The local sensitivity analysis of project portfolio selection problem with divisibility

The data about investment and benefit of every project is often acquired by estimating and forecasting. When external environment changes, inaccuracy of the data will bring great risk to the enterprise decision. So we need to analyze how the range of parameters makes influence on the economic benefits. This will provide an important basis to the enterprise in preventing risk. This paper first presents a new model for project portfolio selection problem with divisibility and parameter variation. On the basis of this model, we also present a new model for project portfolio selection problem with parameter variation without divisibility, and then a numerical example is provided and solved by GAMS/BARON. Similarity, we also calculate sensitivity coefficients on the basis of the solution. After that, we make local sensitivity analysis on several important parameters. We also compare the size of parametric sensitivity when considering the divisibility or not considering it. The research results show that: when different parameters vary the same proportion, the size of each parametric sensitivity coefficient is different. Besides, divisibility makes a great influence on the size of sensitivity coefficient. Sensitivity coefficient can reflect the degree of importance of parameters. This has good practical significance in decision-making and risk prevention for enterprise.