Machine Learning Based Prediction of Pyrolytic Conversion for Red Sea Seaweed

Abstract

The complex and non-linear relationship between kinetic parameters and main biomass components is one of the major problem for model development. The application of non-linear modelling would be interesting to predict the behavior of such a complicated model. The machine learning approaches are widely used in many applications for modelling the non-linear relationship between input and output data. In this study, two machine learning tools, namely, Artificial Neural Network (ANN) and Support Vector Machine (SVM) are investigated in order to develop a prediction model for the pyrolytic conversion of Red Sea seaweed (Sargassum sp.). Pyrolysis of Sargassum sp. was carried out in a thermogravimetric analyzer (TGA) where the samples were nonisothermally heated at 10, 15, 20 and 25°C•min −1 from 40 to 900°C in an inert atmosphere. The obtained mass loss-time and temperature series data sets are used for the training, validation and testing the prediction model with ANN and SVM. The results show that ANN provide better accuracy than SVM. More importantly, higher R 2 and lower error rates renders machine learning based tools as reliable means for forecasting pyrolytic conversion.