Estimating the performance of Sino-Hong Kong joint ventures using neural network ensembles

Prediction of the performance of international joint ventures remains a rel atively under-researched area, yet its importance is well recognized due to the tremendous surge in joint venture activities in the past decade. Data on 1,463 Sino-Hong Kong joint ventures were gathered and those that appeare d on the Honor Roll of the China Association of Enterprises with Foreign In vestment are identified. Neural network models were used to relate the post erior probability - the probability that a venture gets on the Honor Roll - and the seven economic variables. A simple and yet powerful method called the ensemble method was used to estimate the posterior probability. The res ults indicate that every variable, except one, has influence on the probabi lity of success. Perhaps more important, the results demonstrate that the m odeling approach is able to mine useful information from the data set.