I had an ML challenge that I hoping for advice on. Overall, my aim is to use yearly real-estate data to predict neighbourhood change, primarily gentrification.
Initially, I have created a label for each year from 2013 to 2019. This label is either 0 (non-gentrification) or 1 (gentrification). Then the data I want to train on is real-estate data, in which I also have yearly data.
One way I have been exploring to learn ML tools has been to simply train a Random Forest real-estate data and predict the 2019 gentrification if it predicts well, then I can use 2019 real-estate data to predict 2025.
I have been looking at Neural Networks LSTMs and thinking more about how to incorporate this time-series element in. I was thinking do I train 2013 real-estate data to predict 2014 gentrification, and then keep combing the yearly models?
Basically I am unsure whether the best route to go down is by having a label for each year and combining the model? OR to have 1 label and train on the yearly data using LSTM?
I have struggled to find many tutorials on such a time series challenge regarding yearly data, and labels for each year, but maybe I’m looking in the wrong place