I have a data set such that each data point is an "event" with features $x_1, x_2, \dots, x_n$ and the year of its occurence $y$. I want to train a forecasting model that predicts when an event of features given is going to happen.
My problem has two parts:
What is the standard approach to such problems? It seems to be conceptually different from time-series forecasting and fitting a Gaussian Process or Hidden Markov Model. I was thinking about training a regression model $y = f(x_1, x_2, \dots, x_n)$ but I am not sure whether this will learn how to predict future events.
How does one measure performance of such method? I was thinking about training the model on events occuring e.g. before 2017 and then using 2018 events as the validation data set and 2019 as the test data set. However, the lack of data from 2018 and 2019 will make the model less powerful in predicting 2020 events.