Revenue Prediction using Sequential Machine Learning
Vijay Mahadeo Mane, Rajat Patil, Swanand Patwardhan, Piyush Pethkar · 2024
Forecasting has been an inherent human activity, practiced since prehistoric times through both scientific and nonscientific methods. With the advancement of science, numerous statistical algorithms and models have been developed. This paper presents a comparative study of two sequential machine learning algorithms, Long Short-Term Memory (LSTM) and Prophet, for revenue prediction. The dataset, sourced from Kaggle, includes daily records and common parameters of a company. After preprocessing and analysis, the models' outputs were compared. LSTM achieved a mean absolute percentage error (mAPE) of 12.49, while Prophet recorded a mAPE of 14.41. The trained models were evaluated, and trends were observed. As revenue prediction becomes increasingly critical for organizations aiming to optimize production and maximize profit, this study provides valuable insights for selecting effective sequential machine learning algorithms.