Go & Google Yourself!

Description

This study examined the development of a Recommender System based on Google Takeout data. This system is built on a web application that is will be documented on this paper. The main motivation to develop it is this following concept: Make the Google Data accessible to everyone. We already know that every person that has a Google account is able to download their information, but not everyone has the ability to handle or even understand what mbox or json files are. Then we created a web app that allows every person can "read" their own data. First, is been used the anonymized data from project's volunteers, it is been downloaded the historical records of Gmail, Location and Searching from Google. Then It was driven an exploratory data analysis in order to understand the nature of features. After that, it is been predicted the most frequent places the individual has been, this step using Non-supervised Machine Learning algorithms based on density. It is been found the network that the individual can be influenced by, and all the important keywords used, in order to filter topics the person is likely to find out using text mining algorithms. All the process it is been documented and it is open code, the reader can easily try the beta version or reproduce the app instead.

Questions

• Which steps you need to take in account? Could you have some rules for doing it? Or using some ML algorithm for doing that?
• Get your network from your emails
• Get your interests from your searches
• Get your path of life from your locations
• Are there any relations?
• Create this a product, i.e. any could put their data in your pipeline and see his life
• Are this related to your other social networks?
• How you could modify your tool so it could be used for a group?

Steps

• Exploratory Data Analysis ¿What do we have?
• Project Scope and Data Stories ¿What's the goal?
• Mockup and Data Pipeline ¿What's the path?
• Model Features ¿What we gonna use to predict?
• Model Insights ¿Does it works?
• User interface design ¿A human can use it?
• Final Report

Source Google Takeout