Procedure for Creating a Virtual Multibank Agent: Services
For the financial institutions that will be subject to the “sweeping” of information and its subsequent classification, the presence of the intelligent agent will also benefit them in the sense that the number of users who leave the site will be decreased, and hence the migration of their pages, at the same time that these institutions may reach more users given that the agent would be performing much broader searches on the Internet than those that a user would be able to perform individually, except in the case where the user were to invest a considerable amount of time navigating the Internet.
We are aware that the search systems that we establish will not be very exhaustive, even when handing very large databases, given that the Internet grows everyday; although without pursuing high percentages of thoroughness, what we aim to do is ensure that the results are precise and represent the solutions closest to the objective established by the user.
The problem that we are faced with is how a user is going to obtain information about a banking product at various online financial institutions. Each request for information will have an assigned access frequency, and based on that frequency, some possible combinations of dates on which to obtain the information.
This has to do with optimizing the navigation time in the period of time considered at the same time as minimizing the number of routes to follow in order to arrive at the desired information. The tabu search is a heuristic adaptive memory procedure that was first presented in research performed by Glover, and has been gaining wide acceptance in recent decades when it comes to resolving complex problems (continuous and discrete, convex and non-convex, linear and non-linear problems, etc.) in different scientific areas. The tabu algorithm is used for exploring the space of solutions through repeated movements from a solution to the best of its neighbors, trying to avoid the local optimums, which thus enables movements that worsen the solution once arriving at a local optimum [Costa, D. ]. The main attributes of each solution visited are stored in a list and are classified as tabu during subsequent iterations in order to prevent the algorithm from cycling—that is, preventing the solutions from being revisited. Once a movement that generates a new solution is accepted, its inverse movement is added to the tabu list and it remains on this list for a certain number of iterations. Generally, the search ends after a certain number of iterations, or a predetermined amount of navigation time, or when reaching a given number of iterations that do not improve the solution found as the most adequate. The solutions are determined by the times assigned to the navigation and by the routes designed for each access. In the problem of a user navigating through an online banking website, penalty costs due to delays and updating costs have been considered, both of a linear type. payday loans online direct lenders only
We begin our study with the consideration of a maximum time dedicated by the user for navigating through the site in search of financial information. This time is broken down into 5 possible fields: 1 (obtain information about products), 2 (use simulators), 3 (entry of data), 4 (solve problems), 5 (buy product). At time zero, N accesses arrive at the page.