Electronic tongue is a membrane-based sensor made with composition and recipe from some published paper. The sensor consists of 4 sets with each set supposed to sensitive to one specific taste. The tastes specified are umami, salty, bitterness, and sour. By adapting into this idea, the artificial tongue-based sensor can 'felt' the difference components in a solution. For this case, the substances to identify is mineral (Calsium, Potassium, Magnesium) and lead.
The purpose are to identify substances in the solution and compare supervised with unsupervised method (clustering). The expected outputs are 4 model class for each substance classification case or one model multi-class classification.
This initial idea is one small step towards bigger developed product of electronic tongue. By identifying mineral and lead in a conditioned solution, the tools can be more developed by trial into more realistic solution so it will help standardize mineral drinking water and detecting toxic lead in water.
- Data cleaning : relative value calculation
- Data visualization
- Training Scheme : ML with PCA and clustering
- Evaluation