Machine learning application for multiple grade latex pricing

Aripin, Amir Asyraf (2022) Machine learning application for multiple grade latex pricing. [Project Paper] (Submitted)

Text FS 2022 11.pdf Download (3MB)

Abstract

Machine learning is an area of artificial intelligence (AI) and computer science that focuses on using data and algorithms to mimic the way humans learn, to steadily improve accuracy. In this study, machine learning is used to predict the multiple grade latex prices. The algorithm models used in this study are BSVR, SVR-PSO, and ARIMAX. The multiple grade latex consists of ‘CV’, ‘L’, ‘5’, ‘GP’, ‘10’, ‘20’, and ‘Bulk Latex’. However, only ‘CV’ and ‘L’ grade is going to be discussed in this study. Some limitation in this study are there has yet to be correlation coefficient (cc) between probable input factors and the pricing of multiple grades of latex, only a small amount of research has been done on the various input variables that impact latex pricing, and the use of BSVR, SVR-PSO, and ARIMAX in forecasting latex prices is not being investigated. The objective in this study are to analyze the cc between each input and the price of multiple grade latex. The positive cc value for monthly are ‘Production’, ‘Exports’, ‘Rubber Thread’, ‘Tapping Area’, and ‘Palm Oil Price’ while for daily, the input factors are ‘CV’, ‘5’, ‘GP’, ‘10’, ‘20’, and ‘Bulk Latex’. The next objective is to determine which input parameters may be combined to get the most accurate latex pricing. The three algorithms were used to test all possible combinations for the input factors with positive cc value. For monthly, it can be determined that 'Exports’, ‘Rubber Thread’, and ‘Palm Oil Price' are the finalize combination of input factors while for daily, the combinations of ‘5', 'GP', '10', and '20' can be used for each other, whereas 'CV', 'L', and 'Bulk Latex' can be used for each other to get the least amount of error and the highest cc value. Final objective is to compare the performance of BSVR, SVR-PSO, and ARIMAX in predicting the price of latex. The comparison is based on the RMSE, MAE, and cc value. For monthly, the algorithm model with low RMSE and MAE value in average is SVR-PSO with RBF kernel function type. The value of RMSE and MAE for ‘CV’ grade for this algorithm is 63.1215 and 54.4225 respectively. While for ‘L’, The value of RMSE and MAE is 53.7942 and 47.7345 respectively. The highest cc value for both grade in monthly is also SVR-PSO (RBF) with 0.8463 for ‘CV’ and 0.8686 for ‘L’. While for daily, the algorithm model with low RMSE and MAE value in average is SVR-PSO with Polynomial kernel function type. The value of RMSE and MAE for ‘CV’ grade for this algorithm is 12.3144 and 12.2899 respectively. While for ‘L’, The value of RMSE and MAE is 27.7739 and 27.6406 respectively. The highest cc value for both grade in monthly is also SVR-PSO (RBF) with 1.000 for ‘CV’ and 1.000 for ‘L’.

Item Type:	Project Paper
Faculty:	Fakulti Sains
Depositing User:	Ms. ROHANA ALIAS
Date Deposited:	25 Jun 2024 10:11
Last Modified:	25 Jun 2024 10:11
URI:	http://psaspb.upm.edu.my/id/eprint/1977

Actions (login required)

View Item