Sean Taylor — Business Decision Problems

Prophet Development

Sean Taylor shares insights into the development of Facebook's Prophet, a time series forecasting tool. He explains that Prophet was created to address the lack of effective tools for business time series forecasting, particularly those with multi-period seasonality like yearly and weekly cycles 1. Sean notes that Prophet is designed for human behavior-generated data, such as web traffic, which exhibits predictable patterns 2.

Prophet provides a pretty easy way to get there.

--- Sean Taylor

He emphasizes that while Prophet may not be suitable for chaotic or non-human periodic data, it excels in scenarios where local smoothness assumptions hold true 2.

   

Election Forecasting

In discussing election forecasting, Sean Taylor highlights the divergence between traditional poll aggregation and prediction markets. He suggests that prediction markets may be influenced by emotional hedging, where participants bet against their preferences to mitigate potential disappointment 3. Sean expresses confidence in polls, citing their scientific foundation and long-standing use, but acknowledges potential biases in how people respond to them 4.

I am a big believer in polls.

--- Sean Taylor

He notes that while prediction markets offer an alternative perspective, their effectiveness is limited by participant motivations and transaction costs 4.