A Markov Process for Modeling Cash Flow and Analyzing of Financial Risk: Case of National Bank of Algeria Agency of Bejaia

This paper explores the use of a Markov process to model the dynamic of reserves dedicated to cash flow and liquidity levels at a National Bank of Algeria agency in Bejaia. To achieve our main objective of analyzing financial risk, we require data on daily deposits, and withdrawals. We model both the daily deposits, and withdrawals using kernel density estimation, a non-parametric method that specifically employs a normal kernel with a logarithmic transformation, while the spreads between them is modeled using a logistic distribution. We evaluate the effectiveness and robustness of the cash flow model data using some performance metrics such as accuracy, mean squared error, mean absolute error, log-likelihood, etc. Cross-validation techniques are applied to ensure consistent and reliable performance across different data subsets. Next, we evaluate the behavior of our model under realistic conditions. This allows for risk prediction of exceeding fund thresholds set by Financial Movements Directorate, which are dynamic throughout the month. To ensure efficient liquidity management, we propose a threshold system that reduces risk based on successive assessments.