The goal of this project is to show the fundamental steps for the construction of a forecast model. Starting with the study of the time series, going through the metrics and validation processes, to the construction of the model and finally making predictions and analyzing results.
To make the study robust, we will build two different models, always analysing why we chose the model and their parameters, their strengths and weaknesses, and what are the improvements of one model in relation to another, that is, what aspect the model in question intends to improve in relation to its predecessor.
We will work with a dataset of daily minimum temperatures in Melbourne, Australia (1981-1990) and our main goal will be to perform one-step forecasts (the minimun temperature of the next day).
The project will be developed in 4 steps (and files):
- 1º Time series analysis;
- 2º Build Exponential Smoothing model;
- 3º Build ARIMA model;