Cash Flow Planning And Optimization Through Genetic Algroritms

This article describes an intelligent system for financial planning and cashflow optimization named ICF: Intelligent Cash Flow. ICF is a computational tool for decision support which provides short-term and long-term financial managing strategies, considering financial products of the market. The ICF system makes use of Genetic Algorithms to elaborate cash flow projections which improve the company's profit for a specific period. ICF helps to deal with the complex aspects of cash flow planning: the large number of alternatives to consider, i. e. the mix of investments which offer the higher profit rates over a period; the intensive numerical processing involved; the dynamic changes in the Financial Market (e.g. rates, terms and tax regulations); and the changes in the company's daily financial position. The ICF system integrates two models: the financial and the genetic models. The financial model is used to calculate the cash flow profitability, based on the IDC (Interbank Deposit Certificate), by projecting profits and taxes for each kind of investment, for any term in the considered period. The genetic model, on the other hand, is used to search for cash flow plannings which promote profitability and liquidity. The chromosome of the ICF genetic model consists of n genes. Each gene stands for a day in the considered period and has four fields. The first two identify an investment option and its term; the last two identify a resource taking option and its term. For each analyzed day, only two of these fields are used, which depends whether the operational balance is positive or negative that day.According to the Evolutionary Computation theory, problems such as the optimization of the cash flow are highly epistatic, which means that there is a strong interdependency between genes of the respective representation (for example, the investment on day d depends on the availability of financial resources that day, which can be due to the redemption made on day d-n). Such genes consist of genetic patterns that can be set apart by the crossover operator. In order to deal with the epistasy in this problem, the chromosome has been adapted in way that each gene is represented by its allele and by its locus (position in the chromosome). This kind of representation has the objective to relax the positional rigidity of the genes, increasing the chances of distant interdependent genes to come closer to each other. Thus, genetic patterns with high fitness have more chances to proliferate in forthcoming generations. To manipulate this chromosome structure we have employed an extension of the partially-mapped crossover (PMX) operator proposed by Goldberg, which explores important similarities of value and order simultaneously. The mutation operator applied in the ICF implements a random choice of a gene (day) and the random assignment of a new term and a new type of financial application (investment or loan). The fitness function calculates the liquid returns (profit or tax) of each suggested application/loan for each day in the considered period, projected to the last day of the same. A more satisfactory planning is obtained by finding the maximum return value to this function.ICF has been tested and is currently in use by a Brazilian company. The model manages to find cash flow plannings which present profits, in average, 38% higher during evolution, making evident the importance of a such decision-supporting system. Comparing to the random search, the ICF in average leads to profits 50% higher. Many experiments were made for different periods of the year. The results show that the profitability is obviously affected by the company's operational balance, but it is also strongly influenced by the planning strategy. In this point, the ICF was capable of identifying strategies, with matched operations of application and redemption, which increased the cashbox in days of the flow, in which there was the option of highly profitable investments.