File 031521
Email Correspondence Regarding 2016 US Election Data Analysis and Social Media Predictions (File 031521)
Email exchange between Zubair Khan and Jeffrey E. discussing a data analysis project that predicted the 2016 US presidential election outcome using social media data, with discussion of project limitations and future plans.
Summary
Zubair Khan reports to Jeffrey E. on the accuracy of their election prediction model, which correctly predicted 11 out of 13 states. Khan discusses how their analysis showed Trump dominated social media positively in the pre-election period, leading to their October 30th prediction of Trump's victory. He reflects on model limitations regarding the optimal amount of historical data needed for accurate predictions and mentions the tool's potential applications for voter tracking in other elections, while noting the project is being shut down due to expensive data acquisition costs from Twitter.
From: Zubair KhanSent: 11/11/2016 6:34:29 AMTo: jeffrey E. [jeeyacation@gmail.com]Subject: Re:Importance: HighJust checked report again. NY is correct. We only got WI wrong.Regards,ZubairOn Fri, 11 Nov 2016 at 10:29 am, jeffrey E. <jeevacation@gmail.com> wrote:New York ? How ? It was hillary no doubt alwaysOn Friday, 11 November 2016, Zubair Khan ‘1> wrote:Out of 13 states we got 11 right. The ones we got wrong are NY and WI.In the last report, analyzed data was of 3 days in which Clinton got a positive spike after she got clearancefrom FBI. However Trump dominated the social media positively if we analyze one week data before startingtime of election, monthly data and entire data which we have gathered for months. Based on this we predictedon our Facebook page on 30th October that Trump will be next president of US.Our model works but I think the important question which we failed to answer is that how much historic datamust be analyzed to make the right decision. This could have been answered if we had a Data Scientist. Also Ibelieve its a great tool to track voters and get them to vote. It can be used in French elections next year orelections of Brazil in 2018 but unfortunately I have to shut down this project as getting data from Twitter isexpensive and the start up is running out of money.I am going to focus on cyber security again and will definitely reach out if I have a powerful investable idea.Regards,Zubair KhanOn Nov 11, 2016, at 4:49 AM, jeffrey E. <jeevacation@gmail.com> wrote:not so goodplease noteThe information contained in this communication isconfidential, may be attorney-client privileged, mayconstitute inside information, and is intended only forthe use of the addressee. It is the property ofJEEUnauthorized use, disclosure or copying of thiscommunication or any part thereof is strictly prohibitedand may be unlawful. If you have received thiscommunication in error, please notify us immediately byHOUSE OVERSIGHT 031521return e-mail or by e-mail to jeevacation@gmail.com, anddestroy this communication and all copies thereof,including all attachments. copyright -all rights reservedplease noteThe information contained in this communication isconfidential, may be attorney-client privileged, mayconstitute inside information, and is intended only forthe use of the addressee. It is the property ofJEEUnauthorized use, disclosure or copying of thiscommunication or any part thereof is strictly prohibitedand may be unlawful. If you have received thiscommunication in error, please notify us immediately byreturn e-mail or by e-mail to jeevacation@gmail.com, anddestroy this communication and all copies thereof,including all attachments. copyright -all rights reservedHOUSE OVERSIGHT 031522