CBRi: A Case Based Reasoning-Inspired Approach for Univariate Time Series Imputation
Anibal Flores, Hugo Tito, Carlos Silva · 2019
This paper presents a new approach for univariate time series imputation inspired by Case Based Reasoning (CBR), hence the name CBRi. Time series of maximum temperatures were used and two case bases were implemented, the first one was based on a time series of daily data corresponding to 4 years; and the second one from the same previous time series but with a greater amount of data, in this case 9 years. The results of the proposal are compared with different imputation algorithms, both univariate and multiple. CBRi achieved very good results, outperforming in most cases the algorithms with which it was compared.