Optimizations of modified machine learning algorithms using k-fold cross validations for wheat productivity: a hyper parametric approach

Farrukh Shehzad; Muhammad Islam; Muhammad Omar; Syed Ijaz Hussain Shah; Rizwan Ahmed; Naeem Sohail

doi:10.17582/journal.sja/2022/38.5.271.278

No Detail Found!

Optimizations of modified machine learning algorithms using k-fold cross validations for wheat productivity: a hyper parametric approach

Author(s)

Farrukh Shehzad The Islamia University of Bahawalpur,,Pakistan

Muhammad Islam The Islamia University of Bahawalpur,,Pakistan; Agriculture Department Bahawalpur, Punjab, Pakistan

Muhammad Omar The Islamia University of Bahawalpur,,Pakistan

Syed Ijaz Hussain Shah Agriculture Department Mandi BahuDin,Punjab,Pakistan

Rizwan Ahmed Agriculture Department Khanewal,Punjab,Pakistan

Naeem Sohail Agriculture Department Gujrat,Punjab,Pakistan

Abstract

An optimized wheat crop productivity model can play a crucial role for evolving efective agricultural policy decisions for food concerns and trepidation. This study measures the eficacies of modified machine learning algorithms using multiple linear regression (MLR), decision tree regression (DTR) and random forest regression (RFR) for wheat productivity using 75% and 25% randomized partitions. The 26,430 field of wheat crop cut experiments (C.C.E) is taken from crop reporting service (CRS), Punjab for the years 2016-17 to 2019-2020. Three generated datasets (D2, D3 and D4) were used to optimize the model performance. The heat plot map shows very strong significance of correlation matrix for D3 and D4, while it was low for D1 and D2. The modified RFR produced lowest values of error for all the datasets, comparing with benchmark DTR and MLR (ErrorMLR > ErrorDTR > ErrorRFR). The modified RFR found best fitted model for the prediction of wheat productivity. The hyper parametric tuning K-Fold cross validation is applied to get the most optimized sub fold for the modified models. It is demonstrated that modified RFR provides superior performance as we advanced from D1 to D4. The results got best when it used D4 for random forest regression with the K Fold-6.

Publication Details

Page(s) 271-278

DOI 10.17582/journal.sja/2022/38.5.271.278

Published Journal: Sarhad Journal of Agriculture, Volume: 38, Issue: 5, Year: 2022

Request the full text from PASTIC

Keywords

Hyper parametric K-fold cross validation, Wheat productivity, Modified machine learning model optimizations