Startup firm’s data analysis tool makes surprising predictions on Delhi elections

TechnoparkToday.com > Riafy Technologies, a company incubated at Kochi’s Startup Village has predicted the results of the Delhi Assembly elections using a revolutionary new data analysis tool — and they run contrary to nearly every forecast made so far for the hotly contested polls.

According to data crunched by RIA, a relationship engine patented by Riafy, the Congress will retain power in Delhi and the newly-formed Aam Aadmi Party, which is expected by many to make significant inroads into the 70-seat state legislature will be unable to open its account this time.

RIA’s predicted share of votes and seats for each of the major contesting parties in the 2013 Delhi Assembly Elections is: Congress at 38%(37 seats), BJP a close second with 37% (27 seats), BSP around 14% (3 seats) and AAP with 3.5% votes which won’t be enough to get a majority in any constituency; with others taking the remaining three seats.

This is RIA’s first attempt at poll predictions after forecasting the box office performance of Hindi movies with an accuracy rate of more than 86% consistently for the past one year.

RIA analyses historical as well as online data for arriving at results and it intuitively modifies search results based on the user’s web footprints, according to John Mathew, the CEO of Riafy.

“We work in Relational Intelligence; as in we try to find relationship patterns from data which are then used in prediction,” he says.

He is aware that predicting election outcomes is a far precarious business than predicting the success of a film. “The data on Delhi elections that we got out of RIA was surprising for many of us who personally expect a different outcome. But this is pure science, it has nothing to do with personal beliefs or political affiliations.”

RIA used historical data and current political sentiment to calculate the number of people who would be voting in Delhi this time. Then it estimated vote share by analysing a number of factors including vote banks, religion/caste, voter inertia, candidate relationship value (indicates how well a candidate is connected to voters in his constituency), media value (indicates the influence that ‘news about a candidate’ has on voters in his constituency) and so on.

A large share of the online chatter around the Delhi elections was generated by automated software or bots and ‘non-voters’ which has little influence on the ground, says John. This also makes their analysis different from exit polls.

“Exit polls are conducted on a limited sample space. RIA on the other hand tries to aggregate sentiment from a large set,” says John. “It tries to understand ‘why’ a person writes a comment rather than trying to understand ‘what’ they wrote. This makes the user’s intent more clear and hence the conclusions drawn more accurate.”

While the exit poll collects opinions from, at the most, 1,00,000 people; RIA quantifies its results from more than 10,00,000, he adds.

RIA’s predictions for the 2013 Delhi Assemble Elections which have been posted on the Startup Village blog — a democratic platform for startup teams to connect to the world — have brought in a fair amount of criticism, indignation, scorn and accusations of bias from readers. But Riafy’s team is unfazed. The election results which are expected to be declared on December 8, whether or not they match RIA’s forecast, will be just another set of data for Riafy to work with to improve their software.