This paper seeks to identify whether the analysis of Volatile Organic Compound (VOC) signals using a gas chromatograph mass spectrometer (GCMS) will enable the use of biomedical applications in recognising the presence of certain bacteria in infected wounds. Auto Machine Learning (AutoML) is used in this report to determine the most appropriate model for the data, firstly looking into the specific strain of bacteria for each sample and then focusing on the medium used for each sample.
Complete
- Python
- Tpot AutoML
- Jupyter notebook
- Scikit-learn (random forest classifier)
The results from the baseline system and AutoML system using the strain as labels shows an increase in accuracy of predicting the correct strain while using AutoML. The baseline returned 76% to an improved 89% with AutoML.