Files in this directory illustrate how to use Predictive Maintenance Modeler on an example multivariate time series dataset.
Data for this example are publicly-available multivariate time-series data from Scania trucks, which was specifically created for time-to-event analysis. This whitepaper explains the background on the dataset, which can be downloaded here.
This notebook merges these raw data into one dataset and carries out some initial preprocessing steps, including:
- creating the
target_feature - using groupby forward-fill for missing values
- to backward fill missing values, I use MICE imputation using Bayesian ridge regression.
This preprocessed dataset can be downloaded here:
Exploratory data analysis of the preprocessed dataset is done in this notebook.
This notebook provides a walkthrough of train and predict functionality