Predictive Modeling to Measure The Productivity of Working From Home in The Software Industry Based on External Environmental Factors

Abstract

The recent COVID-19 pandemic affected many office cultures all around the world. Out of all the industries, Software/IT rapidly adapted to the new remote working culture. Many researchers believe that these moves towards remote work will be permanent in the future. There are few investigations of how people make working from home, work. This study aimed to identify external environmental factors that affect the software/IT industryWorking From Home culture in Sri Lanka. And to analyze the degree to which these factors affect work productivity to make the most accurate predictions possible. An intense literature review revealed potential environmental factors affecting Software/ IT knowledge workers Working From Home productivity mainly in five areas. Such as Respondent Background, Family Members Interactions while Working From Home, Working From Home setup, Working Practices, Surrounding Environments. A survey was conducted on the Sri Lankan software/ IT community to collect data about how they interpret the perceived productivity using a mixed-method approach and, 498 responded. Out of six classification models Random Forest classification model was chosen to predict the productivity of an employee. The trained model was then integrated into an application to predict the work productivity range as a percentage according to the user input about their Working From Home environment. Working from home setup was identified to have the highest importance by carrying out machine learning feature importance techniques. Spearman’s Correlation techniques further revealed that changing the workplace often and location of the work setup has the highest positive correlation. Chair setup and loud people at the work setup showed the highest negative correlation to the Working From Home productivity. The dataset of this study had a limited distribution. Including personality-related factors that affect Working From Home productivity will be an improvement to the study.