Coursera Data Science Capstone Project

Scope of this presentation is to describe the way this predictive application works and what options users have.

This is the final assignment for the Capstone project of Data Science Specialization

• Application is divided into 2 sections , left section and right section. Within left section user types a single word or a sentence
• They have the ability to selet how many words will be predicted (up to 10) through a silder. Default value is 4
• Regardless of the slider bar , the word with the bigger probability is shown in top right section , highlighted in blue
• Above this , a bar chart shows all the words (depending on the slider bar selection) that have probability to be shown ordered respectively.

• US Dataset provided had to cleaned up , punctuation , special characters and non-English characters had to be removed
• All words should be presented in lowercase while numbers and years should be categorized properly.
• A list of profanity words was used from a website to remove these appropriate words
• The full dataset was too big to be fully processed hence the biggest available subset was selected through R functions
• Next step involves tokenization and creation of n-grams
• Based on our model design , 4 n-grams have been created that predict words according to the number of words typed.

This app is hosted on ShinyApps.io.

ui.R and Server.R files , n-grams files used in the app and rest of R script can be found at GitHub.

R Presentation is available at RPubs.