Risk-Based Audit Engagement Planning: Incorporation of Predictive Analytics

Abstract

The inclusion of predictive analytics in auditing engagement planning is a revolution in risk identification and evaluation. The conventional risk-based audit practices are not well-suited to handle the large amounts of data and speed of data in contemporary business organisations, which may negatively impact the identification of emerging risks or misuse of resources. The research paper constructs and justifies a hybrid model that consists of machine learning algorithms and best practices of audit risk procedures. With 847 audit engagements with seven years of data across industries, we utilize Random Forest, Gradient Boosting and Neural Network models to forecast audit risk using much more accurate forecasts as compared to conventional methods. The findings show that predictive analytics can identify risks with an accuracy of F1-score of 0.847 vs 0.689 and cut the time spent on planning by 31%. Utilizing the feature importance analysis, cash flow volatility, the complexity of governance, and industry-adjusted ratios are found to be the most important predictors. We determine such critical success factors as data infrastructure preparation and maintenance of professional judgment through case studies and practitioner interviews. The study shows that data science methods can enhance human knowledge in the workplace.