This repository is a personal project for a centralized documentation about the concepts and implementation examples of some of the most relevant methods and tools currently used for time series forecasting.
###Topics:
-
Introduction to time series (TS)
- Description of data structure, AC Plots
- Component decomposition
- Stationary propoerties, Dickey-Fuller test
- Target transformations (Log, Cox-box)
- Residuals
- Metrics to measure forecasting performance
-
Statistical Models
- Univariate TS modelling: Average, Exponential smoothing, ARIMA models
- Multivariate TS & External-regressors
- Component-based TS modelling(tools): Facebook's Prophet, RSTL
- Structural TS modelling: Hierarchical time series, Bayesian structural TS
-
Machine Learning Models
- (Linear) Regression models & standard toolbox (SVM, GBDT)
- Hidden Markov Models
- Seq2seq models
- Recurrent NN (RNN,LSTM)
-
Use Bayesian forecasting