2024 : 11 : 21
Mohammad Hossein Fatemi

Mohammad Hossein Fatemi

Academic rank: Professor
ORCID:
Education: PhD.
ScopusId:
HIndex: 0/00
Faculty: Faculty of Chemistry
Address: http://rms.umz.ac.ir/~mhfatemi/en/
Phone: 01135342931

Research

Title
Prediction of Kovats Retention Indices of Some Aliphatic Aldehydes and Ketones on Some Stationary Phases at Different Temperatures Using Artificial Neural Network
Type
JournalPaper
Keywords
I Kovats retention indicesوmultiple linear regression وartificial neural networkو aliphatic ketones
Year
2008
Journal Journal of Chromatographic Science
DOI
Researchers Elahe Knoz ، Mohammad Hossein Fatemi ، Raziieh Faraji

Abstract

In this work, the Kovats retention indices of aliphatic ketones and aldehydes on four stationary phases at different temperatures are predicted. The data set consists of retention indices of 35 aldehydes and ketones on HP-1, HP-50, DB-210, and HP-Innowax stationary phase. The molecular descriptors that appear in this model are: path one connectivity index, fractional atomic charge weighted by partial positive surface area, and dipole moment, which are selected by stepwise multiple linear regression (MLR). The selected descriptors encode steric and electronic aspects of molecules. These descriptors, together with column temperature, are used as inputs of constructed artificial neural network (ANN). The optimized network has 4-3-4 topology, in which its outputs are retention indices of molecules at four stationary phases at the desired temperature. Comparison between statistical results calculated for MLR and ANN models reveals that all statistics have improved considerably in the case of the ANN model. The improved statistics for the ANN would suggest the existence of a nonlinear relation between selected molecular descriptors and their retention in gas chromatography. Also, the simultaneous prediction of retention indices for aldehydes and ketones at four stationary phases at different temperatures using only three molecular descriptors shows the capability of the obtained ANN model