Skip to content

Classify chemiotherapy molecules and predict their concentrations using their Raman Spectrum

Notifications You must be signed in to change notification settings

petermartigny/datacamp2016

Repository files navigation

DataCamp2016 @ Ecole polytechnique

In this challenge, we are given Raman spectra for 1000 chemical solutions with different containers, solutes, molecules (4 different molecules for chemiotherapy usage) and concentrations. Providing the right molecule at the right concentration is a crucial point for the chemiotherapy to be successful.

This project is about optimizing these two problems:

Classification task

Predict the molecule present in the solution

Regression task

Predict the concentration of the molecules

Metrics used

Classification: 0/1 loss

Regression: MARE

Combined: 2/3 * clf + 1/3 * reg

Scores (combined)

Frome the starting kit: 0.15

Final individual score: 0.053

Score after collaborative session: 0.040

About

Classify chemiotherapy molecules and predict their concentrations using their Raman Spectrum

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published