Predicting Operating Cash Flows: A Comparative Study of Machine Learning Algorithms

Operating cash flows (OCF) are considered a financial metric to determine a firm’s decisions regarding financing projects, repaying debts, and dividend policies. With the advancement of machine learning, the method for predicting the operating cash flows using machine learning techniques remains a gap for studying. Therefore, a comparative study is proposed in this paper to find out a predictive accuracy for operating cash flows using different machine learning algorithms, which include Decision Trees (DCT), K-Nearest Neighbors (KNN), Random Forest (RF), and Deep Learning (DL). The dataset used for the experiment comprises $\mathbf{1, 6 3 0}$ samples, which are the statements of cash flow of 163 non-financial companies during the years 2014-2023, gathered from the Stock Exchange of Thailand. The comparison results showed that KNN with $\mathbf{k} \boldsymbol{=} \mathbf{1 0}$ achieves the best performance in predicting the change in OCF with the highest accuracy of 84.28%, compared to other algorithms.