Analytics based study of Bangalore real estate market

Abstract

In this project we have modelled the Apartment prices in Bangalore. Our approach is based on hedonic regression. We have taken a two-step approach to modelling. First we model the Apartment prices based on amenities that are provided in the Apartment. This model does not have a time component and is fixed for each year. The average price of the locality, serves as an index price in this model. We use the Magic Bricks data from 2010 to2015 to model this problem. Next we model the index price using a panel data regression across multiple years. For this augment the online data with the manually collected offline data. The following are the questions that we try to answer in our work:· What are the amenities that contribute to the pricing model of the apartment?· Problems that are encountered in a modelling scheme based on online listing data and mitigation steps including principal component analysis and step wise regression.· The trend in index/average price in multiple localities and the connection to demand and supply over multiple years. Modelling this using panel data regression.· The impact of price and the distance to work, business and school locations? Which parameters are relevant and which are not? The following insights were arrived at based on the work:· Following variables were found to be significant in the locality pricing model: supply, proximities to work and school and presence of public infrastructure.· A locality wise pricing index model which can be potentially used to update the year on year guidance value (adjusted R-square = 0.68)· An overall pricing model for an apartment which also consider the locality pricing index into account apart from the apartment specific attributes (adjusted R-square= 42.1%).