Prediction based Enterprise Cash Holding Management by using Back Propagation Neural Network

In recent days, efficient cash holding management is important for enterprises to monitor liquidity, reduce financial risk, and increase investment returns. However, the existing Deep Q-Network (DQN) heavily depended on human judgement which led to inaccurate predictions and suboptimal decisions. Hence, this research proposes Back Propagation Neural Network (BPNN) for accurate prediction of enterprise cash holding management. The proposed BPNN provide accurate predictions by minimizing errors in cashflows and this procedure begins with the collection of input data from Financial and Total Asset Holdings of a Startup (FTAHS) dataset. Then, the data is preprocessed with modified z-score filtering technique to identify and remove the outliers. After that, Recursive Feature Elimination with Cross-Validation (RFECV) is employed to determine the optimal subset of features by eliminating the irrelevant features through validation. Finally, the proposed BPNN is introduced to predict the enterprises of cashflow values by adjusting the weights and biases for optimal liquidity management. From the results, the proposed BPNN established better results in terms of accuracy, precision, recall, and F1-score with 98.85%, 97.06%, 95.48%, and 99.85% when compared to existing Long Short-Term Memory (LSTM).