Approach to text data clustering based on molecular chemical reactions in agricultural and food industries

Abstract

We propose an algorithm for clustering short text fragments such as lists of ingredients in food products using the k-means algorithm in combination with the molecular chemical reactions algorithm. The solution obtained with k-means is used as an initial molecular structure solution to optimize chemical reactions in combination with k-means by generating new solutions using the four main types of chemical reactions. The computational experiment showed the comparative efficiency of new algorithm, taking into account numerical metrics: accuracy, precision, recall, and F-measure.