Targeted prevention of risky deals for improper granular data with deep learning

Deciding whether to accept or reject business deals is a complex task that traditionally demands considerable time and effort due to multiple attribute evaluations. While machine learning models have improved decision-making in this area, the success of such models heavily depends on data granularity. This paper presents a deep learning approach to enhance risk prediction in deal management by streamlining data granularity. In our experiments on a large dataset with 2 million records, our proposed model achieved an accuracy rate of 94%, outperforming traditional ensemble methods that reached only 76% and optimized models achieving up to 93%. Key metrics for our Artificial Neural Network model demonstrated high reliability, with a specificity of 0.96 for low risk, 0.85 for medium risk, and 0.99 for high risk. Furthermore, our model showed an F1 score of 0.86 for low risk, 0.93 for medium risk, and 0.94 for high risk. These improvements enable businesses to predict and mitigate risks with greater accuracy, ultimately saving time, reducing costs, and improving overall business outcomes.