Business Context:

• Prospective modern-day customers read the product reviews and feedback given by experts and other users on social media before the purchase of new gadgets.

Problem Description:

• This is an analysis of the tech review and customer feedback on twitter, to get insights into what the customers seek from Apple’s iPhone Series.
• To identify the areas of concern that need more focus.

• We have stored the data which we collected using the Twitter API in the below mentioned Google drive link: https://drive.google.com/uc?export=download&id=1kOlxzs8Bq9 WzXl8UlZDD8mOrTE9pyq6s

• Use topic modelling techniques to create topics based on co-occurence of words in the document.
• Use text mining and sentiment analysis.
• Use visualization techniques to demonstrate distribution of sentiments, frequency of the tweets, frequent terms in the tweets and geography-wise tweet analysis.

• Maybe try and add some extra visualization to the project analysis -We have added a wide range of visual images to depict the information in the best way possible.
• Could you also include region specific most popular #hashtags in your analysis. -We have included region-wise frequency of tweets and most popular hashtags have been visualized using a word cloud.
• Will you be performing analysis on the launch of a product by location? -Yes, we have analyzed the location-wise performance with regards to launch of a product by location.
• what amount of tweet data are you considering to use for your project purposes? -We have considered tweets around 23k for our analysis.

• Which statistical model are you using for topic modelling and what are different topics that you are expecting from the topic modelling of your tweets? -We have used LDA topic modelling technique and identified top 5 topics using it.
• Can you not pull tweets with a date filter on them? -We have pulled tweets using a date filter.

• Data has been uploaded at https://drive.google.com/uc?export=download&id=1kOlxzs8Bq9 WzXl8UlZDD8mOrTE9pyq6s
• Data consists of following columns: Sno,user_id,created_at,screen_name,text,source,is_retweet, favorite_count,retweet_count,reply_count,hashtags,place_fu ll_name,place_type,country,country_description,followers_c ount,account_lang

• Frequency of tweets based on the source attributes

• Frequency of tweets based on user location

• Day-to-Day trend of tweets regarding iPhone11 series

• As the analysis is about the data extracted from twitter, there are no predefined sentiments tagged to the tweets in the data. So, Sentimental Analysis is performed.
• Since there is no scope for building models based on training and validation data sets, topic modelling techniques are used on the tweets.
• Latent Dirichlet allocation (LDA) topic modelling technique has been used in the analysis.
• Steps involved: Creating a corpus,Tokenization,Removing Stop Words,Removing Numbers,Removing rare words,Finding unigrams and bigrams,Finding correlation between words,Generating Document-Term Matrix,tf-idf.

Alt text

Alt text

Alt text

Alt text

• NLP has become increasingly popular over the past few years, and NLP researchers have achieved very insightful insights
• The Natural Language Tool Kit (NLTK) is one of the most popular Python libraries for NLP
• Regular Expressions are an important part of NLP, which can be used for pattern matching and filtering
• Common feature engineering techniques are removing stop words, stemming, lemmatization, and n-grams
• How you clean and preprocess your data will have a major effect on the conclusions you’ll be able to draw in your NLP classification problems