Imputação de valores faltantes em série histórica no cenário automotivo

Abstract

Machine learning models have been helping to create solutions for years, in many different areas of knowledge. Whether in the creation of a sales prediction, or the classification of an image. In time series specifically, we can predict the value of a product in the coming months, and thus anticipate any action to be more accurate. However, one of the most frequent problems encountered is the lack of data, that is, when we have a database with little information or with holes in its history. Due to this, some models do not work well and lose value due to lack of data. This work brings the proposal to solve the problem of missing data in a database with vehicle information. Through machine learning algorithms the missing data is filled in, based on the few filled in data.