Title: Feature engineering-based predictive modelling using a non-imputed dataset

Authors: Partha Sarathi Mishra; Debabrata Nandi

Addresses: Department of Computer Science, North Orissa University, Sriram Chandra Vihar, Takatpur, Baripada, Odisha-757003, India ' Department of Remote Sensing and GIS, North Orissa University, Sriram Chandra Vihar, Takatpur, Baripada, Odisha-757003, India

Abstract: This paper presents random investigations of different regression algorithms of machine learning techniques by considering some of the feature engineering principles using a major nutrient-based soil dataset containing missing values and noisy instances. Here, one of the feature engineering techniques like imputation is considered and emphasised for the efficient random investigation analysis using seven regression machine learning algorithms. The subsequent dataset then participates without losing its basic features and negotiating accuracy results. Finally, the transformed input dataset tuned by feature engineering principles is used by the different machine learning algorithms for random investigations. The result obtained from the different experiments ensures that the feature engineering-based linear regression, Ridge regression, and Classification and Regression tree regression algorithms powered by explanatory are one of the significant approaches in the field of machine learning in building a potent explanatory model with improved performances.

Keywords: feature engineering; imputation; feature selections; regression algorithms; data mining.

DOI: 10.1504/IJDS.2021.121097

International Journal of Data Science, 2021 Vol.6 No.3, pp.241 - 256

Accepted: 20 Oct 2021
Published online: 24 Feb 2022 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article