I am attempting to build a predictive model based on the past historical data. I have details of specific machine failure based on the past year data. I have data from some months of 2016 and from 2017 January to November. I am attempting to predict if the machine fails in December. I have attached the past historical data based on the data transformation I performed. I am stuck at a point on how to build a model to identify if the machine throws up any repair or replace in next one month.

I need to build a classification model which will identify if the machine fails. I don't understand on how to convert this multiple rows to single row record for each machine. Do I need to create categorical for day and month. How would I represent each sensor data column. Essentially look at the right way of representing data so that it is easier for classification to predict for next one month.

Any sort of guidance would be of great help dataset.csv train.csv

train.csv contains the 1 month data of December for the machines which are already predicted. The target column is to predict in the month of December, if the machine requires a repair or replacement or no issue.


1 Answer 1


In essence you want to shift your outcome variable by 1, that is place december's outcome to november. In doing so you are saying that the values in the current month (november) caused the next month's (december) outcome. As to the question of processing your data that depends entirely on your environment / software used for the analysis, however they seem good as they are now.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.