Multivariate Cuban Consumer Price Index Database, Statistic Analysis and Forecast Baseline Based on Vector Autoregressive

The global Consumer Price Index (CPI) is a monthly multivariate time series, which allows measuring the variation of the final consumer prices of a given set of goods and services of households living in a given geographic region, city or country. The present work addresses the problem of the multivariate time series database of Cuba’s CPI and a respective forecasting model based on Vector Autoregressive to establish a baseline for this dataset. An statistical analysis of the data will allow characterizing each variable of the series in terms of relevance to the multivariate problem, its causal relationships and the respective stationary analysis to evaluate the best lag to be considered in the forecasting model. The main statistics evidences of each test were reported in the paper as starting point for futures researches in the field of deep learning.