Using machine learning models but worrying about CENSORING?

CondiS got you covered!



What you'll find here


An interactive tool to help you impute censored observations in your dataset and make predictions using machine learning models.


How it can help you


With a 'normal', no-censoring, complete dataset, you can apply any algorthom you want to apply, including mainstream machine learning/deep learning models.




Author: Yizhuo Wang (maintainer), Xuelin Huang, Ziyi Li, Christopher R. Flowers
Please direct any question to ywang70@mdanderson.org , thank you!

PID:1106; Version: V1.1.0.0; Last Updated: 3/10/2022

CondiS

CondiS imputes survival times for censored observations based on their conditional survival distributions derived from Kaplan-Meier estimators.

CondiS-X

When covariates are available, CondiS is extended by incorporating the covariate information through regression modeling (CondiS-X), which further increases the accuracy of the imputed survival time.

Downstream application

With the censored observations being properly imputed, machine learning algorithms can be directly used to predict survival times for newly enrolled patients.

Below is a schematic diagram of this application:


* Note that this data file must include a censoring/event indicator and a survival time variable. The other covariate variables are optional.
Click here to see an example dataset.
* 1: event, 0: censoring. Please adjust this variable of your input data accordingly.



Please wait for a couple seconds if your dataset is large.

Download

* Note that here CondiS imputes survival times based on the maximum follow-up time. If you are interested in a specific time range, i.e., if you want to set up a time point instead of the default maximum follow-up time, please install our package (CondiS) from CRAN. Thank you!
Click here to check CondiS package.

Please wait for a couple seconds if your dataset is large.

Download

* Note that here CondiS-X improves the imputed survival time using a generalized linear model. If you want to use another machine learning algorithom to incorporate the covariates information, please install our package (CondiS) from CRAN. Thank you!
Click here to check CondiS package.

* Note that this data file should contain the same variables (except for the censoring/event indicator & the survival time) as the data file you upload in the previous section.
Click here to see an example dataset.

Download

* WARNING: our application doesn't have the hyperparameter tuning feature. Hyperparameters control the overall performance of a machine learning model. If you see any unreasonable predictions, please download the imputed survival times and resort to a programming language to tune your model.Thank you!