In this chapter, we learned about a very popular Bayesian forecasting model known as the Gaussian process and used it to predict stock prices.
In the first part of this chapter, we looked at the forecasting problem by sampling an appropriate function from a multivariate Gaussian rather than use predicting point forecasts. We looked at a special kind of non-parametric Bayesian model named Gaussian processes.
Thereafter, we used GP to predict the prices of three stocks, namely Google, Netflix, and GE, for 2017 and Q4 2018. We observed that our predictions were mostly within a 95% confidence interval, but far from perfect.
Gaussian processes are used widely in applications where we need to model non-linear functions with uncertainty with very few data points. However, they sometimes fail to scale to very high dimensional problems in which other deep learning algorithms, such as LSTM, would perform better.
In the next chapter, we will take a closer look at an unsupervised approach to detecting...