Regresar

Data analysis algorithms for revenue assurance

Abstract:

In companies there are dissimilar factors that influence the loss of income, such as errors in planning, in decision-making, inadequate control of projects and poor treatment of inaccuracy and uncertainty in the data. Some of these causes can be mitigated if anomalous data contained in the information systems of the organizations themselves are analyzed. Present research proposes several algorithms to support revenue assurance in project management organizations. As a novelty, this proposal combines outlier mining, proactive risk management and soft computing techniques. A variety of algorithms for detection of anomalous data in revenue assurance is implemented in a library based on free software. These algorithms apply methods based on data spatial analysis, K-means, Mahalanobis and Euclidean distances, partial clustering with automatic estimation of the clusters number, pattern recognition techniques, heuristics, among others. In the research, cross-validation tests, non-parametric tests and evaluation by a group of experts are carried out. For the application of the proposal in a real environment, databases of finished projects are used and the identification of situations generating anomalous data in project-oriented organizations is achieved. In addition, the proposal has been implemented in a computer tool dedicated to project management with which several companies and centers of development of information technologies have benefited, with more than 300 projects and 5000 users.