In the second part of our interview with Housing’s Data Science Team, Nitin andPaul talk about the role data will play for Housing and the real estate industry, its use in psephology, and what it takes to be a data scientist!
Want to recap Part 1 of the interview before reading this one? You can find ithere.
Q.7) A lot of credit for Barack Obama’s successful re-election in 2012 has been given to a data science approach led by a 100-strong analytics team. When will we see this in India?
Paul: In 2013, I was a Data Science for Social Good Fellow at the University of Chicago. The fellowship director, Rayid Ghani, had been the Chief Data Scientist for President Obama’s reelection campaign. He was just starting to speak and write publicly about the work his team did during the campaign. I learned a whole lot from him. Their main problem was to identify people who would support the campaign and persuade others in their social network to take action. They called the approach that really defined their success Targeted Sharing. Using a variety of algorithms and predictive models, they identified the subset of campaign supporters who would be most likely to really take action. Then they identified the people in those supporters’ networks who would be most likely to respond positively if the supporters reached out to them. In other words, rather than just asking people to support the campaign, they helped them take action effectively by giving them a list of people they should spend their limited time and effort on recruiting.
I don’t know when that kind of analytics will be applied to political campaigns in India. But getting elected is easier and less important than implementing good policies after you’re elected. So I’m less interested in whether analytics are used for political campaigns in India. Instead, it’s interesting to think about how analytics can be used to create positive social change and better opportunities for people across society. I’m very optimistic about that kind of work. There are some well-established efforts in India like ipaidabribe.com which make great use of data. I recently spoke with one of the founders of socialcops.com and I’m excited to see the kind of data-driven and technology-enabled work that they’re beginning. I’m also keen to see what comes out of the new DataKind chapter in Bangalore. DataKind is a great organisation that helps grow communities of data scientists focused on helping non-profit and public organisations and causes and on generally contributing to social good. In other words, it’s already very exciting times for social-related data science in India.
Q.8) What role will data play in the days to come for Housing and the industry?
Nitin: Imagine this – Mr. X wants to move. He visits Housing.com and selects the city and neighbourhood where he’d like to live. Like an efficient personal assistant, we understand his requirements, social network, and community. We use this personalised understanding to recommend houses. He explores his top choices virtually, selecting furnishings that suit his tastes. Maybe someday he can go outside and look around, all in virtual reality.
Once he has decided, we instantly connect him to the bank. The bank checks his credit history and approves the loan. The sale is done. He schedules a move-in date and we send movers to shift his belongings from his old house to his new home.
This is the experience we want to build in the future. Each of these steps is connected to the next with a huge amount of data about the customer, real estate, credit histories, loan criteria, and much more.
Data science will be the key differentiator for all platforms providing real estate services, or really, for any services-based industry.
Q.9) What would you advise people, especially those who don’t have a computer science or engineering background, before diving into the field?
Paul: I’d give the same advice to people wanting to dive into the field: develop a strong curiosity and seek out problems to solve. For someone without a technical background, I’d recommend getting involved in a local data science community. Join data-related meetup groups, participate in hackathons, take MOOCs (Massive Open Online Courses) and listen to data science talks on YouTube. It’ll take up your evenings and weekends, but it’ll be worth it. And if you want a very specific suggestion, I’d say read ‘Data Science for Business’ by Foster Provost and Tom Fawcett.
Q.10) What are the gaps in the data Housing has? How do you plan to address them?
Paul: At Housing, we think of data as an investment not just as something we passively collect. You have to put effort and resources into acquiring data before you see it bring returns a little further down the road. Housing is the only online real estate platform in India that actually sends out data collection teams to collect data on every single listing on our site. That’s an investment that has already paid off.
We’re constantly thinking about acquiring and developing new kinds of data. For example, in India there’s a huge problem with developers demanding part-payment in cash – basically black money. Besides being a significant social problem it’s also a data problem. We want to provide accurate and useful property valuation estimates to both buyers and sellers. That effort is degraded when substantial portions of transactions are made in cash and go unreported. We’re developing a number of ways to help limit and overcome that problem.
In addition to collecting data, we also do a lot to create it. Running experiments and surveys that produce new data will be a big part of data science at Housing. We use crowdsourcing platforms like Amazon’s Mechanical Turk in order to generate data for a variety of our products. We’re developing a major new effort, using approaches and methods like active learning and deep learning, to turn our massive data-store of images from listings into structured data that we can feed into our algorithms and predictive models.
Q.11) How do you keep up with the latest developments in your field?
Nitin: There are a number of channels we tap. Some good examples would be: reading up and attending machine learning and data management conferences like the ones from MLconf and Fifth Elephant; participating in Kaggle competitions and hackathons; subscribing to feeds from sources like KDnuggets. We also try to stay aware of the technologies being used by other leading startups and technology companies. Whenever we visit a webpage, we know this feature is powered by a machine learning engine. We then analyse its behaviour and read up on the maths behind it. We also make great use of MOOCs. At any given time, about half of our team is taking a MOOC from Coursera, or EdX, or Udacity.
Q.12) Apart from technical skills, what is the one thing you look out for in prospective employees? How do you know if someone has an eye for data?
Nitin: Our data scientists are involved in all aspects of a data-based startup – from research to conceptualisation to experimentation to deployment and maintenance of services. We like individuals who can be a part of that full process – spot a neat place to make a dent and then just go about executing on the idea. In other words, we want people who love to design data products end-to-end.
Courtesy : www.Housing.com/blog
EmoticonEmoticon