A Design and analysis of classification models for the gelification of alkoxybenzoates using the kNN algorithm.

Classification models of the states produced in the gelation tests of organic molecules require designing several corpora of data based on their characteristics. This work studies 15 solvents characterized by Hansen Solubility Parameters and a series of alkoxybenzoates. Characterization of the alkoxybenzoates has as distinctive feature the number of carbons on its alkyl tails. Solvent and molecule properties were evaluated as attributes on corpus through the kNN algorithm. Three corpora were tested in different algorithm configurations, varying each corpus content according to the solvents and molecules attributes. Relevance of some attributes over others on the performance prediction of the products class can be appreciated. Significant instances were correctly classified on corpora when the HSP and the alkoxybenzoate alkyl ether tail length were considered, thus, stating the influence of these properties on classification accuracy. The most suitable configurations in kNN as metric, k value, and attribute weight were determined according to each corpus.