Decision Trees in Time Series Reconstruction Problems

This work proposes to use a decision tree classifier for time series data reconstruction. Object of this analysis is to study environmental data acquired by a distributed multi-sensors monitoring system placed in Taranto. The performance obtained in data reconstruction using the proposed decision tree is compared with those obtained using two well known signal reconstruction methods: mean value and polynomial interpolation. The results show that the decision tree outperforms the other two methods in almost all the analyzed cases.