
consumerSecret <- "fPYPpmAOtK5NxHLDfOvTwwTHzcQUO0yriaNRniAJKnligCRFUH"
my_oauth <- OAuthFactory$new(consumerKey = consumerKey,
consumerSecret = consumerSecret,
requestURL = requestURL,
accessURL = accessURL,
authURL = authURL)
my_oauth$handshake(cainfo = system.file("CurlSSL", "cacert.pem", package = "RCurl"))
save(my_oauth, file = "my_oauth.Rdata")The codes below can be used to collect streaming data on tweets mentioning either fashion, clothing or clothes for 10 minutes, save tweets in a json file, parse the json file, and save to a data frame called tweets_stream_fashion. The dimension of the data frame and column names are also shown as below.
load("my_oauth.Rdata")
filterStream(file.name = "tweets_fashion.json",
track = c("fashion", "clothing", "clothes"),
locations = c(-180, -90, 180, 90),
timeout = 60*10,
oauth = my_oauth)tweets_stream_fashion <- parseTweets("tweets_fashion.json", simplify = FALSE)## Warning in vect[notnulls] <- unlist(lapply(lst[notnulls], function(x)
## x[[field[1]]][[field[2]]][[as.numeric(field[3])]][[field[4]]])): number of
## items to replace is not a multiple of replacement length
## 50302 tweets have been parsed.
dim(tweets_stream_fashion)## [1] 50302 42
colnames(tweets_stream_fashion)## [1] "text" "retweet_count"
## [3] "favorited" "truncated"
## [5] "id_str" "in_reply_to_screen_name"
## [7] "source" "retweeted"
## [9] "created_at" "in_reply_to_status_id_str"
## [11] "in_reply_to_user_id_str" "lang"
## [13] "listed_count" "verified"
## [15] "location" "user_id_str"
## [17] "description" "geo_enabled"
## [19] "user_created_at" "statuses_count"
## [21] "followers_count" "favourites_count"
## [23] "protected" "user_url"
## [25] "name" "time_zone"
## [27] "user_lang" "utc_offset"
## [29] "friends_count" "screen_name"
## [31] "country_code" "country"
## [33] "place_type" "full_name"
## [35] "place_name" "place_id"
## [37] "place_lat" "place_lon"
## [39] "lat" "lon"
## [41] "expanded_url" "url"
Besides text mining, we can also do many other data analysis using this data frame. For example, the variable country_code can be used to count the number of tweets sent from different countries. According to the result below, we can see that most tweets regarding fashion and clothing were sent from United States.
tweets_stream_fashion %>% group_by(country_code) %>% summarise(tweets_count=n()) %>% arrange(desc(tweets_count)) %>% top_n(20)## Selecting by tweets_count
## # A tibble: 20 x 2
## country_code tweets_count
## <chr> <int>
## 1 US 20508
## 2 BR 7029
## 3 <NA> 4307
## 4 GB 1611
## 5 AR 1579
## 6 JP 1570
## 7 TR 1052
## 8 MX 1036
## 9 CA 847
## 10 FR 761
## 11 SA 752
## 12 ES 718
## 13 PH 717
## 14 CO 488
## 15 PT 440
## 16 MY 402
## 17 ID 373
## 18 VE 315
## 19 EG 311
## 20 RU 304
We can also capture and visualize tweets sent from the United States as below
map.data <- map_data("state")##
## Attaching package: 'maps'
## The following object is masked from 'package:plyr':
##
## ozone
points <- data.frame(x = as.numeric(tweets_stream_fashion$place_lon), y = as.numeric(tweets_stream_fashion$place_lat))
points <- points[points$y < 50 & points$y > 25 & points$x < -66 & points$x > -125, ]
ggplot(map.data) + geom_map(aes(map_id = region), map = map.data, fill = "white", color = "grey20", size = 0.25) + expand_limits(x = map.data$long, y = map.data$lat) + theme(axis.line = element_blank(), axis.text = element_blank(), axis.ticks = element_blank(), axis.title = element_blank(), panel.background = element_blank(), panel.border = element_blank(), panel.grid.major = element_blank(), plot.background = element_blank(), plot.margin = unit(0 * c(-1.5, -1.5, -1.5, -1.5), "lines")) + geom_point(data = points, aes(x = x, y = y), size = 1, alpha = 1/5, color = "darkblue")## Warning: Removed 4307 rows containing missing values (geom_point).
According to the plot above, it is obvious that people live in large cities, such as Greater Los Angeles Area, Bay Area, Seattle, Chicago, Dallas, Miami, and Northeastern Coast have more interets on fashion and clothing.
In this project, recent tweets and streaming data are collected from twitter for the purpose of exploratory analysis and text mining. Through the tweets collected, we can find that ASOS, Forever 21, and Rag & Bone are the most popular clothing brands among all the 15 clothing brands we discussed in this project. In addition, the tweets collected indicate that many people are talking about ASOS Curve (women's plus size clothing with ASOS), ASOS midi dress, and ASOS maternity dress in Twitter recently. Furthermore, the USA tweets density map indicates that people live in large cities, such as Greater Los Angeles Area, Bay Area, Seattle, Chicago, Dallas, Miami, and Northeastern Coast have more interets on fashion and clothing.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent