Improving the Performance of Machine Learning Model Selection for Electricity Cost Forecasting in Homebased Small Businesses via Exploratory Data Analysis

Abstract

An Exploratory Data Analysis (EDA) remains a major tool proving a better understanding of the time series dataset with assumptions that could culminate in obvious errors being eliminated. Yet this step during modeling is often done in a less exhaustive manner. This has resulted in assumptions without effective uncovering of the relationships and structure of the underlying dataset. Besides, possible trends and patterns that might not be readily apparent are often not exposed due to the lack of a comprehensive EDA. The primary step to forecasting entails the selection of the appropriate forecasting approach which remains a major challenge for analysts performing the forecasts. This challenge is the reason why many researchers continue to propose new approaches with the objective of forecasting accuracy enhancements. This study examines the impact of applying a robust EDA in improving Machine Learning model selection for electricity cost forecasting in home-based small businesses.