I remember reading somewhere that there is no better time to be a politician. But it’s an even better time to be a machine learning engineer. In this talk, I would like to discuss how Data Mining and other Data Science techniques help in understanding public sentiments based on geography, profession, community and customize every campaign based on the audience.

Here I would like to discuss the following points:

  1. Data's impact on the election.
  2. Smarter targeting of voters.
  3. Various channels to reach the voter.
  4. Social Media to influence Voter behaviour.
  5. How political parties are moving away from Traditional Models.
  6. A Case Study
  7. Using it all for Good !!

Demo: How I modeled an awesome election outcome predictor using Neural Networks!!

 
1 favorite thumb_down thumb_up 3 comments visibility_off  Remove from Watchlist visibility  Add to Watchlist
 

Outline/Structure of the Case Study

  1. Intro & first dive
  2. Data's impact on the election.
  3. Smarter targeting of voters.
  4. Various channels to reach the voter.
  5. Social Media to influence Voter behaviour.
  6. How political parties are moving away from Traditional Models.
  7. A Case Study & Demo
  8. Using it all for Good !!
  9. Q&A

Learning Outcome

  • Understand and improve on behavioral analysis.
  • Various points to consider when picking data samples.
  • Identify Dimensions that create huge impact in the outcome.

Target Audience

Data Engineers, Data Analysts, Machine Learning Engineers, Any one who likes behavioral analysis

schedule Submitted 3 months ago

Public Feedback

comment Suggest improvements to the Speaker
  • Dr. Vikas Agrawal
    By Dr. Vikas Agrawal  ~  2 months ago
    reply Reply

    Dear Gautam: Thanks for the proposal! Do you happen to have a video of one of your previous talks or could you please record a video introducing the topic for the ODSC audience and post it here?

    Warm Regards

    Vikas

    • Anoop Kulkarni
      By Anoop Kulkarni  ~  2 months ago
      reply Reply

      Thanks for your proposal. Being the election year, this is also a timely presentation. Curious to know what case study and what demo you plan to include? Which data you would be using? What are your data sources? 

      Would it be possible to include these details?

      ~anoop

      • Gautam Anand
        By Gautam Anand  ~  2 months ago
        reply Reply

        Hi Anoop, thank you for looking into my proposal.

        My case study is how the social media cells of the political parties make can use of the data coming in from various sources such as tweets, FB posts and news feeds to come up with targeted campaigns in every geo and region. 

        For eg: In bangalore, the major pain point can be commute, traffic etc. In some other area, it can be employment, in other area it can cleanliness or health. Every area, geo or region has its own pain points and positives. This info will help the political parties adopt an issue based campaign rather than just rhetorics. 

        The overall intent is for the parties to understand various areas of concerns per locality, type of occupation and then campaign to those individuals and localities in a targeted manner rather than going ahead with a generic agenda and a generic campaign model.

        While pursuing my masters from BITS Pilani, as part of the dissertation i was granted access to unlimited twitter feeds starting from the very first tweet back in 2006 by the twitter dev team when I submitted my proposal. I used the twitter data from 2013 to 2014 to come up with predictions during the 2014 elections and they were pretty accurate. I plan to do the same activity, but this time with more number of data sources such as Facebook posts, tweets and news feeds and for the data collected for the 2019 elections. Since the ODSC is in August, I thought I will be able to come up with the observations for all the use cases mentioned above for the latest Lok Sabha election that would have had just wrapped up.

        Also, my demo would consist of

        1. a step by step approach on how to handle this data
        2. Operations performed on the data to bring it into a trainable form.
        3. Top attributes to consider that add more value to the data than just weight.
        4. Mark/Discard polarities. Identify opinions only from unbiased poster.
        5. Identify a troll/spam tweet and discard from our model.
        6. List the ML and NLP Models used and why were they used in specific.
        7. All the observations that were noted down, and some hard facts around how ML and recommendations in social media during elections polarises a part of the society.
        8. Finally we will present our models to validate if our predictions were in line with the reality. (If time permits).

        Hope I have answered your question. As of now, I don't have a structure for the presentation in my mind. But overall this is what I want to present.