Classify solid samples according to their diffuse reflectance infrared spectra
The Near Infrared Spectroscopy Branch of the China Instrument Society holds the data modeling contest (2022 创和亿杯近红外光谱数据建模竞赛). The organizer provides a set of near-infrared spectral data from the actual application scenario. Different data preprocessing technologies (e.g. PCA, normalization) and machine learning models (AutoGluon) were tried for the classification problem. ExtraTreesClassifier was used as the final model.
Our work won an excellent award in this competition.
Dataset | Input | Output |
---|---|---|
Training Set | 301 samples x 700 features | 301 samples x 1 (of 7 categories) |
Testing Set | 76 samples x 700 features | 76 samples x 1 (of 7 categories)* |
* Need to be classified for competition evaluation
The project was implemented in Colab with python 3.7
!pip install --upgrade mxnet
!pip install autogluon