Implementation of Machine Learning Techniques to Predict Briskness and Brightness of Tea Liquor using Factory Data

This paper documents the prediction of two important quality attributes of tea liquor viz. briskness and brightness using process parameters of fermentation room of tea factory. Different machine learning techniques are implemented for this work. Temperature and relative humidity are two important parameters of the fermentation room. These data are collected from the tea factory by the developed instrument during factory hours. Corresponding tea samples are also collected and tea quality is evaluated by tea tasters. This tea quality is a numeric value called overall liquor rating which is a function of different tea quality attributes like briskness, body, brightness etc. This work is an attempt to predict briskness and brightness from the collected factory data by implementing different machine learning techniques viz. k- Nearest Neighbor, decision tree and random forest. Average accuracy, f1-score and recall are found more than 95% for k-Nearest Neighbour, decision tree and random forest in training and validation for both the attributes.